U.S. patent application number 14/310683 was filed with the patent office on 2015-04-30 for mr spectroscopy system and method for diagnosing painful and non-painful intervertebral discs.
The applicant listed for this patent is NOCIMED, LLC, THE REGENTS OF THE UNIVERSITY OF CALIFORNIA, A CALIFORNIA CORPORATION. Invention is credited to John Patrick Claude, Paul Henry Kane, Jeffrey C. Lotz, James Clayton Peacock, III.
Application Number | 20150119688 14/310683 |
Document ID | / |
Family ID | 47744643 |
Filed Date | 2015-04-30 |
United States Patent
Application |
20150119688 |
Kind Code |
A1 |
Peacock, III; James Clayton ;
et al. |
April 30, 2015 |
MR SPECTROSCOPY SYSTEM AND METHOD FOR DIAGNOSING PAINFUL AND
NON-PAINFUL INTERVERTEBRAL DISCS
Abstract
An MR Spectroscopy (MRS) system and approach is provided for
diagnosing painful and non-painful discs in chronic, severe low
back pain patients (DDD-MRS). A DDD-MRS pulse sequence generates
and acquires DDD-MRS spectra within intervertebral disc nuclei for
later signal processing and diagnostic analysis. An interfacing
DDD-MRS signal processor receives output signals of the DDD-MRS
spectra acquired and is configured to optimize signal-to-noise
ratio by an automated system that selectively conducts optimal
channel selection, phase and frequency correction, and frame
editing as appropriate for a given acquisition series. A diagnostic
processor calculates a diagnostic value for the disc based upon a
weighted factor set of criteria that uses MRS data extracted from
the acquired and processed MRS spectra for multiple chemicals that
have been correlated to painful vs. non-painful discs. A display
provides an indication of results for analyzed discs as an overlay
onto a MRI image of the lumbar spine.
Inventors: |
Peacock, III; James Clayton;
(San Carlos, CA) ; Claude; John Patrick; (Redwood
City, CA) ; Kane; Paul Henry; (Albuquerque, NM)
; Lotz; Jeffrey C.; (San Mateo, CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
NOCIMED, LLC
THE REGENTS OF THE UNIVERSITY OF CALIFORNIA, A CALIFORNIA
CORPORATION |
REDWOOD CITY
OAKLAND |
CA
CA |
US
US |
|
|
Family ID: |
47744643 |
Appl. No.: |
14/310683 |
Filed: |
June 20, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13444731 |
Apr 11, 2012 |
8825131 |
|
|
14310683 |
|
|
|
|
PCT/US2010/052737 |
Oct 14, 2010 |
|
|
|
13444731 |
|
|
|
|
12579371 |
Oct 14, 2009 |
8761860 |
|
|
PCT/US2010/052737 |
|
|
|
|
Current U.S.
Class: |
600/410 |
Current CPC
Class: |
A61B 5/748 20130101;
A61B 2576/02 20130101; G01R 33/485 20130101; A61B 5/4514 20130101;
A61B 5/4566 20130101; A61B 5/743 20130101; A61B 5/4824 20130101;
A61B 5/055 20130101 |
Class at
Publication: |
600/410 |
International
Class: |
A61B 5/055 20060101
A61B005/055; A61B 5/00 20060101 A61B005/00 |
Claims
1. An MRS system comprising an MRS pulse sequence, MRS signal
processor, and MRS diagnostic processor, and which is configured to
generate, acquire, and process an MRS spectrum representative of a
region of interest in a body of a patient for providing
diagnostically useful information associated with the region of
interest.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This Application is a continuation of U.S. patent
application Ser. No. 13/444,731, filed Apr. 11, 2012, and titled
"MR SPECTROSCOPY SYSTEM AND METHOD FOR DIAGNOSING PAINFUL AND
NON-PAINFUL INTERVERTEBRAL DISCS," which is a continuation of
International Patent Application No. PCT/US2010/052737, filed on
Oct. 14, 2010, and titled "MR SPECTROSCOPY SYSTEM AND METHOD FOR
DIAGNOSING PAINFUL AND NON-PAINFUL INTERVERTEBRAL DISCS," which
designates the United States, and which is a continuation-in-part
of U.S. patent application Ser. No. 12/579,371, filed Oct. 14,
2009, and titled "MR SPECTROSCOPY SYSTEM AND METHOD FOR DIAGNOSING
PAINFUL AND NON-PAINFUL INTERVERTEBRAL DISCS," each of which is
hereby incorporated by reference in its entirety and made a part of
this specification for all that it discloses.
BACKGROUND OF THE INVENTION
[0002] 1. Field of the Invention
[0003] This disclosure relates to systems, processors, devices, and
methods for measuring chemical constituents in tissue for
diagnosing medical conditions. More specifically, it relates to
systems, pulse sequences, signal and diagnostic processors,
diagnostic displays, and related methods using novel application of
nuclear magnetic resonance, including magnetic resonance
spectroscopy, for diagnosing pain such as low back pain associated
with degenerative disc disease.
[0004] 2. Description of the Related Art
[0005] While significant effort has been directed toward improving
treatments for discogenic back pain, relatively little has been
done to improve the diagnosis of painful discs.
[0006] Magnetic resonance imaging (MRI) is the primary standard of
diagnostic care for back pain. An estimated ten million MRIs are
done each year for spine, which is the single largest category of
all MRIs at an estimated 26% of all MRIs performed. MRI in the
context of back pain is sensitive to changes in disc and endplate
hydration and structural morphology, and often yields clinically
relevant diagnoses such as in setting of spondlyolesthesis and disc
herniations with nerve root impingement (e.g. sciatica). In
particular context of axial back pain, MRI is principally useful
for indicating degree of disc degeneration. However, degree disc
degeneration has not been well correlated to pain. In one regard,
people free of back pain often have disc degeneration profiles
similar to those of people with chronic, severe axial back pain. In
general, not all degenerative discs are painful, and not all
painful discs are degenerative. Accordingly, the structural
information provided by standard MRI exams of the lumbar spine is
not generally useful for differentiating between painful and
non-painful degenerative discs in the region as related to chronic,
severe back pain.
[0007] Accordingly, a second line diagnostic exam called
"provocative discography" (PD) is often performed after MRI exams
in order to localize painful discs. This approach uses a needle
injection of pressurized dye in awake patients in order to
intentionally provoke pain. The patient's subjective reporting of
pain level experienced during the injection, on increasing scale of
0-10, and concordancy to usual sensation of pain, is the primary
diagnostic data used to determine diagnosis as a "positive
discogram"--indicating painful disc--versus a "negative discogram"
for a disc indicating it is not a source of the patient's chronic,
severe back pain. This has significant limitations including
invasiveness, pain, risks of disc damage, subjectivity, lack of
standardization of technique. PD has been particularly challenged
for high "false+" rates alleged in various studies, although recent
developments in the technique and studies related thereto have
alleged improved specificity of above 90%. (Wolfer et al., Pain
Physician 2008; 11:513-538, ISSN 1533-3159). However, the
significant patient morbidity of the needle-based invasive
procedure is non-trivial, as the procedure itself causes severe
pain and further compromises time from work. Furthermore, in
another recent study PD was shown to cause significant adverse
effects to long term disc health, including significantly
accelerating disc degeneration and herniation rates (on the lateral
side of needle puncture). (Carragee et al., SPINE Volume 34, Number
21, pp. 2338-2345, 2009). Controversies around PD remain, and in
many regards are only growing, despite the on-going prevalence of
the invasive, painful, subjective, harmful approach as the
secondary standard of care following MRI. PD is performed an
estimated 400,000 times annually world-wide, at an estimated total
economic cost that exceeds $750 Million Dollars annually. The need
for a non-invasive, painless, objective, non-significant risk, more
efficient and cost-effective test to locate painful intervertebral
discs of chronic, severe low back pain patients is urgent and
growing.
[0008] A non-invasive radiographic technique to accurately
differentiate between discs that are painful and non-painful may
offer significant guidance in directing treatments and developing
an evidence-based approach to the care of patients with lumbar
degenerative disc disease (DDD).
SUMMARY OF THE INVENTION
[0009] One aspect of the present disclosure is a MRS pulse sequence
configured to generate and acquire a diagnostically useful MRS
spectrum from a voxel located principally within an intervertebral
disc of a patient.
[0010] Another aspect of the present disclosure is an MRS signal
processor that is configured to select a sub-set of multiple
channel acquisitions received contemporaneously from multiple
parallel acquisition channels, respectively, of a multi-channel
detector assembly during a repetitive-frame MRS pulse sequence
series conducted on a region of interest within a body of a
subject.
[0011] Another aspect of the present disclosure is an MRS signal
processor comprising a phase shift corrector configured to
recognize and correct phase shifting within a repetitive
multi-frame acquisition series acquired by a multi-channel detector
assembly during an MRS pulse sequence series conducted on a region
of interest within a body of a subject.
[0012] Another aspect of the present disclosure is a MRS signal
processor comprising a frequency shift corrector configured to
recognize and correct frequency shifting between multiple
acquisition frames of a repetitive multi-frame acquisition series
acquired within an acquisition detector channel of a multi-channel
detector assembly during a MRS pulse sequence series conducted on a
region of interest within a body of a subject.
[0013] Another aspect of the present disclosure is a MRS signal
processor comprising a frame editor configured to recognize at
least one poor quality acquisition frame, as determined against at
least one threshold criterion, within an acquisition channel of a
repetitive multi-frame acquisition series received from a
multi-channel detector assembly during a MRS pulse sequence series
conducted on a region of interest within a body of a subject.
[0014] Another aspect of the present disclosure is an MRS signal
processor that comprises an apodizer to reduce the truncation
effect on the sample data. The apodizer can be configured to
apodize an MRS acquisition frame in the time domain otherwise
generated and acquired by via an MRS aspect otherwise herein
disclosed, and/or signal processed by one or more of the various
MRS signal processor aspects also otherwise herein disclosed.
[0015] Another aspect of the present disclosure is an MRS
diagnostic processor configured to process information extracted
from an MRS spectrum for a region of interest in a body of a
subject, and to provide the processed information in a manner that
is useful for diagnosing a medical condition or chemical
environment associated with the region of interest.
[0016] Another aspect of the present disclosure is an MRS system
comprising an MRS pulse sequence, MRS signal processor, and MRS
diagnostic processor, and which is configured to generate, acquire,
and process an MRS spectrum representative of a region of interest
in a body of a patient for providing diagnostically useful
information associated with the region of interest.
[0017] Still further aspects of the present disclosure comprise
various MRS method aspects associated with the other MRS system,
sequence, and processor aspects described above.
[0018] Each of the foregoing aspects, modes, embodiments,
variations, and features noted above, and those noted elsewhere
herein, is considered to represent independent value for beneficial
use, including even if only for the purpose of providing as
available for further combination with others, and whereas their
various combinations and sub-combinations as may be made by one of
ordinary skill based upon a thorough review of this disclosure in
its entirety are further contemplated aspects also of independent
value for beneficial use.
BRIEF DESCRIPTION OF THE DRAWINGS
[0019] These and other features, aspects, and advantages of the
present disclosure will now be described with reference to the
drawings of embodiments, which embodiments are intended to
illustrate and not to limit the disclosure.
[0020] FIGS. 1A-C show respective MRI images of an intervertebral
disc region of a lumbar spine with overlay features representing a
voxel prescription within a disc for performing a DDD-MRS exam
according to one aspect of the disclosure, in coronal, sagittal,
and axial imaging planes, respectively.
[0021] FIG. 2 shows an example of the sectional deployment in one
commercially available MR spine detector coil assembly, and with
which certain aspects of the present disclosure may be configured
to interface for cooperative operation and use, and have been so
configured and used according to certain Examples provided
elsewhere herein.
[0022] FIG. 3A shows an example of a CHESS water suppression pulse
sequence diagram representing certain pulse sequence aspects
contemplated by certain aspects of the present disclosure.
[0023] FIG. 3B shows certain aspects of a combined CHESS-PRESS
pulse sequence diagram also consistent with certain aspects of the
present disclosure.
[0024] FIG. 3C shows various different aspects of a combined
CHESS-VSS-PRESS pulse sequence diagram also illustrative of certain
aspects of the present disclosure.
[0025] FIGS. 4A-B show two examples of respective planar views of a
very selective saturation (VSS) prescription for a voxelated
acquisition series in an intervertebral disc to be conducted via a
DDD-MRS pulse sequence according to further aspects herein.
[0026] FIG. 5 shows Real (Sx) and imaginary (Sy) parts of an FID
(right) that correspond to x and y components of the rotating
magnetic moment M (left).
[0027] FIG. 6 shows an amplitude plot of complex data from a
standard MRS series acquisition of multiple frame repetitions
typically acquired according to certain present embodiments, and
shows amplitude of signal on the y-axis and time on the x-axis.
[0028] FIG. 7 shows a graphical plot of an MRS absorption spectrum
from an MRS pulse sequence acquisition from a lumbar disc using a 3
T MR system, and which is produced from the transform of the
complex data as the output average after combining all of 6
activated acquisition channels and averaging all frames, such as
typically provided in display by a commercially available MRS
system, and is generated without applying the various signal
processing approaches of the present disclosure.
[0029] FIG. 8 shows a graphical display of individual channel MRS
spectra of all uncorrected channels of the same MRS acquisition
featured in FIG. 7, and is shown as "real part squared"
representation of the acquired MRS spectral data prior to combining
the channels, and is also prior to pre-processing according to the
signal processing approaches of the present disclosure.
[0030] FIG. 9A shows a schematic flow diagram of one DDD-MRS
processor configuration and processing flow therein, first
operating in DDD-MRS signal processor mode by conducting optimal
channel (coil) selection, phase correcting, then apodizing, then
transforming domain (from time to frequency), then frame editing
(editing out poor quality frames while retaining higher quality),
then frequency error correction (correcting for frequency shifts),
then averaging of all selected coils, and then followed by a
DDD-MRS diagnostic processor and processing flow that comprises
data extraction related to MRS spectral regions of diagnostic
interest, then applying the diagnostic algorithm, then generating a
diagnostic patient report.
[0031] FIG. 9B shows a schematic flow diagram of further detail of
various component parts of the DDD-MRS signal processor and
respective steps taken thereby as shown more generally in FIG.
9A.
[0032] FIG. 9C shows a schematic flow diagram of further detail of
various component parts of the DDD-MRS diagnostic processor and
processing flow taken thereby as also shown more generally in FIG.
9A.
[0033] FIG. 10 shows a plot of phase angle pre- and post-phase
correction for an acquisition series example, and as is similarly
applied for a DDD-MRS acquisition such as for a disc according to
certain aspects of the present disclosure.
[0034] FIG. 11 shows the serial acquisition frame averages for each
of 6 individual acquisition channels as shown in FIG. 8, but after
phase correction consistent with the signal processing flow shown
in FIGS. 9A-B and phase-correction approach illustrated in FIG.
10.
[0035] FIG. 12 shows the frame-averaged real part squared MRS
spectrum after combining the strongest two channels (channels 1 and
2) selected among the 6 phase-corrected frame-averaged channel
spectra shown in FIG. 11 using a channel selection approach and
criterion according to a further aspect of the current disclosure,
but without frequency correction.
[0036] FIG. 13 shows an example of a time-intensity plot for a
DDD-MRS acquisition similar to that shown in FIG. 17D for the
acquisition shown in FIGS. 7-8 and 11-12, except that the plot of
FIG. 13 relates to another MRS pulse sequence acquisition series of
another lumbar disc in another subject with corrupted frames midway
along the temporal acquisition series in order to illustrate frame
editing according to other aspects of the disclosure.
[0037] FIG. 14A shows confidence in frequency error estimate vs.
MRS frames temporally acquired across an acquisition series for a
disc, as plotted for the DDD-MRS series acquisition shown in
different view in FIG. 13.
[0038] FIG. 14B shows a frame by frame frequency error estimate of
the acquisition series featured in FIG. 14A.
[0039] FIG. 15 shows all 6 frame-averaged acquisition channels for
the series acquisition conducted on the disc featured in FIGS.
13-14B, prior to correction.
[0040] FIG. 16A shows phase corrected, frequency corrected, but not
frame edited spectral average combining all of acquired series
frames for channels 3 and 4 as combined after optimal channel
selection, for the same series acquisition featured in FIGS.
13-15.
[0041] FIG. 16B shows phase corrected, frequency corrected, and
frame edited spectral average combining the partial retained frames
not edited out from the acquired series for channels 3 and 4 as
combined after optimal channel selection, also for the same series
acquisition featured in FIGS. 13-15.
[0042] FIG. 17A shows a 2-dimensional time-intensity plot similar
to that shown in FIG. 13, but for yet another DDD-MRS acquisition
series of another disc in another subject and to illustrate another
mode of frame editing aspects of the present disclosure.
[0043] FIG. 17B shows a waterfall plot in 3-dimensions for the
DDD-MRS acquisition series shown in FIG. 17A, and shows the
chemical shift spectrum as a running cumulative average at discrete
points over time of serial frames acquired, with spectral amplitude
on the vertical axis.
[0044] FIG. 17C shows an average DDD-MRS spectrum across the full
acquisition series shown in FIGS. 17A-B, without frame editing, and
plots both phase only and phase+frequency corrected formats of the
spectrum.
[0045] FIG. 17D shows a 2-dimensional time-intensity plot similar
to that shown and for the same DDD-MRS acquisition series of FIG.
17A, but only reflecting retained frames after editing out other
frames according to the present aspect of the disclosure and
referenced to FIGS. 17A-C.
[0046] FIG. 17E shows a similar waterfall plot of cumulative
spectral averages and for the same DDD-MRS acquisition series shown
in FIG. 17B, but according to only the retained frames after frame
editing as shown in FIG. 17D.
[0047] FIG. 17F shows a similar average DDD-MRS spectrum and for
the same acquisition series shown in FIG. 17C, but only for the
retained frames after frame editing as shown in various modes in
FIGS. 17D-E.
[0048] FIGS. 18A-B show time-intensity plots of the same MRS series
acquisition for the same disc featured in FIGS. 7-8 and 11-12 as
pre- (FIG. 18A) and post- (FIG. 18B) frequency correction according
to a further aspect of the present disclosure, and shows each
acquisition frame as a horizontal line along a horizontal frequency
range with brightness indicating signal amplitude (bright white
indicating higher amplitude, darker indicating lower), and shows
the series of related repetitive frames in temporal relationship
stacked from top to bottom, e.g. top is time zero).
[0049] FIGS. 19A-B show the same respective time-intensity plots
shown in FIG. 19A (pre-) and FIG. 19B (post-) frequency correction,
but in enhanced contrast format.
[0050] FIG. 20 shows spectral plots for 6 frame-averaged
acquisition channels for the same acquisition shown in FIGS. 7-8
and 11-12, except post phase and frequency correction and prior to
optimal channel selection and/or combination channel averaging.
[0051] FIG. 21 shows a spectral plot for phase and frequency error
corrected channels 1 and 2 selected from FIG. 20 as averaged,
according to a further aspect of the disclosure.
[0052] FIG. 22 shows a bar graph of mean values, with standard
deviation error bars, of Visual Analog Scale (VAS) and Oswestry
Disability Index (ODI) pain scores calculated for certain of the
pain patients and asymptomatic volunteers evaluated in a clinical
study of Example 1 and conducted using certain physical embodiments
of a diagnostic system constructed according to various aspects of
the present disclosure.
[0053] FIG. 23 shows a Receiver Operator Characteristic (ROC) curve
representing the diagnostic results of the DDD-MRS diagnostic
system used in the clinical study of Example 1 with human subjects
featured in part in FIG. 22, as compared against standard control
diagnostic measures for presumed true diagnostic results for
painful vs. non-painful discs.
[0054] FIG. 24 shows a partition analysis plot for
cross-correlation of a portion of the clinical diagnostic results
of the DDD-MRS system under the same clinical study of Example 1
and also addressed in FIGS. 22-23, based on partitioning of the
data at various limits attributed to different weighted factors
used in the DDD-MRS diagnostic processor, with "x" data point plots
for negative control discs and "o" data point plots for positive
control discs, also shows certain statistical results including
correlation coefficient (R.sup.2).
[0055] FIG. 25A shows a scatter plot histogram of DDD-MRS
diagnostic results for each disc evaluated in the clinical study of
Example 1 and also addressed in FIGS. 22-24, and shows the DDD-MRS
results separately for positive control (PC) discs (positive on
provocative discography or "PD+"), negative control (NC) discs
(negative on provocative discography or "PD-", plus discs from
asymptomatic volunteers or "ASY"), PD- alone, and ASY alone.
[0056] FIG. 25B shows a bar graph of the same DDD-MRS diagnostic
results shown in FIG. 25A across the same subject groups of Example
1, but shows the mean values with standard deviation error bars for
the data.
[0057] FIG. 26 shows a bar graph of presumed true and false binary
"positive" and "negative" diagnostic results produced by the
DDD-MRS system for painful and non-painful disc diagnoses in the
clinical study of Example 1, as compared against standard control
diagnostic measures across the positive controls, negative controls
(including sub-groups), and all discs evaluated in total in the
study.
[0058] FIG. 27 shows diagnostic performance measures of
Sensitivity, Specificity, Positive Predictive Value (PPV), Negative
Predictive Value (NPV), and area under the curve (AUC) which in
this case is equivalent to Global Performance Accuracy (GPA) for
the DDD-MRS diagnostic results in the clinical study of Example
1.
[0059] FIG. 28 shows a bar graph comparing areas under the curve
(AUC) per ROC analysis of MRI alone (for prostate cancer
diagnosis), MRI+PROSE (MRS package for prostate cancer diagnosis),
MRI alone (for discogenic back pain or DDD pain), and MRI+DDD-MRS
(for discogenic back pain or DDD pain), with bold arrows showing
relative impact of PROSE vs. DDD-MRS on AUC vs. MRI alone for the
respective different applications and indications, with DDD-MRS
results shown as provided under Example 1.
[0060] FIG. 29 shows positive predictive value (PPV) and negative
predictive value (NPV) for MRI alone and for MRI+DDD-MRS (per
Example 1 results), both as applied for diagnosing DDD pain, vs.
standard control measures such as provocative discography.
[0061] FIG. 30A shows a plot of DDD-MRS algorithm output data for a
series of 8 L4-L5 lumbar discs in 8 asymptomatic human control
subjects per clinically acquired and processed DDD-MRS exam under
Example 1, and plots these results twice for each disc on first (1)
and second (2) separate repeat scan dates in order to demonstrate
repeatability of the DDD-MRS exam's diagnostic results.
[0062] FIG. 30B shows a plot of PG/LAAL ratio data for 3 discs per
DDD-MRS pulse sequence and signal processing data of Example 1, and
shows the clinically acquired results via 3 T DDD-MRS exams of the
discs in vivo in pain patients (y-axis) against acquired
measurements for the same chemicals in the same disc material but
flash frozen after surgical removal and using 11 T HR-MAS
spectroscopy.
[0063] FIG. 31A shows a digitized post-processed DDD-MRS spectrum
(in phase real power) as processed according to certain of the MRS
signal processor aspects of the present disclosure, and certain
calculated data derived therefrom as developed and used for
calculated signal-to-noise ratio (SNR) of the processed result, as
taken across a sub-set of samples evaluated under Example 1.
[0064] FIG. 31B shows a digitized pre-processed DDD-MRS spectrum
(absorption) as 6 channel spectral average without deploying the
MRS signal processing aspects of the present disclosure (e.g.
"pre-processing"), and certain calculated data derived therefrom as
developed and used for calculated signal-to-noise ratio (SNR) of
the processed result.
[0065] FIG. 31C shows a scatter plot histogram of signal-to-noise
ratio (SNR) for standard "all channels, non-corrected" frame
averaged MRS spectra (absorption) produced by the 3 T MR system for
a subset of discs evaluated using the DDD-MRS pulse sequence in the
clinical study of Example 1, and the SNR of MRS spectra (in phase
real power) for the same series acquisitions for the same discs
post-processed by the DDD-MRS signal processor configured according
to various of the present aspects of this disclosure, as such SNR
data was derived for example as illustrated in FIGS. 31A-B.
[0066] FIG. 31D shows the same data shown in FIG. 31C, but as bar
graph showing mean values and standard deviation error bars for the
data within each pre-processed and post-processed groups.
[0067] FIG. 31E shows a scatter plot histogram of the ratio of SNR
values calculated post- versus pre-processing for each of the discs
per the SNR data shown in FIGS. 31C-D.
[0068] FIG. 31F shows a bar graph of mean value and standard
deviation error bar of the absolute difference between post- and
pre-processed SNR values for each of the discs shown in different
views in FIGS. 31C-E.
[0069] FIGS. 31G-H respectively show the mean and standard
deviation for absolute improvement between pre- and post- processed
SNR (FIG. 31F), the mean ratio improvement of
post-processed/pre-processed SNR (FIG. 31G), and the mean
improvement of post-processed vs. pre-processed SNR (FIG. 31H).
[0070] FIG. 32A shows a mid-sagittal T2-weighted MRI image of a
patient evaluated under the clinical study of Example 1 and
comparing the diagnostic results of the operating embodiment for
DDD-MRS system developed according to various aspects herein
against provocative discography results for the same discs, and
shows a number-coded (and also may be color coded) diagnostic
legend for the DDD-MRS results (on left of image) and discogram
legend (top right on image) with overlay of the DDD-MRS results and
discogram results on discs evaluated in the patient.
[0071] FIG. 32B shows a mid-sagittal T2-weighted MRI image of
another patient evaluated under the clinical study of Example 1 and
comparing the diagnostic results of the physical embodiment DDD-MRS
system developed according to various aspects herein against
provocative discography results for the same discs, and shows a
number-coded (and also may be color coded) diagnostic legend for
the DDD-MRS results (on left of image) and discogram legend (top
right on image) with overlay of the DDD-MRS results and discogram
results on discs evaluated in the patient.
[0072] FIG. 33A shows a scatter plot histogram plot of DDD-MRS (or
"Nociscan") diagnostic results against control groups for various
discs evaluated in vivo according to the data set reviewed and
processed under Example 2, as similarly shown for the data
evaluated in Example 1 in FIG. 25A (plus the further addition of
certain additional information further provided as overlay to the
graph and related to another aspect of data analysis applied
according to further aspects of the present disclosure under
Example 2).
[0073] FIG. 33B shows another scatter plot histogram of another
processed form of the DDD-MRS diagnostic results also shown in FIG.
33A and per Example 2, after transformation of the DDD-MRS
diagnostic algorithm results for the discs into "% probability
painful" assigned to each disc as distributed across the positive
(POS) and negative (NEG) control group sub-populations shown.
[0074] FIG. 34A shows a scatter plot histogram of signal-to-noise
ratio (SNR) for standard "all channels, non-corrected" frame
averaged MRS spectra (absorption) produced by the 3 T MR system for
a subset of discs evaluated using the DDD-MRS pulse sequence and
signal processor in the clinical study of Example 2, and the SNR of
spectra (absorption) for the same series acquisitions for the same
discs post-processed by the DDD-MRS processor, as such SNR data was
derived for example as illustrated in FIGS. 31A-B.
[0075] FIG. 34B shows the same data shown in FIG. 34A, but as bar
graph showing mean values and standard deviation error bars for the
data within each pre-processed and post-processed groups.
[0076] FIG. 34C shows a scatter plot histogram of the ratio of SNR
values calculated post- versus pre-processing for each discs per
the SNR data shown in FIGS. 34A-B.
[0077] FIG. 34D shows a bar graph of mean value and standard
deviation error bar of the absolute difference between post- and
pre-processed SNR values for each of the discs shown in different
views in FIGS. 34A-C.
[0078] FIG. 34E shows a bar graph of mean value and standard
deviation error bar of the ratio of post- to pre-processed SNR
values for each of the discs shown in different views in FIGS.
34A-D.
[0079] FIG. 34F shows a bar graph of the mean value and standard
deviation error bar for the percent increase in SNR from pre- to
post-processed MRS spectra for each of the discs further featured
in FIGS. 34A-E.
[0080] FIG. 35 shows a DDD-MRS spectrum illustrative of a perceived
potential lipid signal contribution as overlaps with the regions
otherwise also associated with lactic acid or lactate (LA) and
alanine (AL), according to further aspects of the present
disclosure and as relates to Example 3.
[0081] FIG. 36 shows a scatter plot histogram of DDD-MRS diagnostic
algorithm results for the test population of in vivo discs, as
calculated for a defined Group A evaluated for diagnostic purposes
via Formula A, under the Example 3.
[0082] FIGS. 37A-C show scatter plot histogram of certain
embodiments for the DDD-MRS diagnostic processor for discs
designated as Group B under Example 3, including as shown with
respect to PG/LAAL ratio results for the discs (FIG. 17A), logistic
regression generated Formula B results for the discs (FIG. 17B),
and the transformed % probability pain distribution for the same
Group B discs as a result of the results in FIG. 17B (FIG.
17C).
[0083] FIG. 38 shows a scatter plot histogram of certain
embodiments for DDD-MRS diagnostic processor for discs designated
as Group C discs under Example 3, after applying logistic
regression generated Formula C to the DDD-MRS spectral data
acquired for the group of discs.
[0084] FIG. 39 shows a scatter plot histogram of another embodiment
for DDD-MRS diagnostic algorithm, as applied to Group C discs under
Example 3 according to a Formula B "hybrid" illustrative of yet a
further embodiment of the present disclosure.
[0085] FIGS. 40A-B show MRI images of two lumbar spine phantoms
according to another Example 4 of the disclosure.
[0086] FIGS. 40C-D show graphical plots of n-acetyl (NAA) and
lactic acid (LA) concentrations in discs from phantoms shown in
FIGS. 40A-B as measured according to certain DDD-MRS aspects of the
present disclosure, versus known amounts, per Example 4.
[0087] FIGS. 41A-B show schematic flow diagrams of a DDD-MRS exam,
including DDD-MRS pulse sequence, DDD-MRS signal processing, and
DDD-MRS algorithm processing, and various data communication
aspects, according to certain further aspects of the present
disclosure.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0088] Previously reported lab experiments used 11 T HR-MAS
Spectroscopy to compare chemical signatures of different types of
ex vivo disc nuclei removed at surgery. (Keshari et al., SPINE
2008) These studies demonstrated that certain chemicals in disc
nuclei, e.g. lactic acid (LA) and proteoglycan (PG), may provide
spectroscopically quantifiable metabolic markers for discogenic
back pain. This is consistent with other studies that suggest DDD
pain is associated with poor disc nutrition, anaerobic metabolism,
lactic acid production (e.g. rising acidity), extracellular matrix
degradation (e.g. reducing proteoglycan), and increased enervation
in the painful disc nuclei. In many clinical contexts, ischemia and
lowered pH cause pain, likely by provoking acid-sensing ion
channels in nociceptor sensory neurons.
[0089] The previous disclosures evaluating surgically removed disc
samples ex vivo with magnetic resonance spectroscopy (MRS) in a
laboratory setting is quite encouraging for providing useful
diagnostic tool based on MRS. However, an urgent need remains for a
reliable system and approach for acquiring MRS signatures of the
chemical composition of the intervertebral discs in vivo in a
readily adoptable clinical environment, and to provide a useful,
clinically relevant diagnostic tool based on these acquired MRS
signatures for accurately diagnosing discogenic back pain. A
significant need would be met by replacing PD with an alternative
that, even if diagnostically equivalent, overcomes one or more of
the significant shortcomings of the PD procedure by being
non-invasive, objective, pain-free, risk-free, and/or more
cost-effective. Magnetic resonance spectroscopy (MRS) is a medical
diagnostic platform that has been previously developed and
characterized for a number of applications in medicine. Some of
these have been approved such as for example for brain tumors,
breast cancer, and prostate cancer. Some MRS platforms disclosed
have been multi-voxel, and others single voxel. None of these have
been adequately configured or developed for in vivo clinical
application to reliably diagnose medical conditions or chemical
environments associated with nociceptive pain, and/or with respect
to intervertebral discs such as may be associated with disc
degeneration and/or discogenic back pain (including in particular,
but without limitation, with respect to the lumbar spine).
[0090] Various technical approaches have also been alleged to
enhance the quality of MRS acquisitions for certain purposes.
However, these approaches are not considered generally sufficient
to provide the desired spectra of robust, reliable utility for many
intervertebral discs in vivo, at least not at field strengths
typically employed for in vivo spectroscopy, e.g. from about 1.2
tesla (T) or about 1.5 T to about 3.0 T or even up to about 7 T.
Furthermore, while individual techniques have been disclosed for
certain operations that might be conducted in processing a given
signal for potentially improved signal:noise ratio (SNR), an MRS
signal processor employing multiple steps providing significant MRS
signal quality enhancement, in particular with respect to improved
SNR for multi-channel single voxel pulse sequence acquisitions,
have yet to be sufficiently automated to provide robust utility for
efficient, mainstream clinical use, such as in primary radiological
imaging centers without sophisticated MR spectroscopists required
to process and interpret MRS data. This is believed to be generally
the case as a shortcoming for many such in vivo MRS exams in
general. Such shortcomings have also been observed in particular
relation to the unique challenge of providing a robust MRS
diagnostic system for diagnosing medical conditions or otherwise
chemical environments within relatively small voxels, areas of high
susceptibility artifact potential, and in particular with respect
to unique challenges of performing MRS in voxels within
intervertebral discs (including with further particularity,
although without necessary limitation, of the lumbar spine). In
solving many of these challenges according to certain aspects of
the present disclosure, such as those providing particular utility
for diagnosing discogenic low back pain and/or chemical
environments within discs, additional beneficial advances have also
been made that are also considered more broadly applicable to MRS
in general, and as may become adapted for many specific
applications, as are also herein disclosed.
[0091] Certain aspects of the current disclosure therefore relate
to new and improved system approaches, techniques, processors, and
methods for conducting in vivo clinical magnetic resonance
spectroscopy (MRS) on human intervertebral discs, in particular
according to a highly beneficial mode of this disclosure for using
acquired MRS information to diagnose painful and/or non-painful
discs associated with chronic, severe axial lumbar (or "low") back
pain associated with degenerated disc disease (or "DDD pain"). For
purpose of helpful clarity in this disclosure, the current aspects,
modes, embodiments, variations, and features disclosed with
particular benefits for this purposed are generally assigned the
label "DDD-MRS." However, other descriptors may be used
interchangeably as would be apparent to one of ordinary skill in
context of the overall disclosure. It is also further contemplated
within the scope of this present disclosure that, while this
disclosure is considered to provide particular benefit for use
involving such human intervertebral discs (and related medical
indications and purposes), the novel approaches herein described
are also considered more broadly and applicable to other regions of
interest and tissues within the body of a subject, and various
medical indications and purposes. For purpose of illustration, such
other regions and purposes may include, without limitation: brain,
breast, heart, prostate, GI tract, tumors, degeneration and/or
pain, inflammation, neurologic disorders, alzheimers, etc.
[0092] Various aspects of this disclosure relate to highly
beneficial advances in each of three aspects, and their various
combinations, useful in particular for conducting a DDD-MRS exam:
(1) MRS pulse sequence for generating and acquiring robust MRS
spectra; (2) signal processor configured to improve signal-to-noise
ratio (SNR) of the acquired MRS spectra; and (3) diagnostic
processor configured to use information from the acquired and
processed MRS spectra for diagnosing painful and/or non-painful
discs on which the MRS exam is conducted in a DDD pain patient.
[0093] Several configurations and techniques related to the DDD-MRS
pulse sequence and signal processor have been created, developed,
and evaluated for conducting 3 T (or other suitable field strength)
MRS on human intervertebral discs for diagnosing DDD pain. A novel
"DDD" MRS pulse sequence was developed and evaluated for this
purpose, and with certain parameters specifically configured to
allow robust application of the signal processor for optimal
processed final signals in a cooperative relationship between the
pulse sequence and post-signal processing conducted. These
approaches can be used, for example, with a 3 Tesla (3 T) "Signa"
MR system commercially available from General Electric (GE). Highly
beneficial results have been observed using the current disclosed
application technologies on this particular MR platform, as has
been demonstrated for illustration according to Examples provided
herein, and it is to be appreciated that applying the present
aspects of this present disclosure in combination with this one
system alone is considered to propose significant benefit to pain
management in patients requiring diagnosis. Accordingly, various
aspects of the present disclosure are described by way of specific
reference to configurations and/or modes of operation adapted for
compatible use with this specific MR system, and related
interfacing components such as spine detector coils, in order to
provide a thorough understanding of the disclosure. It is to be
appreciated, however, that this is done for purpose of providing
useful examples, and though significant benefits are contemplated
per such specific example applications to that system, this is not
intended to be necessarily so limited and with broader scope
contemplated. The current disclosure contemplates these aspects
broadly applicable according to one of ordinary skill to a variety
of MR platforms commercially available that may be different
suitable field strengths or that may be developed by various
different manufacturers, and as may be suitably adapted or modified
to become compatible for use with such different systems by one of
ordinary skill (with sufficient access to operating controls of
such system to achieve this). Various novel and beneficial aspects
of this present disclosure are thus described herein, as provided
in certain regards under the Examples also herein disclosed.
[0094] A DDD-MRS sequence exam is conducted according to one
example overview description as follows. A single three dimensional
"voxel," typically a rectangular volume, is prescribed by an
operator at a control consul, using 3 imaging planes (mid-sagittal,
coronal, axial) to define the "region of interest" (ROI) in the
patient's body, such as shown in FIGS. 1A-C, for MR excitation by
the magnet and data acquisition by the acquisition channel/coils
designated for the lumbar spine exam within the spine detector coil
assembly. The DDD-MRS pulse sequence applies a pulsing magnetic and
radiofrequency to the ROI, which causes single proton combinations
in various chemicals within the ROI to resonate at different
"signature resonant frequencies" across a range. The amplitudes of
frequencies at various locations along this range are plotted along
a curve as the MRS "spectrum" for the ROI. This is done iteratively
across multiple acquisitions for a given ROI, typically
representing over 50 acquisitions, often 100 or more acquisitions,
and often between about 200 and about 600 acquisitions, such as
between 300 and 400 acquisitions for a given exam of a ROI. One
acquisition spectrum among these iterations is called a "frame" for
purpose of this disclosure, though other terms may be used as would
be apparent to one of ordinary skill. These multiple acquisitions
are conducted in order to average their respective acquired
spectra/frames to reduce the amplitudes of acquired signal
components representing noise (typically more random or
"incoherent" and thus reduced by averaging) while better
maintaining the amplitudes of signal components representing target
resonant chemical frequencies of diagnostic interest in the ROI
(typically repeatable and more "coherent" and thus not reduced by
averaging). By reducing noise while maintaining true target signal,
or at least resulting in less relative signal reduction, this
multiple serial frame averaging process is thus conducted for the
primary objective to increase SNR. These acquisitions are also
conducted at various acquisition channels selected at the detector
coils, such as for example 6 channels corresponding with the lumbar
spine area of the coil assembly used in the Examples (where for
example 2 coils may be combined for each channel).
[0095] The 3 T MRI Signa system ("Signa" or "3 T Signa"), in
standard operation conducting one beneficial mode of DDD-MRS
sequence evaluated (e.g. Examples provided herein), is believed to
be configured to average all acquired frames across all acquisition
channels to produce a single averaged MRS curve for the ROI. This
unmodified approach has been observed, including according to the
various Figures and Examples provided herein, to provide a
relatively low signal/noise ratio, with low confidence in many
results regarding data extraction at spectral regions of diagnostic
interest, such as for example and in particular regions associated
with proteoglycan or "PG" (n-acetyl) and lactate or lactic acid
(LA). Sources of potential error and noise inherent in this
imbedded signal acquisition and processing configuration of the
typical MR system, for example were observed in conducting the
DDD-MRS pulse sequence such as according to the Examples. These
various sources of potential error or signal-to-noise ratio (SNR)
compromise were determined to be mostly correctable--either by
altering certain structures or protocols of coil, sequence, or data
acquisition, or in post-processing of otherwise standard protocols
and structures used. Among these approaches, various
post-acquisition signal processing approaches were developed and
observed to produce significantly improved and highly favorable
results using otherwise un-modified operation pre-processing. In
particular, various improvements developed and applied under the
current post-signal processor disclosed herein have been observed
to significantly improve signal quality and SNR.
[0096] Certain such improvements advanced under the post-signal
processor configurations disclosed herein include embodiments
related to the following: (1) acquisition channel selection; (2)
phase error correction; (3) frequency error correction; (4) frame
editing; and (5) apodization. These modules or steps are typically
followed by channel averaging to produce one resulting "processed"
MRS spectrum, when multiple channels are retained throughout the
processing (though often only one channel may be retained). These
may also be conducted in various different respective orders,
though as is elsewhere further developed frame editing will
typically precede frequency error correction. For illustration, one
particular order of these operations employed for producing the
results illustrated in the Examples disclosed herein are provided
as follows: (1) acquisition channel selection; (2) phase
correction; (3) apodization; (4) frame editing; (5) frequency
correction; and (6) averaging.
[0097] While any one of these signal processing operations is
considered highly beneficial, their combination has been observed
to provide significantly advantageous results, and various
sub-combinations between them may also be made for beneficial use
and are also contemplated. Various illustrative examples are
elsewhere provided herein to illustrate sources of error or "noise"
observed, and corrections employed to improve signal quality.
Strong signals typically associated with normal healthy discs were
evaluated first to assess the signal processing approach. Signals
from the Signa that were considered more "challenged" for robust
data processing and diagnostic use were evaluated for further
development to evaluate if more robust metabolite signal can be
elicited from otherwise originally poor SNR signals from the
Signa.
[0098] Additional description further developing these aspects
according to additional embodiments, and other aspects, is provided
below.
[0099] Spine Detector Coil and Patient Positioning
[0100] A typical DDD-MRS exam according to the present embodiments
will be conducted in an MR scanner in which the patient lies still
in a supine position with a spine detector coil underneath the
patient's back and including the lower spine. While this scanner
applies the magnetic and RF fields to the subject, the spine
detector coil functions as an antenna to acquire signals from
resonating molecules in the body. The primary source of MRS signals
obtained from a Signa 3 T MR scanner, according to the physical
embodiments developed and evaluated in the Examples herein this
disclosure, are from the GE HD CTL 456 Spine Coil. This is a
"receive-only" coil with sixteen coils configured into eight
channels. Each channel contains a loop and saddle coil, and the
channels are paired into sections. For lumbar (and thoracic) spine
coverage, such as associated with lumbar DDD pain diagnosis,
sections 4, 5, and 6 are typically deployed to provide six
individual channel signals, as shown for example in FIG. 2.
[0101] Defining the Voxel (Voxel Prescription)
[0102] Certain embodiments of this disclosure relates principally
to "single voxel" MRS, where a single three dimensional region of
interest (ROI) is defined as a "voxel" (VOlumetric piXEL) for MRS
excitation and data acquisition. The spectroscopic voxel is
selected based on T2-weighted high-resolution spine images acquired
in the sagittal, coronal, and axial planes, as shown for example in
FIGS. 1A-C. The patient is placed into the scanner in a supine
position, head first. The axial spine images acquired are often in
a plane oriented with disc angle (e.g. may be oblique) in order to
better encompass the disc of interest. This voxel is prescribed
within a disc nucleus for purpose of using acquired MRS spectral
data to diagnose DDD pain, according to the present preferred
embodiments. In general for DDD-MRS applications evaluating disc
nucleus chemical constituents, the objective for voxel prescription
is to capture as much of the nuclear volume as possible (e.g.
maximizing magnitude of relevant chemical signals acquired), while
restricting the voxel borders from capturing therewithin structures
of the outer annulus or bordering vertebral body end-plates (the
latter being a more significant consideration, where lipid
contribution may be captured and may shroud chemical spectral
regions of interest such as lactate or alanine, as further
developed elsewhere herein). In fact, the actual operation may not
exactly coincide with acquiring signal from only within the voxel,
and may include some bordering region contribution. Thus some
degree of spacing between the borders and these structures is often
desired. These typical objectives may be more difficult to achieve
for some disc anatomies than others, e.g. relatively obliquely
angled discs. For example, L5-S1 may be particularly challenging
because in some patients it can frequently be highly angulated,
irregularly shaped, and collapsed as to disc height.
[0103] In certain voxel prescriptions, the thickness is limited by
the scanner's ability to generate the magnetic gradient that
defines the Z-axis (axial plane) dimension. For example, a minimum
thickness limit is pre-set to 4 mm on the GE Signa 3 T. While such
pre-set limits of interfacing, cooperative equipment and related
software may result in limits on the current application's ability
to function in that environment outside of these limits, the broad
aspects of the current disclosure should not be considered
necessarily so limited in all cases, and functionality may flourish
within other operating ranges perhaps than those specifically
indicated as examples herein, such as in cases where such other
imparted limitations may be released.
[0104] These usual objectives and potential limitations in mind,
typical voxel dimensions and volumes (Z-axis, X-axis, Y-axis, Vol)
may be for example 5 mm (thick) by 14 mm (width) by 16 mm (length),
and 1.12 cc, though one may vary any or all of these dimensions by
operator prescription to suit a particular anatomy or intended
application. The Z-axis dimension is typically limited maximally by
disc height (in order to exclude the end-plates, described further
herein), and minimally by either the set minimum limitations of the
particular MR scanner and/or per SAR safety considerations, in many
disc applications (such as specific indication for pain diagnosis
or other assessment of disc chemistry described herein). This
Z-axis dimension will typically be about 3 mm to about 6 mm
(thick), more typically between about 4 mm to about 6 mm, and most
typically will be suitable (and may be required to be, per anatomy)
between about 4 mm to about 5 mm. The other dimensions are
typically larger across the disc's plane, and may be for example
between about 15 mm to about 20 mm (width and/or length), as have
been observed suitable ranges for most observed cases (e.g. per the
Examples herein). While the higher dimension of these ranges is
typically limited only by bordering tissues desirable to exclude,
the opportunity for patient motion to alter the relative location
of the target voxel relative to actual anatomy may dictate some
degree of "spacing" from such bordering structures to ensure
exclusion. The smaller dimensions of the ranges are more related to
degraded signal quality that comes with excessively small voxel
volume, whereas signal amplitude will typically be directly related
to voxel dimension and volume. Accordingly, voxels within discs
will generally provide robust results, at least with respect to
signal quality, at volumes of at least about 0.5 cc, and in many
cases at least about 0.75 cc or 1 cc. This typically will be
limited by bordering anatomy to up to about 2 ccs, or in some less
typical cases up to about 3 ccs for exceptionally large discs.
These voxel volume ranges will typically be achieved with various
combinations of the typical axis dimensions as also stated
above.
[0105] Also according to the typical voxel prescription objectives
and limitations stated above, an initial prescription may not be
appropriate for achieving acceptable results, though this may not
be known until a sequence is begun to allow observation of acquired
signal quality. Accordingly, further aspects of the present
disclosure contemplate a voxel prescription protocol which
prescribes a first prescription, monitors results (either during
scan or after completion, or via a "pre-scan" routine for this
purpose), and if a lipid signature or other suspected signal
degradation from expected results is observed, re-prescribe the
voxel to avoid suspected source of contaminant (e.g. make the voxel
smaller or adjust its dimensions, tilt, or location) and re-run an
additional DDD-MRS acquisition series (retaining the signal
considered more robust and with least suspected signal degradation
suspected to be voxel error). According to still a further mode, a
pre-set protocol for re-prescribing in such circumstances may
define when to accept the result vs. continue re-trying. In one
embodiment, the voxel may be re-prescribed and acquisition series
re-run once, or perhaps twice, and then the best result is to be
accepted. It is to be appreciated, as with many technology
platforms, that operator training and techniques in performing such
user-dependent operations may be relevant to results, and optimal
(or conversely sub-optimal) results may track skill levels and
techniques used.
[0106] To further illustrate this current aspect of the present
disclosure, the example of a single voxel prescription according to
the typical three planar slice images is shown in FIGS. 1A-1C as
follows. More specifically, FIG. 1A shows a coronal view oriented
aspect of the voxel prescription. FIG. 1B shows a sagittal view
oriented aspect of the voxel prescription. FIG. 1C shows an axial
view oriented aspect of the voxel prescription.
[0107] The "DDD" MRS Pulse Sequence--PRESS
[0108] The DDD-MRS pulse sequence according to one embodiment
shares certain similarities, though with certain differences and
modifications defined herein, with another MRS pulse sequence
called "PROSE". PROSE is primarily intended for use for diagnosing
prostate cancer, and is approved for use and sale and available
from GE on 1.5 T GE MR systems. The DDD-MRS pulse sequence of the
present embodiments, and PROSE for further reference, employ a
sequence approach called Point RESolved Spectroscopy (PRESS). This
involves a double spin echo sequence that uses a 90.degree.
excitation pulse with two 180.degree. slice selective refocusing
radio frequency (RF) pulses, combined with 3D chemical shift
imaging (CSI) phase encoding gradients to generate 3-D arrays of
spectral data or chemical shift images. Due to the small size,
irregular shape, and the high magnetic susceptibility present when
doing disc spectroscopy for DDD pain, the 3D phase encoding option
available under PROSE is not an approach typically to be utilized
under the current disclosed version of DDD-MRS sequence, and single
voxel spectra are acquired by this version embodiment of DDD-MRS
pulse sequence. This unique relative configuration for the DDD-MRS
pulse sequence can be accomplished by setting the user control
variables (CVs) for the matrix acquisition size of each axis to 1
(e.g., in the event the option for other setting is made
available). Further aspects of pulse sequence approaches
contemplated are disclosed elsewhere herein. It is to be
appreciated that while the modified PRESS approach herein described
is particularly beneficial, other approaches may be taken for the
pulse sequence according to one of ordinary skill consistent with
other aspects and objectives herein described and without departing
from the broad aspects of intended scope herein.
[0109] Water and Lipid Signal Suppression--CHESS
[0110] In another sequence called "PROBE" also commercially
available from GE, and which is a CSI sequence used for brain
spectroscopy, the lipid/fat signals are believed to be resolved
through the use of long TE (144 ms) periods and 2 dimensional
transformations (2DJ). These acquisition and signal processing
techniques are believed to be facilitated by the large voxel
volumes prescribed in the brain as well as the homogeneity of the
brain tissue resulting in relatively narrow spectral line widths.
In the prostate region targeted by the different pulse sequence of
PROSE, however, the voxel prescriptions are much smaller and it is
often impossible to place the voxel so as to assuredly exclude
tissues that contain lipid/fat. Therefore, two water and lipid
suppression approaches are available and may be used, if warranted,
in the PROSE sequence: "BASING" and "SSRF" (Spectral Spatial Radio
Frequency). An even more challenging environment of bordering lipid
and reduced homogeneity has been observed with the current DDD pain
application of the lumbar intervertebral discs where the current
ROI within disc nuclei are closely bordered by vertebral bodies
with bone marrow rich in lipid content. However, due both to the
desire to use short TE times (e.g. 28 ms) for the current DDD pain
application in lumbar spine, and the desire to observe MRS
signatures of other chemicals within disc nuclei that may overlap
with lipid signal contribution along the relevant DDD-MRS spectrum,
these water/lipid suppression approaches as developed for brain and
prostate application are not necessarily optimized for DDD-MRS
application in many circumstances. While a SSRF suppression
approach for lipid resonances may be employed in the DDD-MRS
sequence, the narrow band RF pulse required for this may require a
long RF period and amplitude that will exceed the SAR level for
many MR systems.
[0111] Water suppression is also provided by a CHESS sequence
interleaved or otherwise combined in some manner with the PRESS
sequence in order to provide appropriate results. Optimization of
the residual water spectral line for frequency correction is done,
according to one highly beneficial further aspect of the present
disclosure, with the setting prescribed for the third flip angle.
The angle is lowered to reduce the water suppression function which
increases the residual water spectral line amplitude. Conversely
higher relative third flip angles will increase water suppression
for reduced water signal in an acquired MRS spectrum. A particular
flip angle for this purpose may be for example about 125, though
may be according to other examples between about 45 degrees and
about 125 degrees (or as much as about 145 degrees). Accordingly,
in one aspect, the flip angle may be for example at least about 45
degrees. In another aspect, the flip angle may be up to 125 degrees
or even up to 145 degrees. Notwithstanding these examples, an
expanded experimental data set of 79 discs in 42 subjects
represented under Example 3 included robust, reliable results
across this population with an average flip angle of about 120 or
121 degrees +/- about 33 degrees. Moreover, a later component of
that population conducted with further refinements revealed a
majority of cases suitable at a flip angle between about 65 degrees
and about 125 degrees, and in fact with most found to be sufficient
at about 85 degrees. It is to be appreciated that despite these
specific number and range examples, and robust results observed
therefrom, it is also believed that flip angles within about 5 or
10 degrees apart are likely to produce susbstantially similar
results for purposes intended herein.
[0112] This flip angle aspect of the present disclosure is another
example where some degree of customization may be required, in
order to optimize water signal for a given disc, in a given
particular MR system. As some discs may be more dehydrated or
conversely more hydrated than others, the water suppression may be
more appropriate at one level for one disc, and at another level
for another disc. This may require some iterative setting and
acquisition protocol to optimize, whereas the angle example
described herein is considered appropriate for most circumstances
and may be a pre-defined starting place for "first try."
[0113] For further clarity and understanding of the present DDD-MRS
pulse sequence embodiments introduced above and also elsewhere
herein described, FIG. 3A shows an example of a CHESS water
suppression pulse sequence diagram, whereas FIG. 3B shows an
example of a combined CHESS-PRESS pulse sequence diagram.
[0114] Very Selective Saturation (VSS) Bands
[0115] The volume excitation achieved using PRESS takes advantage
of three orthogonal slices in the form of a double spin echo to
select a specific region of interest. In some embodiments, the
range of chemical shift frequencies (over 400 Hz for proton at 3.0
T) is not insignificant relative to the limited band width of most
excitation pulses (1000-2000 Hz). The result can be a
misregistration of the volume of interest for chemical shift
frequencies not at the transmitter frequency. Thus, when a PRESS
volume is resolved by MRS, the chemical levels may be not only
dependent on tissue level, T1 and T2, but also dependent on
location within the volume of interest. In some embodiments, due to
imperfections in the RF pulse, out of volume excitation may occur
which can present signals from chemicals that are not in the
frequency/location range of interest.
[0116] Accordingly, another feature that is contemplated according
to a further mode of the DDD-MRS sequence is the use of very
selective saturation (VSS) pulses. This is often beneficial to
deploy for example for removal of signal contamination that may
arise from chemical shift error due to the presence of lipids
within the voxel as well as outside the selected ROI or voxel in
the disc nuclei. In the default operating mode of one DDD-MRS
sequence approach, in some regards sharing some similarities with
PROSE, for example, multiple pairs of VSS RF suppression bands are
placed symmetrically around the prescribed DDD-MRS voxel. In
certain embodiments, the voxel in this approach is oversized, such
as for example by 120% (e.g. PRESS correction factor=about 1.2).
The DDD-MRS sequence according to this mode uses the VSS bands to
define the actual DDD-MRS volume. It is believed that up to about
six additional VSS bands may be prescribed (each consisting of
three VSS RF pulses) graphically in PROSE, with the goal of
reducing the chemical shift error that can occur within the voxel
as well as suppress excitation of out of voxel tissue during the
PRESS localization of the voxel. According to some observations in
applying DDD-MRS to disc spectroscopy, these additional graphic VSS
pulses were found to not significantly improve the volume
selection. In other observations, some benefit is suspected to have
resulted. Accordingly, while they may provide benefit in certain
circumstances, they also may not be necessary or even desired to be
used in others.
[0117] FIG. 3C shows a schematic diagram of certain aspects of a
combined CHESS-VSS-PRESS pulse sequence diagram also consistent
with certain aspects of the present disclosure. As also shown for
further illustration in multiple respective image planes in FIGS.
4A-B, multiple VSS bands are placed around the voxel prescription
in each plane to reduce out of voxel excitation and chemical shift
error present during the PRESS localization of the voxel.
[0118] PRESS Timing Parameters
[0119] For purpose of comparative reference, the echo time (TE) of
about 130 ms is believed to be the default selection typically used
for PROSE data acquisitions. This echo time is typically considered
too long for DDD-MRS pulse sequence applications for acquiring
robust disc spectra due to the small voxel volume and shorter
T.sub.2 relaxation times of the chemical constituents of lumbar
intervertebral discs, leading to a dramatic decrease in signal to
noise in long echo PRESS spectra. Therefore a shorter echo time
setting for the scanner, such as for example about 28 milliseconds,
is generally considered more appropriate and beneficial for use in
the current DDD-MRS sequence and DDD pain application (though this
may be varied as would be apparent to one of ordinary skill based
upon review of this total disclosure and to suit a particular
situation). A frame repetition time (TR) of for example about 1000
ms provides sufficient relaxation of the magnetic dipoles in the
ROI and leads to reasonable acquisition times and is believed to
represent a beneficial compromise between short acquisition times
and signal saturation at shorter values of TR (though this may also
be varied, as also elsewhere herein described). Other appropriately
applicable operating parameter settings for PRESS spectra suitably
applicable to the DDD-MRS sequence may be, for example: number of
data points equal to about 1024, number of repetitions equal to
about 300, and example typical voxel size of about 4 mm.times.18
mm.times.16 mm (1.12 cc). Furthermore, first, second, and third
flip angles of PRESS for the current DDD-MRS sequence embodiment
may be for example 90, 167, and 167, respectively (though these may
slightly vary, and user-defined settings may not always reflect
actual angle--for example the latter two values may be exchanged
with or represent one example of an actual result of a 180 degree
setting).
[0120] Summary of User Control Variable (CV) Examples for DDD-MRS
Sequence
[0121] The foregoing disclosure describes various user controllable
sequence settings observed to be appropriate and of particular
benefit for use in an example DDD-MRS sequence according to the
current disclosure and for use for diagnosing DDD pain, as
contemplated under the preferred embodiments herein. These are
further summarized in Table 1 appended herewith at the end of this
disclosure.
[0122] One or more of these CVs may comprise modifications from
similar settings that may be provided for another CHESS-PRESS or
CRESS-VSS-PRESS pulse sequence, such as for example PROSE, either
as defaults or as user defined settings for a particular other
application than as featured in the various aspects herein this
disclosure. These CV settings, in context of use as modifications
generally to a sequence otherwise sharing significant similarities
to PROSE, are believed to result in a highly beneficial resulting
DDD-MRS sequence for the intended purpose of later signal
processing, according to the DDD-MRS signal processor embodiments
herein described, and performing a diagnosis of DDD pain in discs
examined (the latter according for example to the DDD-MRS
diagnostic processor aspects and embodiments also herein
disclosed). However, it is also appreciated that these specific
settings may be modified by one of ordinary skill and still provide
highly beneficial results, and are also contemplated within the
broad intended scope of the various aspects of this present
disclosure.
[0123] Data Acquisition of DDD-MRS Pulse Sequence
[0124] The signal detected in the MR spectrometer in the receiving
"detector" coil assembly, after exposing a sample to a radio
frequency pulse, is called the Free Induction Decay (FID) for
purpose of this disclosure. In modern MR spectrometers the MR
signal is typically detected using quadrature detection. As a
result, the acquired MR signal is composed of two parts, often
referred as real and imaginary parts of the FID. A schematic
example of the time domain FID waveform is shown in FIG. 5, which
shows the real (Sx) and imaginary (Sy) parts of an FID (right) that
correspond to x and y components of the rotating magnetic moment M
(left).
[0125] FIDs are generated at the period defined by TR. Thus a TR of
about 1000 milliseconds, according to the example embodiment
described above, equals a rate of about 1 Hz (about one FID per
second). The FID signal received from each coil channel is
digitized by the scanner to generate a 1024 point complex number
data set or acquisition frame. An MRS scan session consists of a
number of frames of unsuppressed water FIDs (such as for example
may be about 16 frames) and up to 368 or more (as may be defined by
an operator or setting in the pulse sequence) frames of suppressed
water FIDs, which together are considered an acquisition series.
The unsuppressed water FIDs provide a strong water signal that is
used by the signal processing to determine which coils to use in
the signal processing scheme as well as the phase information from
each coil (and in certain embodiments may also be used for
frequency error correction). However, due to gain and dynamic range
in the system these high water content unsuppressed frames do not
typically provide appropriate resolution in the target biomarker
regions of the associated spectra to use them for diagnostic data
purposes. The suppressed water FIDs are processed by the DDD-MRS
processor to obtain this spectral information, although the
unsuppressed frames may be used for certain processing approaches
taken by the processor.
[0126] For further illustration, FIG. 6 shows a plot of all the
FIDs obtained in a DDD-MRS pulse sequence scan according to certain
present illustrative embodiments, and is an amplitude plot of
complex data from a standard DDD-MRS acquisition with the y-axis
representing the magnitude of FID data and the x-axis representing
serial frame count over time.
[0127] DDD-MRS Pulse Sequence Data Transfer from MR System to
Post-Processor
[0128] The MR scanner generates the FIDs using the defined
sequences to energize the volume of interest (VOI), digitizes them
according to the defined data acquisition parameters, and stores
the data, typically as floating point numbers. While this data may
be packaged, e.g. in "archive file," and communicated in various
formats and methods, one example is provided here. A data
descriptor header file (DDF) with all the aforementioned parameters
along with voxel prescription data is appended to the data to
generate the archive file. Examples of certain parameters provided
in a DDF, are as follows: studyID (String); seriesNum (Integer for
assigned Series Number); studyDate (String date code); seriesDesc
(String for series description); rootName (String); nSamps (Integer
for number of complex samples, typically 1024); nFrames (Integer
for number of frames or reps); coilName (String); pulseSeqName
(String); Te (Float, echo time, in ms); Tr (Float, repetition time,
in ms); TxFreq (Float, in MHz); nSatBands (Integer, number of
saturation bands); voxTilt (Float, voxel tilt about x-axis, in
degrees); voxVol (Float, Voxel volume in cc); voxX (Float, Voxel X
dimension, in mm); voxY (Float, Voxel Y dimension, in mm); voxZ
(Float, Voxel Z dimension, in mm). The archive file can include
data received from the MR scanner that is representative of the
anatomy of a patient (e.g., representative of the chemical makeup
of tissue inside the area of interest inside an intervertebral disc
of the patient's spine).
[0129] The archive file may then be transferred to another computer
running an application written in a language, such as for example
Matlab.RTM. R2009a (e.g. with "Image Processing Toolbox" option,
such as to generate time-intensity plots such as shown in various
Figures herein), which opens the archive file. The Matlab
application may be user-configurable, or may be configured as full
or partial executables, and is configured to signal process the
acquired and transferred DDD-MRS information contained in the
archive file, such as according to the various signal processing
embodiments herein. Other software packages, such as "C," "C+," or
"C++" may be suitably employed for similar purposes. This
application, subsequently referred to as the DDD-MRS signal
processor, parses information pertinent to the signal processing of
the data from the data description header, and imports the FID data
acquired at each detector coil for subsequent signal processing. It
will be understood that the DDD-MRS signal processor can be
implemented in a variety of manners, such as using computer
hardware, firmware, or software, or some combination thereof. In
some embodiments, the DDD-MRS signal processor can comprise a
computer processor configured to execute a software application as
computer-executable code stored in a non-transitory
computer-readable medium. In some embodiments, the computer
processor can be part of a general purpose computer. In some
embodiments, the DDD-MRS signal processor can be implemented using
specialized computer hardware such as integrated circuits instead
of computer software. It will be understood that the DDD-MRS signal
processor, as well as other components described herein that may be
implemented by a computer, can be implemented by multiple computers
connected, for example, by a network or the internet. Thus,
algorithms, processes, sequences, calculations, and tasks, etc.
described herein can be divided into portions to be performed by
multiple computer processors or other hardware located on multiple
physically distinct computers. Also, some tasks that are described
herein as being performed by distinct computers or systems may be
performed by a single computer or a single integrated system.
[0130] The archive file and related MRS data may be communicated
via a number of available networks or methods to external source
for receipt, processing, or other form of use. In one particular
typical format and method, the information is communicated via
picture archiving and communication system (PACS) that has become
ubiquitous for storing and communicating radiologic imaging
information. In addition to the archive file with DDF and stored
MRS data, accompanying MRI images may also be stored and
communicated therewith, e.g. in standardized "digital imaging and
communications in medicine" or "DICOM" format.
[0131] The data transfer described may be to a local computer
processor for processing, or more remotely such as via the web
(typically in secure format). In alternative to data transfer of
acquired MRS data pre-processing to an external system for
post-processing as described above, e.g. MRS signal processor and
diagnostic processor aspects of this disclosure, all or a portion
of the various aspects of the present embodiments may be installed
or otherwise integrated within the MR system itself, e.g. a
computer based controller or processor embedded therewithin or
otherwise connected thereto, for operation prior to packaging
results for output (and any remaining portions might be performed
peripherally or more remotely).
[0132] DDD-MRS Signal Processing
[0133] Upon the acquisition of all MRS data from a DDD-MRS pulse
sequence exam, according to certain aspects of the present
embodiments, the MR scanner system will typically provide the
operator with a spectral image that is the averaged combination of
all frames across all the 6 detection channels (coils). An example
of such a waveform from an MRS pulse sequence exam acquired for an
ROI in a disc nucleus via a GE Signa 3 T MR system is shown in FIG.
7, which shows a typical scanner-processed spectral signal plot of
combined, averaged channels. FIG. 8 shows the magnitude only (no
correction) MRS spectral images of each of the six channels which
are aggregated to form the output from the example MR system as
shown in FIG. 7, and thus this raw uncorrected individual channel
spectral data output provides the input to the DDD-MRS signal
processor of the present embodiments.
[0134] According to one highly beneficial mode, the DDD-MRS signal
processor is configured to conduct a series of operations in
temporal fashion as described herein, and as shown according to the
present detailed embodiments in the flow charts illustrated in
general to increasing detail for the various component modes and
operable steps in FIGS. 9A-C. More specifically, FIG. 9A shows a
general schematic overview for the flow of a diagnostic system 2
that includes a signal processor 4 and a diagnostic processor 6.
Signal processor includes various sub-components and processors
that carry out certain steps, such as a channel selector that
conducts channel or "coil" selection step 10, phase corrector that
does phase correction step 20, apodizer that conducts apodizer step
30, domain transformer that conducts the domain transform step 44
such as from time domain to frequency domain, frame editor that
conducts frame editing steps 50, frequency corrector that conducts
frequency correction steps 60, and channel combiner that conducts
combining or averaging steps 70 to aggregate retained channels into
one final post-processed spectral results (not shown). Following
signal processing steps 6, a diagnostic processor conducts
diagnostic processing of the signal processed signals through data
extraction steps 110, diagnostic algorithm application steps 120,
and patient or diagnostic report generation 130. While this
configuration is considered highly beneficial, these same or
similar tasks may be performed in different order, as would be
apparent to one of ordinary skill.
[0135] For illustration, FIGS. 9B-C show further details regarding
some of these specific steps, and also illustrate a different order
than has been shown and referenced to FIG. 9A. More specifically,
the signal processor 4 in shown to include the main primary steps
shown in FIG. 9A, but in finer detail and different order. It some
embodiments, the steps shown in FIGS. 9A-C may be performed in an
order different than those shown in FIGS. 9A-C. Also, in some
embodiments, steps that are shown in FIGS. 9A-C can be omitted,
combined with other steps, or divided into additional sub-steps.
Additionally, in some cases, additional steps not specifically
shown in FIGS. 9A-C can be performed in addition to the steps shown
in FIGS. 9A-C.
[0136] Channel selection includes the following steps: signal power
measurement step 11 measures signal power for SNR calculation, as
shown here in the specific embodiment in first 100 points of FID
with unsuppressed water. Noise power measurement step 13 measures
noise in the last 100 points, for example, of the FID with
unsuppressed frames. SNR estimate 15 is then conducted, at which
point thereafter channel selection step 17 is conducted per the
channel with the maximum or highest signal. Channel selection
includes an additional step 18 where additional channels are
selected if within range of the strongest, e.g. about 3 dB. Upon
completing channel selection, an index of selected channels is
generated (step 19).
[0137] Frame editing operation 50 is also shown, with
transformation of unsuppressed water frame to frequency domain 51,
locate water peak in +/-40 Hz per peak location step 53, frame
confidence level calculation 55, frame frequency error and store
57, and actual frame selection step 59 based upon minimum
confidence level threshold (e.g. 0.8) Phase correction 20 is also
done per applying 21 1.sup.st order linear curve fit to the FID of
each unsuppressed water frame (e.g. n=16), obtaining average of
zero order terms from the curve fit 23, and rotate 25 all
suppressed water frame FIDs by zero order term. Apodization 30
includes for each selected channel and each frame indexed for
frequency correction 31, then apply 250 point boxcar function to
the FID (step 33). In addition, frequency correction 60 entails for
each selected channel and each frame indexed for frequency
correction and apodization 61, transform 63 the frame (FID) to the
frequency of domain, and locate the frequency error 65 for the
frame as identified during frame editing. The spectrum is shifted
67 by the frequency error value to frequency correct the spectrum.
Step 69 adds frequency corrected spectrum to many to spectral
average for all selected channels.
[0138] As also shown in FIG. 9C for the diagnostic processor 6,
data extraction 110 involves opening 111 the acquisition spectral
average lot generated in frequency correction, scan step 113 of
spectral plot for metabolite features and apply to bins, and record
data bins to acquisition metabolite 115. The diagnostic algorithm
120 itself involves opening the acquisition metabolite file
extracted from spectral average 121, extract 123 metabolite
features required by diagnostic linear regression equations, and
generate 125 acquisition diagnostic score and store to file. Report
generation 130 includes open acquisition diagnostic score fill 131,
open DICOM file 133 for Patient ID associated with acquisition
diagnostic score, extract 135 sagittal image from the DICOM report,
apply 137 acquisition score to sagittal image for each scanned
disc, and return sagittal image to DICOM report 139.
[0139] According to the current example embodiment, a first
operation of the DDD-MRS processor assesses the SNR of each
channel. This is done to determine which channels have acquired
sufficiently robust signal to use for data processing and
averaging. The result may produce one single channel that is
further processed, or multiple channels that are later used in
combination under multi-channel averaging. In the majority of
acquired signals observed according to the Examples disclosed
herein, only a subset of the 6 lumbar acquisition channels were
determined to be sufficiently robust for use. However, the standard
system output averages all 6 channels. Accordingly, this filtering
process alone--removing poor signal channels and working with only
stronger signal channels--has been observed to dramatically improve
processed spectra for diagnostic use in some cases. While various
techniques may be suitable according to one of ordinary skill, and
thus contemplated herein, according to the present illustrative
embodiment the SNR is calculated by obtaining the average power in
the first 100 data points (the signal) and the last 100 points (the
noise) of the unsuppressed water FID. The unsuppressed water FIDs
signals are used because of the strong water signal. The channel
with the greatest SNR, and channels within a predetermined
threshold variance of that strongest one, e.g. within about 3 dB
for example, are preserved for further processing and as candidates
for multi-channel averaging--other channels falling below this
range are removed from further processing (though may be used for
further processing, yet removed from final results).
[0140] Further examples and embodiments for evaluating relative
channel quality are provided as follows. One additional indication
of channel quality that may be observed and used is the line width
of the unsuppressed water signal based on the averaged frequency
and phase corrected FFTs of the coil channels with the highest
SNRs. This is computed to serve as a general indicator of signal
quality as determined by the quality of the shimming process and to
provide an estimate of the resolution we should expect in the
chemical shift spectrum. Another indication of channel quality is
the degree of water suppression. This has utility in determining
the optimum degree of water suppression to apply in the acquisition
protocol. The water suppression should leave enough residual water
signal to use as a reference to reliably perform frame-by-frame
frequency correction but not so much that water signal artifacts
affect the chemical shift spectrum in the metabolite areas of
interest. Such artifacts include simple spectral leakage as well as
phase modulation sidebands due to gradient vibration induced
B.sub.o modulation.
[0141] Further to the present embodiments and per further reference
to FIGS. 9A-B, a second operation conducted by the DDD-MRS
processor is phase alignment, or phase error correction. This is
performed to support coherent summation of the signals from the
selected channels and the extraction of the absorption spectra.
This is often necessary, or at least helpful, because in many cases
a systemic phase bias is present in the different channels. This
systemic phase bias is best estimated by analysis of the data
frames (e.g. about 16 frames in the DDD-MRS pulse sequence of the
current illustrative detailed embodiments) collected at the
beginning of each scan without water suppression. This operation,
according to one mode for example, analyzes the phase sequence of
the complex samples and fits a polynomial to that sequence. A
first-order (linear) fit is used in one further illustrative
embodiment. This is believed to provide a better estimate of the
offset than simply using the phase of the first sample, as is often
done. This is because eddy current artifacts, if present, will be
most prominent in the first part of the frame. The offset of the
linear fit is the initial phase. Observation has indicated that the
first 150 samples (75 mS at the typical 2000 samples-per-second
rate) typically provide reliable phase data. The fit is performed
on each of the water-unsuppressed frames for each channel and the
mean phase of these is used to phase adjust the data for the
corresponding channel. This is accomplished by performing a phase
rotation of every complex sample in each frame to compensate for
the phase offset as estimated above, setting the initial phase to
zero.
[0142] The offset of the linear fit is the phase bias with respect
to zero and the slope is the frequency error with respect to
perfect center-tuning on the water signal. Only the offset portion
of the curve fit is used to phase correct the data. An illustrative
example of this is shown in schematic form in FIG. 10, which shows
phase angle before and after phase correction. The phase angle
signal is shown as the dotted line. The solid line is the least
squares fit estimate. The dashed line is the phase and frequency
corrected signal, though the offset component is used to phase
correct and frequency correction is performed subsequently in the
temporal process according to the present DDD-MRS processor
embodiment.
[0143] The real-part squared MRS spectral results of phase
correction for each of all the six channels shown prior to
correction in FIG. 8 is shown in FIG. 11, with channels 1-3
indicated from left to right at the top, and channels 4-6 indicated
from left to right at the bottom of the figure. The averaged
spectrum of the selected, phase corrected channels (channels 1 and
2) is shown in FIG. 12, which reflects significant SNR improvement
versus the uncorrected all channel average spectrum shown in FIG.
7.
[0144] Frame Editing
[0145] While it is contemplated that in some circumstances
individual MRS acquisition frames may provide some useful
information, frame averaging is prevalently indicated in the vast
majority of cases to achieve a spectrum with sufficient SNR and
interpretable signal at regions of interest for pathology
assessment. It is, at most, quite rare that an individual frame
will have sufficient SNR for even rudimentary metabolite analysis
to the extent providing reliable diagnostic information. Often
individual frames along an acquisition series will have such low
SNR, or possess such artifacts, that they make no improvement to
the average--and in fact may even degrade it. To the extent these
"rogue" frames may be recognized as such, they may be excluded from
further processing--with only robust frames remaining, the result
should improve.
[0146] Accordingly, a further mode of the present DDD-MRS processor
embodiment utilizes a frame editor to conduct frame editing to
identify those frames which vary sufficiently from the expected or
otherwise observed acquisition results such that they should be
excluded, as is also represented schematically in the flow diagram
examples of FIGS. 9A-B. In one aspect of the underlying concern,
certain patient motions during an acquisition may result in signal
drop-out as well as frequency shifts (e.g. magnetic susceptibility
artifact). While involuntary motion, e.g. respiration, is a common
cause of frequency shifts, these are typically sufficiently minor
and within a range that they are not believed to implicate signal
quality other than the shift itself (which can be significant
source of SNR degradation, but correctable per the present
disclosure). However, other more significant movements (e.g.
voluntary) may cause sufficiently significant shifts to seriously
degrade the acquired spectrum, beyond merely correctable spectral
shifts. For example, such activity may move the voxelated region to
include adjacent tissues versus only the intended VOI upon
prescription prior to the motion. If the salient artifact is
frequency shift, a correction may be applied and the frame can be
used to make a positive contribution to the averaged spectrum. If a
frame is discarded its contribution is lost, and across sufficient
number of discarded frames across a series the result may not
include a sufficient number of frames in the average for a reliable
SNR in the resulting spectrum. The DDD-MRS processor, according to
the current embodiment, analyzes the residual water signal in each
frame to determine if it is of sufficient quality to support
frequency correction.
[0147] FIG. 13 shows a time-intensity plot which illustrates a scan
series with frequency shifts and "drop outs" with SNR changes
considered to represent corrupted frames due to patient motion.
More specifically, this shows 1 dimensional horizontal lines for
each frame, with signal amplitude reflected in "brightness" or
intensity (e.g. higher values are whiter, lower are darker), with
time across the serial acquisition of the series progressing top to
bottom vertically in the Figure. A vertical band of brightness is
revealed to the left side of the plot. However, in this particular
example, there is a clear break in this band as "drop out" frames.
After excluding the "drop out" frames (center of time sequence
between about 75 and 175 MRS frames, it was still possible to
obtain a high quality final averaged spectrum from this scan using
the remaining robust frames, as further developed immediately
below.
[0148] FIGS. 14A and 14B show the confidence level estimate and the
frame by frame frequency error estimate, respectively, which are
used according to the present embodiment for frame editing
according to this acquisition series example of FIG. 13. More
specifically, FIG. 14A shows the frame by frame confidence level,
with confidence level on the Y-axis, and the sequential series of
frame acquisitions along a scan indicated along the X-axis. FIG.
14B shows the actual frequency error along the Y-axis, for the same
frame series along the X-axis. This is based on analyzing the
characteristics of the residual water peak and the noise in a band
80 Hz wide (for 3 T processing, the band would be 40 Hz wide at 1.5
T) around the center-tuned frequency. The largest peak is assumed
to be the water signal and the assumption is qualified by the
confidence estimate. For the purpose of this example, if the
confidence value is above a threshold, i.e. 0.8, the frame is
flagged as a candidate for frequency correction and thus
"retained." As seen from the plots in FIGS. 14A-B, when the
confidence is low, the variance of the frequency error estimate is
greatly increased. The final qualification step, per this example,
is to determine if there are enough qualified candidate frames to
achieve sufficient SNR improvement when averaged. This threshold
limit for proceeding with frequency correction (and thus frame
editing therefore) has been empirically established as 90 frames
meeting the criteria. According to the present embodiments, this
has been observed to provide sufficiently robust results per the
Examples described herein. It is to be appreciated, however, that
other limits may be appropriate in various circumstances. The
number of frames required will be based upon the SNR levels
achievable from the completed signal processing. This will be paced
by SNR of input signal acquisitions to begin with, and performance
of other signal processing modes and steps taken with those
signals. According to the acquisitions under the Examples disclosed
herein, SNR is believed to increase over about 150 frames, and then
with little gained typically thereafter, though the 90 frame
minimum limit has been observed to provide sufficient results when
reached (in rare circumstances). In the event the result drops
below the 90 frame limit, the DDD-MRS processor is still configured
to proceed with other modes of signal processing, signal quality
evaluation, and then diagnostic processor may be still
employed--just without the added benefit of the frame editing and
frequency error correction.
[0149] Further description related to acceptable confidence level
estimate approach according to the present disclosure is provided
as follows, for further illustration of this embodiment for the
frame editing and frequency correction modes of the disclosure. The
discrete amplitude spectrum can be analyzed in the range of the
center-tuned frequency .+-.40 Hz for example in the case of a 3 T
system acquisition, and half this bounded range (e.g. .+-.20 Hz)
for a 1.5 T system acquisition. The highest peak is located to
determine it's width at the half-amplitude point. Next, the total
spectral width of all parts of the spectrum which exceed the
half-amplitude point of the highest peak are determined. The
confidence estimate is formed by taking the ratio of the spectral
width of the greatest peak divided by the total spectral width
which exceeds the threshold. If there is only a single peak above
the threshold, the confidence estimate will be 1.0, if there are
many other peaks or spectral components which could be confused
with the greatest one, then the estimate will approach 0.0. This
provides a simple and robust estimate of the randomness or
dispersal of energy in the vicinity of the water peak. Like another
approach using entropy measurement, e.g. as described below, this
current approach provides at least one desirable characteristic in
that it's performance is substantially invariant with
amplitude.
[0150] Yet another system and method to compute a confidence
estimate that also can be appropriate is provided as follows. The
spectral entropy is computed by normalizing the spectrum to take
the form of a probability mass function. The Shannon entropy or
uncertainty function, H, is then computed as follows:
H=-.SIGMA.p.sub.i log.sub.2 p.sub.i
where p=probability, and i=frequency index value (e.g. -40 to +40
hz).
[0151] It is to be appreciated that other approaches to quantify
randomness or uncertainty of the spectrum may also be suitable for
use with the various DDD_MRS signal processor aspects of the
present disclosure.
[0152] For further understanding and clarity re: the ultimate
impact frame editing as described herein, the unprocessed
absorption spectrum plot for all six channels from the patient
(with the compromised frames included as aggregated in the
respective channel spectra) in various views in prior Figures is
shown for each respective channel in the six indicated panes shown
in FIG. 15. The phase and frequency corrected spectrum averaged for
selected channels 3 and 4, and for all 256 acquired frames
aggregated/averaged per channel, without applying frame editing and
thus including the corrupted frames, is shown in FIG. 16A. In
contrast, FIG. 16B shows a similar phase and frequency error
corrected spectrum averaged for the same selected channels 3 and 4,
but for only 143 of the 256 acquired frames aggregated/averaged per
channel (the remaining 113 frames edited out), per frame editing
applied according to the present embodiments prior to frequency
error correction. The peak value in the combined lactate-alanine
(LAAL) region of the frame edited spectrum of FIG. 16B is
significantly increased--with corresponding increase in
SNR--relative to the peak value in the same LAAL region of the
non-frame edited spectrum in FIG. 16A (e.g. the peak value
increases from about 3.75.times.10.sup.8 to about
4.4.times.10.sup.8), a nearly 20% SNR increase despite about 40%
corresponding reduction in the number of FID frames used.
[0153] While the examples addressed above by reference to FIGS.
13-16B address a highly beneficial embodiment for frame editing
based upon water signal, other frame editing embodiments are also
contemplated, and many different features of acquired DDD-MRS
signals may be used for this purpose. One such further embodiment
is shown for example by reference to another DDD-MRS pulse sequence
acquisition for another disc in another subject by reference to
FIGS. 17A-F. More specifically, per the time-intensity plot shown
for this acquisition in FIG. 17A, while the water signal region of
the acquired spectral series (bright vertical band on left side of
plot) reveals some shift artifact, another bright band appears at a
broader region on the right side of the spectra, between about 150
and 200 FID frames into the exam. This region is associated with
lipid, and also overlaps with lactic acid (LA) and alanine (AL)
regions of diagnostic interest according to the present detailed
embodiments and Examples. This is further reflected in FIG. 17B
which shows a waterfall plot of running cumulative average of
acquired frames in series, where signal amplitude rises in this
lipid-related spectral region during this portion of the exam. A
resulting average spectral plot for channels 1 and 2 of this
acquisition, post phase and frequency correction (again noting
water signal did not prompt frame editing to remove many frames) is
shown for reference in FIG. 17C. This resulting spectrum has
significant signal peak intensity and line width commensurate with
lipid signal, and which shrouds an ability to assess underlying LA
and/or AL chemicals overlapping therewith in their respective
regions. Accordingly, an ability to measure LA and AL being
compromised may also compromise an ability to make a diagnostic
assessment of tissue based upon these chemicals (as if un
compromised by overlapping lipid). However, as this lipid
contribution clearly only occurs mid-scan, an ability to edit it
out to assess signal without that portion of the exam may provide a
robust result for LA and AL-based evaluation nonetheless.
[0154] This is shown in FIGS. 17D-F where only the first 150 frames
of the same acquisition are evaluated, which occur prior to the
lipid contribution arising in the acquired signals. As is shown
here, no lipid signal is revealed in the time intensity plot of
FIG. 17D, or waterfall plot of FIG. 17E, or resulting final
spectrum of FIG. 17F, though strong proteoglycan peak is shown with
very little (if any) LA or AL in the signal of otherwise high SNR
(e.g. per the PG peak). As this example illustrates a DDD-MRS
processed acquisition for a non-painful control disc, strong PG
signal and little to no LA and/or AL signal is typically expected,
and this thus represents a diagnostically useful, robust signal for
intended purpose (whereas the prior spectra without editing out the
lipid frames may have erroneously biased the results). Accordingly,
it is contemplated that a lipid editor may also be employed as a
further embodiment for frame editing, with approaches for
recognizing lipid signal taken as elsewhere herein described (or as
may be otherwise available to one of ordinary skill and
appropriately applicable to this applied use).
[0155] Frequency Correction
[0156] As noted elsewhere herein, during the course of a typical
single voxel DDD-MRS series acquisition cycle according to the
pulse sequence aspects of the present embodiments (e.g. about 2-4
minutes, depending upon settings chosen for TR and number of
frames), frequency errors can occur due to patient motion and
changes in magnetic susceptibility (respiration, cardiac cycle
etc). In this environment where the acquired spectral signals
"shift" along the x-axis between multiple sequential frames in an
exam series, their subsequent averaging becomes "incoherent"--as
they are mis-aligned, their averaging compromises signal quality.
Unless this is corrected to "coherently" align the signals prior to
averaging, this error can result in an increase in line width,
split spectral peaks and reduced peak amplitudes for diminished
spectral resolution relative between signal peaks themselves (as
well as reduced SNR). Accordingly, the DDD-MRS processor according
to further aspects of this disclosure comprises a frequency error
corrector that performs frequency correction, such as for example
prior to averaging frames, as also represented schematically in the
flow diagrams of FIGS. 9A-B.
[0157] This is performed according to one embodiment in the
frequency domain. This is done by transforming the time domain data
for each frame into frequency domain absorption spectra, locating
the water absorption peaks, and shifting the spectrum to align them
to an assigned center reference location or bin. Once shifted, the
frame spectra are averaged in the frequency domain to generate the
corrected or "coherent" channel spectra. In another embodiment, the
desired frequency shift correction for a frame may be applied to
the time domain data for that frame. The time domain data for all
the frames would then be averaged with the final average then
transformed back to spectra. While the processes are linear and
thus not dependent upon sequence of operation, it is believed in
some circumstances that the latter embodiment may present slightly
increased spectral resolution. In difficult signal acquisition
situations, some of the frames do not have sufficient signal
quality to support frequency correction. More specifically, water
signal in some frames may be insufficiently robust to accurately
"grab" its peak with high degree of confidence. This circumstance
is addressed by another operation of the DDD-MRS processor, frame
editing in which the frames are omitted if the water peak cannot be
identified with sufficient confidence, also described herein
(though may be performed independent of frame editing, which may
not necessarily be required to be performed, despite the distinct
benefits believed and observed to result therefrom).
[0158] The frame editing can be performed distinct from the
frequency correction process (e.g., performed beforehand), or the
frame editing and frequency correction can be performed
simultaneously. The DDD-MRS processor can attempt to identify the
water peak, calculate a level of confidence that the identified
peak is water. If the confidence level is below a threshold, the
frame can be disregarded. If the confidence level is above a
threshold, the water peak, as well as the rest of the spectrum, can
be shifted to its proper alignment. The DDD-MRS processor can then
proceed to the next frame in the sequence.
[0159] Frequency error can be visualized using a time-intensity
plot of the absorption spectra of all the frames in an acquisition
cycle. An example process and related results of frequency error
correction according to this present embodiment is shown and
described by reference to FIGS. 18A-21 for the same DDD-MRS series
acquisition featured in FIGS. 7-8 (prior to any correction) and
FIGS. 11-12 (per prior DDD-MRS signal processing step of phase
error correction). As shown in FIGS. 18A-19B (and similarly for
prior FIGS. 13 and 17A), each acquisition frame is represented by a
horizontal line, with amplitude of signal intensity across the
frequency spectrum indicated by brightness in grey scale (brighter
shade/white designates higher amplitude, darker signal intensity
indicates lower relative amplitude). The horizontal lines
representing individual acquisition frames are displayed in
vertically "stacked" arrangement that follows their temporal
sequence as acquired, e.g. time zero is in the upper left corner
and frequency incremented from left to right. The top 16 lines
represent unsuppressed water frames, with the remainder below
representing suppressed water acquisitions. The brightest portion
of each line (left side of the time-intensity plots) is reliably
recognized as the water peak absorption, typically the strongest
signal of acquired MRS spectra in body tissues.
[0160] Further to FIG. 18A, this plot for the original acquired
sequence of frames from an acquisition series intended to be
averaged is shown pre-frequency correction (e.g. with original
frequency locations), and similar view but post-frequency
correction is shown in FIG. 18B. Shifting of the location of this
bright white water peak region, as observed between vertically
stacked frames, indicates frequency shift of the whole MRS spectrum
between those frames--including thus the peaks of spectral regions
of interest related to chemicals providing markers for pain. The
rhythmic quality observed in this frequency shifting, per the
alternating right and left shifts seen around a center in the
uncorrected plot (left side of figure) shift, remarkably
approximates frequency of respiration--and thus is believed to
represent respiration-induced magnetic susceptibility artifact. The
contrasted plots seen in the pre and post frequency corrected time
intensity plots shown in FIGS. 18A-B reveal the process to achieve
corrected "alignment" of the previously shifted signals for
coherent averaging. For further clarity, each of two similar views
of an enhanced contrast image (FIGS. 19A-B) (though FIG. 19B
reveals wider range of MRS Frames acquired in the series), shows
the original frequency shifted, incoherent mis-alignment (FIG. 19A)
and frequency corrected, coherent alignment (FIG. 19B) of the water
peaks from this same acquisition series. In this example case shown
in FIGS. 18A-B and FIGS. 19A-B, all of the frames were of
sufficient quality to support frequency correction.
[0161] The frequency corrected absorption spectra for each
acquisition cycle are averaged to generate an average frequency
(and phase) corrected spectra for each channel, as is shown in FIG.
20. The selected channels (channels 1 and 2) are then averaged to
produce the final spectra (FIG. 21) used for extraction of data
along spectral regions of interest that are considered relevant to
DDD pain diagnosis. In comparing the phase+frequency error result
of FIG. 21 against the phase error corrected-only result for the
same acquisition series in FIG. 12, a significant increase in SNR
and general signal quality is revealed in the latter more fully
processed case--showing for example an increase from slightly more
than 11.times.10.sup.7 peak intensity with clear doublet in the PG
region in FIG. 12 (where a doublet is not typically found, and
likely reflective result of incoherent averaging of the PG peak) to
nearly 18.times.10.sup.7 or an 80% peak intensity increase with
narrower band and no doublet in FIG. 21, and also clearly higher
PG/LA and/or PG/LAAL ratio, as are signal qualities elsewhere
revealed herein to be of diagnostic relevance in some highly
beneficial applications. Still further comparison against the fully
unprocessed spectral output from the MR scanner in FIG. 7 for the
same acquisition series reveals even more dramatic signal quality,
and in particular SNR, improvement.
[0162] The following documents are herein incorporate in their
entirety by reference thereto: [0163] 1. Bottomley P A. Spatial
localization in NMR spectroscopy in vivo. Ann N Y Acad Sci 1987;
508:333-348. [0164] 2. Brown T R, Kincaid B M, Ugurbil K. NMR
chemical shift imaging in three dimensions. Proc. Natl. Acad. Sci.
USA 1982; 79:3523-3526. [0165] 3. Frahm J, Bruhn H, Gyngell M L,
Merboldt K D, Hanicke W, Sauter R. Localized high-resolution proton
NMR spectroscopy using stimulated echoes: initial applications to
human brain in vivo. Magn Reson Med 1989; 9:79-93. [0166] 4.
Star-Lack J, Nelson S J, Kurhanewicz J, Huang L R, Vigneron D B.
Improved water and lipid suppression for 3D PRESS CSI using RF band
selective inversion with gradient dephasing (BASING). Magn Reson
Med 1997; 38:311-321. [0167] 5. Cunningham C H, Vigneron D B, Chen
A P, Xu D, Hurd R E, Sailasuta N, Pauly J M. Design of
symmetric-sweep spectral-spatial RF pulses for spectral editing.
Magn Reson Med 2004; 52:147-153. [0168] 6. Pauly J, Le Roux P,
Nishimura D, Macovski A. Parameter relations for the Shinnar-Le
Roux selective excitation pulse design algorithm [NMR imaging].
IEEE Trans Med Imaging 1991; 10:53-65. [0169] 7. F. Jim, Europeant
Journal of Radialogy 67, (2008) 202-217 The following U.S. Patent
Application Publications are herein incorporated in their entirety
by reference thereto: US2008/0039710 to Majumdar et al.; and
US2009/0030308 to Bradford et al.
[0170] DDD-MRS Diagnostic Processor and Use for Diagnosing DDD
Pain
[0171] Development, application, and evaluation of a DDD-MRS
diagnostic processor configured for use for diagnosing DDD pain
based upon DDD-MRS acquisition series acquired from discs according
to a DDD-MRS pulse sequence and DDD-MRS signal processor
applications is disclosed by reference to the Examples and other
disclosure provided elsewhere herein.
[0172] The diagnostic processing aspects of the present disclosure
is also represented schematically in the flow diagrams of FIGS. 9A
and 9C, and generally includes multiple individual steps or
operations: (1) regional MRS spectral data extraction; and (2)
diagnostic algorithm application. In addition, the diagnostic
results will be typically displayed or otherwise produced in an
appropriate fashion intended to satisfy an intended use.
Furthermore, despite the many significant benefits of the DDD-MRS
signal processor aspects herein disclosed for producing reliably
robust MRS spectra from such DDD-MRS pulse sequence exams of disc
nuclei, certain results will nonetheless provide insufficient
signal quality, such as due to low SNR below a threshold (e.g. 2 or
3), water "washout" of signal, lipid artifact, or obviously out of
phase outer voxel artifact, for making reliable measurements in
spectral regions of diagnostic interest (e.g. considered to
represent certain chemical biomarker regions). In the event such
poor quality signals were to enter the diagnostic process of
extracting data for diagnostic algorithm purposes, the results
would be much more likely to be corrupted by noise artifact vs.
real signal basis of the measured values, and could potentially
yield diagnostically incorrect results.
[0173] Accordingly, the present disclosure according to further
aspects includes a spectrum quality analyzer which determines which
signals otherwise passed through the DDD-MRS signal processor
modules have sufficient signal quality to perform diagnostic
algorithm, and which do not. As for the latter, these may be
considered "indeterminate" or otherwise "failed test" results and
thus not used diagnostically. This may prompt a repeat exam,
perhaps with modified parameters intended to counteract the
underlying cause of such poor quality (e.g., low SNR or lipid
artifacts), such as by re-voxelating according to a different
prescription (e.g., increasing voxel size, or decreasing voxel
size, or moving its location), adjusting water suppression, etc. In
order to assist in appropriately directing such corrections in a
re-exam, the spectrum quality analyzer may compare certain aspects
of the subject signal against known features associated with such
corruptions, determine the potential source of corruption, and flag
and/or identify to a user a suspected cause (and may further
recommend one or more courses of action to attempt correcting in a
re-exam).
[0174] As this spectrum quality analyzer assesses the result of
signal processing, it may be considered a part of the overall
DDD-MRS signal processor. However, as it also comprises one of
potentially multiple analysis algorithms to determine "procedural
failures" from the processed DDD-MRS acquisition and filter them
out from further diagnostic processing to an affirmative result, it
may also in some regards be considered a portion of the diagnostic
processor.
[0175] As still another embodiment of the diagnostic processor of
the present disclosure, spectral data may be acquired for
diagnostic purposes, such as processing through a diagnostic
algorithm, and thus a data extractor is also provided and as
featured in FIGS. 9A and 9B. The data extraction or acquisition can
typically involve recognizing regions along the spectrum generally
associated with certain specific biomarker chemicals of diagnostic
interest (e.g. spectral regions of diagnostic interest or "SRDI"),
and extracting target data from such SRDIs. These SRDI's will
typically have known ranges, with upper and lower bounds, along the
x-axis of the spectrum, and thus making up "bins" that are defined
for respective data extraction. Examples of such bins are shown
between adjacent vertical overlay lines in spectra shown in FIGS.
16A-B, 17C, and 17F (where top to bottom direction of a legend on
the right of FIGS. 17C and 17F corresponds with right to left
direction of "bins" in those FIGS., though as also alternatively
reflected with lead lines to respective chemical bin regions in
FIGS. 16A-B). The typical SRDIs of various biomarkers of interest
are elsewhere described herein, and as may be otherwise known in
the literature and applicable for a given application of the
present aspects in practice. In some cases, it is to be appreciate
that such bins may provide only an ability to find a certain
feature of the spectrum, e.g. a regional "peak", and this
information can then be used to determine and extract other
information (e.g. power under a peak region, which may be
determined to include spectral power around the peak that extends
outside of the respective "bin"). Furthermore, certain artifacts
may cause chemical shift error in the spectra despite corrections
provided in the signal processing. This data extractor may
recognize a certain feature in one respective SRDI bin, e.g. PG
peak, and then adjust the location for another target SRDI from
where it might otherwise be sought (e.g. based upon a prescribed
distance from the first recognized target peak, vs. fixed relative
locations for the SRDIs along the x-axis). In some embodiments, to
compensate for slight shifts in the spectrum (e.g., chemical shift
errors) after a regional peak is identified in a specified bin, the
bin and/or the spectrum can be shifted to align the regional peak
with the center of the bin, and an area under the curve can be
taken for a region (e.g., in the shifted bin) centered on the
located regional peak.
[0176] Once processed signal quality is confirmed, and spectral
data extraction is performed, diagnostic processing based upon that
extracted data may then be performed, as also per schematic flow
diagrams of FIGS. 9A-C. Such approaches are further developed below
by way of the present Examples, though it is to be appreciated that
various different specific diagnostic approaches, algorithms, uses,
etc. may be performed by one of ordinary skill without departing
from the other broad intended scopes of the current disclosure.
Nonetheless, for purpose of understanding of the present detailed
embodiments, the following bin region "limits" were used for
certain aspects of data extraction in the LA, AL, and PG regions of
acquired and processed DDD-MRS spectra for general purpose of most
data extracted and processed in the Examples: LA: 1.2 to 1.45; AL:
1.45 to 1.6; PG: 2.0 to 2.2.
[0177] It will also be understood that the DDD-MRS diagnostic
processor can be implemented in a variety of manners, such as using
computer hardware, software, or firmware, or some combination
thereof. In some embodiments, the DDD-MRS diagnostic processor can
include one or more computer processors configured to execute a
software application as computer-executable code stored in a
non-transitory computer-readable medium. In some embodiments, the
computer processor can be part of a general purpose computer. The
computer processor(s) used by the DDD-MRS diagnostic processor can
be the same computer processor(s) used by the DDD-MRS signal
processor, or it can be one or more separate computer processors.
In some embodiments, the DDD-MRS diagnostic processor can be
implemented using specialized computer hardware such as integrated
circuits instead of computer software. The DDD-MRS signal processor
may also be implemented by multiple computers connected, for
example, through a network or the internet.
EXAMPLES
Example 1
[0178] A DDD-MRS pulse sequence and signal processor were
constructed to incorporate various aspects of the present
embodiments disclosed herein and were used and evaluated in
clinical experience across a population of discs in chronic, severe
low back pain patients and asymptomatic control volunteers. Various
data extracted from features of interest along the acquired and
processed DDD-MRS acquisition series for discs evaluated in these
subjects were compared against control diagnoses for severe disc
pain vs. absence severe disc pain, in order to develop and
characterize a DDD-MRS diagnostic processor with the highest
possible correlation to the control diagnoses.
[0179] Methods:
[0180] Clinical Study Population: The study included 65 discs from
36 total subjects. Thirty-eight discs were from 17 patients with a
clinical diagnosis of chronic, severe low back pain (LBP group),
and 27 discs were from 19 asymptomatic volunteers (ASY Group). 25
discs in 12 of the LBP patients also received PD (PD Group)
sufficiently contemporaneous with the DDD-MRS exam to provide
appropriate comparison basis. All 65 discs were evaluated for
single voxel magnetic resonance spectroscopy pulse sequence and
data acquisition (DDD-MRS), and signal processor parameter
development of the new DDD-MRS approach. 52 discs from 31 subjects
were considered appropriate and used as controls for developing and
assessing the DDD-MRS diagnostic processor for diagnostic
application of the overall DDD-MRS system and approach. Thirteen
discography positive (PD+) discs from the PD Group were used as
positive control (PC) discs, and 12 discography negative (PD-)
discs from the PD Group plus all the ASY discs were used as
negative control (NC) discs. A breakdown summary analysis of
demographics among and between these groups under this Example is
shown in Table 2.
[0181] Study Design: Standard lumbar MRI was performed on all
subjects. PD performed within the PD Group was conducted by
discographers per their discretionary techniques, and in all cases
was performed blinded to DDD-MRS exam information. However, the PD+
criteria included a pain intensity score of greater than or equal
to 6 concordant to typical back pain on PD; less than or equal to
50 psi above opening pressure (where measured); and a negative
control PD- disc in the same patient (except one). All PD- discs
had less than 6 pain intensity scores per PD. Pain questionnaires,
including Oswestry Disability Index (ODI) and Visual Analog Scale
(VAS), were completed by all subjects, and the PD Group scored
significantly higher than the ASY Group according to both measures
as shown in FIG. 22 (PD Group VAS and ODI on left side of graph,
ASY Group VAS and ODI on right side of graph; VAS shown to left,
ODI shown on right, within each group). The DDD-MRS pulse sequence
and signal processor constructed according to the various present
embodiments herein was used for each series acquisition for each
disc, with data extracted from voxels prescribed at regions of
interest within nuclei of all discs included in the study. A 3.0 T
GE Signa MRI system and 8-channel local spine detector coil were
used with the DDD-MRS package and approach (lower 6 of the 8
channels activated for lumbar signal acquisition). Information
along spectral regions of the acquired DDD-MRS signals and
associated with various chemicals of interest were evaluated
against control diagnoses across the PC and NC groups.
[0182] Multi-variate logistic regression analyses were performed to
fit the dicotomous response (PC vs NC) to the continuous spectral
measures and develop a binary DDD-MRS diagnostic set of criteria
and threshold for determining positive (MRS+) and negative (MRS-)
pain diagnoses. A receiver operator characteristic (ROC) curve was
generated, and area under the curve (AUC) was calculated to assess
the accuracy of the developed test (FIG. 23). Five-fold
cross-validation was performed to assess the generalizability of
the predictive relationship (FIG. 24).
[0183] DDD-MRS diagnostic outcomes for each disc were based on a
single number calculated via the developed set of criteria based
upon four weighted factors derived from regions of the acquired MRS
signals and associated with three chemicals--PG, LA, and alanine
(AL). It is noted, however, that LA and AL regions are relatively
narrow and immediately adjacent to each other, and in some cases
the true respective signals representing these actual chemical
constituents may overlap with each other and/or into the adjacent
region's location. Furthermore, either or both of the LA and AL
regions may also overlap with possible lipid contribution, which
was believed to be observed in some cases (which may include signal
from adjacent tissues such as bone marrow of bordering vertebral
body/s). Positive numerical threshold results were assigned "MRS+"
as severely painful, and negative results were assigned "MRS-" as
not severely painful. Accordingly, the threshold for severely
painful vs. otherwise non-painful diagnostic result is zero (0).
The set of diagnostic criteria used to determine MRS+ vs. MRS-
diagnostic values around this threshold with the most robust
statistical correlation and fit to the control data observed across
the disc population evaluated for this purpose is summarized as
follows:
Threshold=-[log(PG/LA*(0.6390061)+PG/AL*(1.45108778)+PG/vol*(1.34213514)-
+LA/vol*(-0.5945179)-2.8750366)];
[0184] wherein: [0185] PG=peak measurement in PG region, AL=peak
measurement in AL region, LA=peak measurement in LA region, and
vol=volume of prescribed voxel in disc used for MRS data
acquisition.
[0186] The distribution of DDD-MRS results according to these
calculated thresholds were compared against all PC and NC
diagnoses, PD results alone, and portion of the NC group
represented by the ASY group alone. Sensitivity, specificity, and
positive (PPV) and negative (NPV) predictive values were also
calculated per control comparisons.
[0187] Further aspects of the statistical methods herein applied,
with respect to identifying diagnostic algorithm and also
evaluating resulting data, are described in more detail below with
respect to similar approaches also taken in subsequent Examples 2
and 3.
[0188] Results:
[0189] DDD-MRS data demonstrated a strong correlation with the
clinical diagnoses (R.sup.2=0.89, p<0.00001), with Receiver
Operator Characteristic (ROC) analysis yielding an area under the
curve (AUC) of 0.99 (FIG. 23) and cross-validation through
partition analysis resulting in only deminimus variance in the
R.sup.2 (FIG. 24). Tables 3 and 4, and FIGS. 25A-27, show various
aspects of the resulting clinical comparison data for DDD-MRS vs.
control diagnostic data, which data and comparisons are further
described as follows.
[0190] DDD-MRS results, with respect to binary MRS+ and MRS-
diagnoses, correctly matched binary PC and NC diagnoses of
painful/non-painful for 50/52 (96.2%) discs evaluated across the PD
and ASY groups. Of the 13 MRS+ discs, 12 discs were from the PC
group (PPV=92%). Of the 40 discs that were MRS-, 39 were from the
NC group (NPV=97%). DDD-MRS sensitivity was about 92% and
specificity was about 97%. Mean DDD-MRS results for the PC and NC
groups were 0.97.+-.0.77 and -1.40.+-.0.65 (R.sup.2=0.89,
p<0.00001, FIG. 25B). As shown in FIG. 26, DDD-MRS results
matched PD results in 23/25 (92.0%) discs of the PD Group: 12/13
(96.2%) of PD+ and 11/12 (91.7%) of PD-. Mean DDD-MRS algorithm
results for PD+ and PD- groups were 0.97.+-.0.77 and -1.39.+-.0.72
(p<0.00001) (FIG. 25B). DDD-MRS results also correlated with PD
pain intensity scores (R.sup.2=0.73)(not shown). DDD-MRS results
matched all 27/27 (100%) NC results represented by the ASY group
(FIG. 26). The mean DDD-MRS algorithm results for the ASY group
were -1.4.+-.0.63, which differed significantly vs. PD+
(p<0.0001), but were not significantly distinguishable vs. PD-
results (p=0.46) (FIGS. 25A-B).
[0191] As shown in FIGS. 28-29, the DDD-MRS results according to
this study of this Example provided highly favorable improvement
vs. the diagnostic accuracy typically attributed to MRI alone for
diagnosing painful vs. non-painful DDD. More specifically, FIG. 28
(two bars on right side of graph) shows a comparison of the AUC for
MRI alone vs. MRI+DDD-MRS, per meta analysis of previously reported
AUC data for MRI for this indication. This is further compared in
the graph against a recent study reporting AUC for MRI alone vs.
MRI+PROSE for prostate cancer diagnosis (as compared to
histopathological diagnosis of biopsy samples), where no
significant improvement was shown by the additional inclusion of
PROSE application of MRS within the MR-based diagnostic regimen.
While the prostate data reflected within the graph reflects a
larger relative population of samples in multi-center study, and
the DDD-MRS pain diagnostic results shown reflects a smaller
population within single center experience, the dramatic relative
improvement presented by the DDD-MRS approach in the single center
experience is expected to carry over to a significant degree into
larger, multi-center context for this application. Further to FIG.
29, the results of this study additionally show improvement to
positive and negative predictive values by enhancing standard MRI
alone with the addition of the DDD-MRS diagnostic--per meta
analysis of the current data vs. previously published data for MRI
for this purpose.
[0192] While the other information described herein is clearly
sufficient to demonstrate the remarkable utility of the present
embodiments in operation for the indicated purpose of this Example,
further supportive information is also provided as follows. The
DDD-MRS diagnostic exam was also evaluted for and demonstrated
robust repeatability, as reflected in FIG. 30A. More specifically,
FIG. 30A shows DDD-MRS diagnostic algorithmic results according to
this Example for eight (8) L4-L5 discs in eight (8) asymptomatic
pain free volunteers examined twice--each on 2 separate dates, with
trend between sequential results for each disc shown in respective
lines between columns (1) and (2) along the x-axis of the graph.
These were all negative diagnostic results, indicating pain free
diagnosis according to the exams, with relative repeatability and
little variance between exams on average between the group and
individually for the vast majority of the samples (with one obvious
outlier demonstrating more variance than the others, though still
nonetheless representing a repeatable diagnostic result as
negative). In addition, as shown in FIG. 30B, the measured ratios
between metabolite regions for PG and a combination of LA and AL
(alanine) or "LAAL" were compared as per spectral acquisitions and
extracted regional data measurements in vivo, against measurements
taken for the same chemical regions but via 11 T HR-MAS
spectroscopy ex vivo after surgical removal for pain treatment.
These comparisons were highly correlative, with R.sup.2=0.98,
demonstrating the robustness of the measurements taken in vivo by
ex vivo validation measurements for the same disc material.
[0193] Certain benefits provided by the DDD-MRS processor for
post-processing acquired MRS signals were also evaluated across a
sub-set sampling of the DDD-MRS data derived from the clinical
population under this study of this Example. In particular, for
each series acquisition the SNR of the processed DDD-MRS signals
("DDD-MRS spectra/spectrum") was characterized, and compared
against the 6 channel average, non-phase or frequency corrected, GE
Signa output spectra as acquired "pre-processing" according to the
present embodiments (e.g. "input combined spectra/spectrum"). This
SNR characterization and comparison exercise was conducted as
follows.
[0194] A freeware digitization program (WinDIG.TM., Ver 2.5,
copyright 1996, D. Lovy)) was used to digitize both final DDD-MRS
results and "screen shot" images. The "screen shot" images were
reverse-imaged using MS Paint prior to digitization. The output of
the digitizer program is an array of integers in a comma-separated
values (CSV) file format. The CSV data files were imported to
Microsoft.TM. Excel.TM. and re-plotted as shown in FIGS. 31A-B. A
"region of interest" on the chemical shift (CS) axis (x-axis)
pertaining to metabolite proteoglycan (PG, CS=2.11 PPM) was deemed
to be the "signal". A region of interest to the far right (CS=0.5
PPM) which would not typically contain any spectral activity was
deemed to be the "noise". In the event there was not a significant
spectral peak in the PG region which is the often the case on pain
patient discs, then the lactate/Lipid region of interest (CS=1.33
PP) was used as the signal. The "ranges of interest" were visually
determined on both images resulting in sections of the data array.
The SNR of a waveform is expressed as:
10*log.sub.10(RMS signal/RMS noise).
[0195] The RMS value was calculated by taking the sum of squares of
the data section, calculating the mean of the sum of squares, and
then taking the square root of the mean. Since the spectra are
power amplitude plots, the log base 10 of the ratio of the RMS
values is then multiplied by 10 to generate the SNR in dB.
[0196] For further understanding of this approach and examples of
the digitized spectra and information extracted therefrom, FIG. 31A
shows a digitized DDD-MRS spectral plot and accompanying SNR
information, whereas FIG. 31B shows similar views for a digitized
pre-processed all channel (n=6) averaged output spectral plot
output from the respective MR system and related SNR information
for the same acquisition series (without processing according to
the present signal processing aspects of the present
disclosure).
[0197] These pre- and post-processing SNR results are shown in
FIGS. 31C-F. More specifically, FIG. 31C shows the calculated SNR
for the pre- and post-processed spectra, with significant majority
of the pre-processed spectral SNR shown on the left side histogram
distribution of the plot falling below 5 (and also much of the data
below 3), but with a significant majority of the post-processed
spectral SNR shown on the right side histogram distribution of the
plot falling above 3 (all but 1) and even above 5 (all but just 2).
A typical accepted SNR range for confidently measuring chemical
constituents from an MRS plot is in many cases over 5, though in
many cases may be for any data over 3--such that below these
thresholds may be "unquantifiable" or "immeasurable" at least per
such standards (if applied). In such an application of these
thresholds, it is clear that a significant portion of data acquired
pre-processing according to the present embodiments is not
generally useful for interpretting signal regions of interest,
whereas these data as post-processed herein become quite
consistently useful. In fact, as shown in FIG. 31D, the average SNR
across the signals evaluated for this comparison exercise was:
about 3 (e.g. well below 5) pre-processing, and about 13 (e.g. well
above 5) post-processing (p<0.001). As per the ratio of post-
vs. pre-processed signals further shown in FIG. 31E, in all cases
compared the post-processed signals were higher SNR than
pre-processing, generally along a range between 2 to 8 times higher
SNR (with only one point falling below 2.times. improvement, though
still about 50% improved). As further evaluated (e.g. FIGS. 31F-H),
the mean absolute improvement was about 10 dB, the mean ratio
improvement was over 4.times., and the mean % improvement was well
over 300% in converting from pre- to post-processed signals
according to the present embodiments.
[0198] For further illustration of the beneficial results
demonstrated by the DDD-MRS diagnostic exam, FIGS. 32A and 32B show
two different examples of DDD-MRS diagnostic display results for
two different patients in the clinical study featured under this
Example 1. These patients have similar disc degeneration profiles
as seen on the MRI images, with dark disc at L5-S1 and relatively
healthy discs revealed above at L4-L5 and L3-L4 in each patient. As
also shown in each of these figures, both patients also had
positive discogram results at L5-S1. However, as also shown in
these two comparison Figures, the patient featured in FIG. 32A had
a negative discogram result (e.g. non-painful diagnosis) at L4-L5,
whereas the patient featured in FIG. 32B had a positive discogram
result (e.g. painful diagnosis) at that level--despite having
similar disc degeneration profile. As a consequence of both exams,
with modern discography technique guidelines indicating requirement
for a negative control disc before positive levels may be accepted
results, the patients each had another negative discogram done at
the L3-L4 (FIG. 32A) and L4-L5 (FIG. 32B) levels, respectively, to
provide the required negative control level. As an awarded recent
study has shown discography significantly increases disc
degeneration and herniations rates, the result of both of these
studies, if followed for directed intervention, would have resulted
in treating the positive discogram levels, but not the negative
discogram levels--leaving those untreated levels in place to
potentially accelerate in degeneration and toward possible
herniations. As shown in these Figures, the non-invasive DDD-MRS
results matched these invasive discography results at all disc
levels. The DDD-MRS approach provides the distinct benefit of
providing the diagnostic information required, while leaving all
discs uncompromised due to the non-invasive nature of the
approach.
[0199] Discussion:
[0200] The differentiation of painful and non-painful lumbar
degenerative discs is an important goal in the accurate assessment
of pain generators, and in guiding clinical management of patients
with lumbar degenerative disc disease. The novel application of
Magnetic Resonance Spectroscopy developed and evaluated under this
study proposes a non-invasive, objective, and quantifiable measure
of the chemical composition of the lumbar intervertebral disc. The
MRS diagnostic algorithm developed and used in this study
demonstrates a high degree of sensitivity in identifying patients
with a clinical assessment of lumbar discogenic pain and a positive
discogram, and a high degree of specificity in identifying levels
that are not painful, without any false positive results observed
in asymptomatics. This study developing, uniformly applying, and
characterizing the DDD-MRS diagnostic approach retrospectively
across the study population evaluated herein is quite encouraging.
Cross validation also performed on the results predicts the
approach is generalizable to broader population, as may be readily
confirmed in additional prospective study in more subjects, and as
may be conducted by one of ordinary skill.
Example 2
[0201] The 52 disc clinical data set evaluated under the DDD-MRS
system embodiments of the present disclosure and associated with
Example 1 was further expanded with additional new subjects
examined for a total of 74 discs, with additional signal processing
developments performed and diagnostic algorithm development
conducted to determine the optimal correlation to the expanded data
set. The results of this algorithm development and analysis was
then applied to an additional 5 discs in new asymptomatic control
volunteers prospectively, for 79 total discs later evaluated.
[0202] Standard logistic regression procedures were used to develop
a second generation linear regression model between disc variables
obtained from DDD-MRS acquisitions and processed by the DDD-MRS
signal processing engine, to disc pain status (pain/no-pain entered
as a categorical variable based on provocative discography). MR
spectra (in-phase real power format) from a population of 74 discs
(15 painful and 59 asymptomatic) were used for classifier
development and cross-validation partition analysis. The DDD-MRS
data demonstrated a strong correlation with the clinical diagnoses
(R.sup.2=0.76, p<0.00001) with an ROC analysis yielding an AUC
of 0.99. Cross-validation through partition analysis resulted in
only small variance in R.sup.2.
[0203] Materials and Methods
[0204] All statistical analyses were performed using JMP (version
7.0, SAS). Standard logistic regression procedures were used to
relate the disc variables (proteoglycan, lactate, and alanine
spectral peaks entered as continuous variables) to the disc pain
status (pain/no-pain entered as a categorical variable).
Discography performed according to ISIS Guidelines was used as the
reference standard for pain status in low back pain patients. Discs
from asymptomatic volunteers were assumed negative. The discography
status and disc variables were entered into an excel spreadsheet
and imported into JMP.
[0205] The DDD-MRS diagnostic algorithm was determined in a
three-stage process.
[0206] First, the terms were limited to spectral features related
to proteoglycan, lactate and alanine because these were shown to be
important classifiers in prior studies (Keshari, 2008. "Lactic acid
and proteoglycans as metabolic markers for discogenic back pain."
Spine 33(3): 312-317), and fit with biologically-plausible theories
for discogenic pain generation. In addition, normalized values for
these factors were considered. To provide an estimate of metabolite
concentration, the spectral measures were divided by the region of
interest (ROI) volume. Also, given signal strength may vary with
ROI depth, subject body mass index (BMI) was also considered as a
normalizing factor. This was done by taking the BMI for a subject
associated with a given disc sample being evaluated divided by the
average BMI across the data set used for the logistic regression
modeling. Also as raw signal region values represent "amounts" of
respective chemicals associated such regions, dividing such values
by voxel volume may provide a surrogate approach to more closely
approximating "concentration" for such chemicals (amount/unit
volume)--which as biomarkers as mediators to a pain cascade are
likely more relevantly assessed as concentration. For example,
lactic acid is more relevant to disc tissue acidity, which is
believed to be a pain generator, on a concentration basis vs. total
amount in the tissue. Accordingly, voxel volume adjustment for a
signal measurement simply involved dividing the measured factor or
parameter by the voxel volume.
[0207] In the second step, the form of the factor dependence was
estimated using Screening Platform in JMP. Within the Screening
Platform, the dependent variable was chosen to be pain status, and
the candidate independent variables were chosen to be proteoglycan,
lactate, and alanine (either raw values or values normalized by
voxel volume and/or BMI). The Screening Platform then identified
candidate terms with associated p-values. These would include
either individual factors, or products of multiple factors. Terms
with p-values less than 0.05 were selected as candidates for
further consideration.
[0208] In the third step, candidate terms from the Screening
Platform were entered as independent predictors in the Logistic
Platform of JMP. This platform was used to conduct logistic
regression analysis to identify statistically-significant terms
plus their parameter estimates. The Logistic Platform fits the
probabilities for the response category (pain/no-pain status) to a
set of continuous predictors (metabolite terms). The fit quality
was judged by the coefficient of determination R.sup.2 and the
p-value. In an ad-hoc stepwise fashion, candidate terms were
brought into the Logistic model to judge their influence on model
performance.
[0209] Because some metabolite data are not normally distributed,
log and square-root transformations of the candidate terms were
also considered. Candidate terms with p-values less than 0.05 were
removed from the model. The Logistic regression output provided
parameters that are multipliers for each term plus an intercept
term. These formed an algorithm that provides a continuous number
that, if greater than zero would indicate a painful status, and if
less than zero would indicate a non-painful status.
[0210] As an additional summary of the discriminatory accuracy of
the Nociscan diagnostic algorithm, generated standard
Receiver-Operator curves (ROC) that are plots of sensitivity versus
specificity across a rank ordered list of study discs. The area
under the ROC curve (AUC) was used to judge the algorithm accuracy.
The AUC is the probability that test results for a
randomly-selected painful disc and non-painful disc will be rank
ordered correctly. Additionally, continuous correlation procedures
were used to judge whether the output of the diagnostic algorithm
correlates with VAS score, disc degeneration grade, and the
discography pain intensity.
[0211] Results/Data
[0212] Using the aforementioned procedures, a diagnostic algorithm
was developed using a 74 disc (15 pain, 59 control) population. The
best-fit linear regression equation result using this approach was
as follows:
Score = - 4.6010405 + 1.58785166 ( BLA ) - 0.081991 ( VBLAAL -
29.3125 ) * ( VBLAAL - 29.3125 ) + 0.01483355 ( PG / MAXLAAL -
7.14499 ) * ( PG / MAXLAAL - 7.14499 ) * ( PG / MAXLAAL - 7.14499 )
+ 0.1442603 ( MAXLAAL / vol - 16.1202 ) * ( VBLAAL - 29.3125 ) -
0.0008879 ( VBLAAL - 29.3125 ) 2 * ( MAXLAAL / VOL - 16.1202 )
##EQU00001##
where BLA is the BMI corrected LA spectral peak, VBLAAL is the ROI
volume and BMI normalized sum of the LA and AL spectral peaks,
MAXLAAL is the maximum of either the LA or AL peaks, and PG is the
n-acetyl spectral peak.
[0213] The present linear regression equation of this Example 2
uses similar features as its predecessor such as chemical peak
values and peak ratios, but in addition uses features normalized
for voxel volume and BMI (e.g. "VB" designating both). Increased
body fat (increased BMI) will reduce chemical peak values because
the voxel is physically further away from the RF coil resulting in
reduced signal strength and chemical peak values. The BMI value is
adjusted (normalized) by the mean BMI of the population. The
adjusted BMI value thus applies a proportional "gain" to chemical
peak values otherwise reduced by large BMI.
[0214] Similarly small voxel volumes will reduce the chemical peak
values and the inverse of voxel volume is applied as a "gain"
factor. In addition to normalization, the equation also defines a
two new features. The first consists of combined regions of
interest (ROI) lactate (LA) and alanine (AL) regions to create
LAAL. The second is a region called MAXLAAL whose value is the
greater of the two regions.
[0215] A final development to the diagnostic engine is the
application of an indeterminate band to the classification process.
This band lies between the highly probable pain and pain free
states and is statistically determined from the distribution of the
two disc populations. Diagnostic scores that fall within this band
are determined to be procedural failures because of the low
probability to diagnose either way. When applied this band results
in one false negative (a positive discography disc diagnosed as
pain free).
[0216] Results and Discussion
[0217] A second generation diagnostic classifier using DDD MRS
acquisition data as processed by the Nociscan signal processing
average has been developed using an increased disc population (from
n=52 to n=74). The incorporation of BMI adjustment per each
sample's BMI relationship to a mean population BMI, voxel volume
adjustment to more closely approximate concentration aspects of the
target biomarker metabolites, and the use of combined regions of
interest (LAAL, MAXLAAL), has resulted in a linear regression
equation with a significant improvement over the otherwise highly
accurate first generation linear regression equation, with
(R.sup.2=0.89, p<0.00001) with an ROC analysis yielding an AUC
of 0.99.
[0218] For further illustration, FIG. 34A shows the distribution of
this algorithmic data across the combined data set via the
formulaic algorithmic "score" for each disc plotted against
designation of the discs as fitting within the positive controls
(positive discogrammed discs, or PD+), negative controls (negative
discogrammed discs in pain patients, or PD-, combined with discs
from asymptomatic control volunteers or "ASY" discs), and versus
these two negative control sub-populations alone. This also shows
the application of the statistically guided "indeterminate" band
bordering the "zero" line and where n=5 discs fall, with positive
test results above the upper limit of that band (n=12), and
negative results below the lower limit of that band (n=63, of which
62 were negative controls and 1 was a positive control disc).
Excluding these indeterminates as "procedural failures" excludes
5/79 discs or only 6% of the test population, while remaining 94%
are considered procedural successes for making a confident
diagnosis. Among these 94% procedural successes, the results were
99% accurate with 73/74 overall match to controls (only 1
mismatch), R.sup.2=0.91, p<0.0001, and AUC=0.99. These more
detailed breakdown for the matches against controls (e.g. positive
match to positive discogram, or negative match to negative
discogram or discs from asymptomatic subjects) are as follows:
12/13 (92%) of Positive Discography; 13/13 (100%) of Negative
Discography; 48/48 (100%) of Asymptomatics--thus there were no
false positive results in 62 negative combined controls, and only
1/13 presumed false negative result among 13 positive control
discs. These results further provide the following performance
characteristics typically used to evaluate a diagnostic platform:
92% Sensitivity, 100% Specificity, 100% Positive Predictive Value,
and 98% Negative Predictive Value.
[0219] For still further illustration of another highly beneficial
view of these highly accurate results of this current approach of
Example 2 to this test group, FIG. 34B shows another distribution
of the same results for this same data set, but as converted to %
probability prediction a disc is painful (as generated by rank
ordering of the logistic regression analysis and results). As shown
in this Figure, a region between about 80% probability and about
20% probability prediction of pain corresponds with capturing the
same n=5 discs indeterminate zone discs of the other view of the
data distribution in the prior Figure, with greater than about 80%
probability criteria capturing all of the n=12 same positive test
results (all matching positive controls), and less than about 20%
probability criteria capturing all of the n=63 same negative test
results (62 matching all of the negative controls, and 1 a positive
control and thus representing the same single presumed false
negative test result).
Example 3
[0220] Standard Logistic Regression Procedures were Used to Relate
Disc Variables obtained from DDD-MRS acquisitions and processed by
the Nociscan signal processing engine to disc pain status
(pain/no-pain entered as a categorical variable based on
provocative discography). Acquired DDD-MRS spectra were processed,
analyzed, and presented post-processing for diagnostic purposes in
absorption mode- vs. real-part squared power format of prior
Examples. The spectral acquisitions were the same and from the same
population of 79 discs in 42 subjects (15 painful and 64
asymptomatic) as featured in Example 2, used here for further
algorithmic classifier development. Certain signal quality criteria
were also used in this Example 3 to determine each of three
classifications of acquired results--namely recognizing the
following sub-groups: (1) a first spectral group with clearly
apparent lipid signal (then given its own logistic regression model
and resulting algorithm), and (2) a second spectral group absent
any obvious lipid signal that was still further sub-classified into
still further sub-groups: (2)(a) spectra with significant PG/LAAL
peak ratios over a determined criteria threshold, and (2)(b) the
remaining non-lipid signals not meeting this criteria also given
its own second logistic regression model and resulting algorithm.
The three classifier equations that were developed resulted in 100%
procedural success and 100% separation for differentiating painful
from non-painful discs in all 79 discs evaluated.
[0221] Purpose
[0222] The purpose of this study was to evaluate still further
potentially valuable approaches for developing a robust classifier,
including as using features extracted from absorption spectra as
opposed to features formerly extracted from in phase real power
spectra, and also to evaluate a different approach for classifier
modeling based upon a serial application of a limited few tests
applied to what appeared to be unique sub-populations among data.
Absorption power format is the traditional method of displaying
spectra. In phase real power spectra are comprised of the square of
the real component of each spectral point. This format presents
only positive going spectra with minimal baseline shift. This
mitigates the need to fit a spline curve to the baseline as well as
makes the spectra appear more peaked. The overall effect is to
enhance the apparent signal to noise ratio (SNR) and remove the
variability associated with fitting a baseline to the spectral plot
for the purpose of making spectral peak and area under the curve
(AUC) measurements. Nonetheless, the current absorption spectra
approach of this Example 3 is more common to typical MRS analysis
in other applications, and may be more relevant for biomarker
assessment in certain cases, vs. previous classifier development of
prior Examples that has been done using spectra presented in
in-phase, real power squared format.
[0223] Materials and Methods
[0224] A comparison of SNR for post-processed versus pre-processed
DDD-MRS spectra acquired per this Example was performed similarly
as featured above for Example 1 data set (e.g. FIGS. 31A-H), except
using absorption spectra for both pre- and post-processed data, and
per the expanded clinical data set represented in this Example 3.
These were otherwise analyzed similarly as was done in those prior
Figures for the prior Example 1.
[0225] All statistical analyses were performed using JMP (version
7.0, SAS). Standard logistic regression procedures were used to
relate the disc variables (proteoglycan, lactate, and alanine
spectral peaks entered as continuous variables) to the disc pain
status (pain/no-pain entered as a categorical variable). A
significant majority of the discography was performed according to
ISIS Guidelines and was used as the reference standard for pain
status of `positive control` discs in low back pain patients. Discs
from asymptomatic volunteers were assumed negative, and were
combined with negative discography discs from the pain patients as
the negative control group presumed to be non-painful. The
discography status and disc variables were entered into an excel
spreadsheet and imported into JMP.
[0226] The terms in each of the two sub-groups (1) and (2)(b) where
logistic modeling was applied for algorithm development were
determined in a three-stage process. The first step choosing
spectral features of interest for analysis, and corresponding to
the PG, LA, and AL biomarker chemicals, proceeded as per prior
examples, and including BMI and voxel adjustment as described for
Example 2, with the following difference in this Example 3 that
absorption spectra were used for the data extraction and subsequent
inputs into the diagnostic processor.
[0227] In the second step, the form of the factor dependence was
estimated using Screening Platform in JMP. Within the Screening
Platform, the dependent variable was chosen to be pain status, and
the candidate independent variables were chosen to be proteoglycan,
lactate, and alanine (either raw values or values normalized by ROI
volume and/or BMI). The Screening Platform then identified
candidate terms with associated p-values. These would include
either individual factors, or products of multiple factors. Terms
with p-values less than 0.05 were selected as candidates for
further consideration.
[0228] In the third step, candidate terms from the Screening
Platform were entered as independent predictors in the Logistic
Platform of JMP. This platform was used to conduct logistic
regression analysis to identify statistically-significant terms
plus their parameter estimates. The Logistic Platform fits the
probabilities for the response category (pain/no-pain status) to a
set of continuous predictors (metabolite terms). The fit quality
was judged by the coefficient of determination R.sup.2 and the
p-value. In an ad-hoc stepwise fashion, candidate terms were
brought into the Logistic model to judge their influence on model
performance.
[0229] Because some metabolite data are not normally distributed,
log and square-root transformations of the candidate terms were
also considered. Candidate terms with p-values less than 0.05 were
removed from the model. The Logistic regression output provided
parameters that are multipliers for each term plus an intercept
term. These formed an algorithm that provides a continuous number
that, if greater than zero would indicate a painful status, and if
less than zero would indicate a non-painful status.
[0230] As an additional summary of the discriminatory accuracy of
the Nociscan diagnostic algorithm, generated standard
Receiver-Operator curves (ROC) that are plots of sensitivity versus
specificity across a rank ordered list of study discs. The area
under the ROC curve (AUC) was used to judge the algorithm accuracy.
The AUC is the probability that test results for a
randomly-selected painful disc and non-painful disc will be rank
ordered correctly. Additionally, we used continuous correlation
procedures to judge whether the output of the diagnostic algorithm
correlates with VAS score, disc degeneration grade, and the
discography pain intensity.
[0231] In context of the aforementioned methods and procedures
applied to previous classifier development and those receiving
logistic regression modeling in this current Example, a data
partition approach was implemented based on certain spectral
features observed in the current dataset. First, discs with
perceived lipid signal in the acquired DDD-MRS spectra were
partitioned into Group A (n=10). This was given its own logistic
regression modeling as test #1. Next, because many negative
non-painful discs were observed to have uniquely strong n-acetyl
peak (PG) and weak lactate (LA) and/or alanine (AL) peaks, the
PG/LAAL ratios for the remaining non-lipid disc population (n=68)
were evaluated between the positive and negative control groups. A
cut-off in a go/no-go voting method approach of test #2 for
`clearly negative` discs was identified at PG/LAAL peak ratios
above 1.81 to create Group B successes for negative results as
non-painful (n=52 of 68 discs evaluated in the non-lipid
population). The third data analysis and test portion, Group C
(n=16, a subset of non-lipid Group B that did not meet the test #2
criteria as having PG/MAXLAAL <1.85) were analyzed also using
the logistic regression modeling per the three-step process defined
above. Four statistically-significant terms and their parameter
estimates were identified by the Logistic Regression Platform: ROI
(e.g. voxel volume or VV) and BMI adjusted LA absorption peak; VV
and BMI adjusted AL absorption peak; VV and BMI adjusted AL AUC
(area under the curve) or "ALAUC"; and square root of the VV and
BMI adjusted N-acetyl AUC or "NAAAUC").
[0232] Finally with respect to the DDD-MRS diagnostic processor
aspects of the present Example, the spectra with suspected lipid
contamination (Group A) were also analyzed using the three-step
analysis procedure. This resulted in two terms that separated
positive from negative discs: the square root of the VV and BMI
adjusted LA peak, and the VV and BMI adjusted ratio of n-acetyl to
LAAL. When taken together, the partition plus logistic regression
approach success fully separated all negative from all positive
discs.
[0233] Results/Data for Absorption Spectra SNR
[0234] The SNR evaluation of the post-processed versus
pre-processed absorption spectra plots per this Example are shown
in FIGS. 34A-F, and demonstrate signficant SNR increase via the
DDD-MRS signal processor aspects deployed for this data set in the
Example 3, and also shows vast majority of the resulting signals to
have sufficiently robust SNR for target regional chemical signal
feature measurements. More specifically, FIG. 34A shows vast
majority of the post-processed SNR above 3 (except only 2 cases),
and in fact over 5, though much of the pre-processed spectra were
below these levels. FIG. 34B shows the average pre-processed SNR of
only slightly above 4, while the average post-processed SNR was
about 8 and nearly 2/3 of the post-processed SNR of the real-part
squared approach taken in the prior Example despite that approach
squaring the signal:noise values. FIG. 34C shows the vast majority
of the individual points were improved (e.g. ratio of SNR of post-
vs. pre-processed signals), but for only a few (n=3) which were
further observed to be quite high SNR to begin with, and with FIG.
34D showing about a 3.5 dB average SNR increase or about 2.2.times.
(FIG. 34E) versus the pre-processed SNR.
[0235] As for the DDD-MRS diagnostic processor developed and
evaluated per this Example, the best fit linear regresion equations
extracted from the absorption spectra are shown as follows:
[0236] Group A, Test #1:
Score = - ( - 335.51971 + 0.00010632 * ( LAVVBMI ) 2 + 873.744714 *
( PG / ( LAALVVBMI ) ) ) ; ##EQU00002##
where LAVVBMI equals the voxel volume and BMI adjusted LA peak
value.
[0237] Group B, Test #2:
Score=-(-1.4959544+1.72223147*(PG(MAXLAAL)));
where PG/MAXLAAL equals the PG peak value divided by the maximum
peak value of the LAAL region.
[0238] Group C, Test #3:
Score = - 1 * ( - 134.40909800961 + 3.96992556918043 * LAVVBMI -
2.6198628365642 * ALVVBMI + 113.683315467568 * ALAUCVVBMI -
149.65896624348 * SQRT ( PGAUCVVBMI ) ) ; ##EQU00003##
where LAVVBMI is the voxel volume and BMI adjusted LA peak value,
ALVVBMI is the voxel volume and BMI adjusted AL peak value,
ALAUCVVBMI is the AL region area under the curve as voxel volume
and BMI adjusted, and PGAUCVVBMI is the PG region area under the
curve as voxel volume and BMI adjusted.
[0239] Results/Data and Discussion--Diagnostic Processor
[0240] The default model used by JMP is to distribute data around
0. Results will typically provide negative results above 0, and
positive results below 0. However, as this is inverse to logical
presentation to match the classifications, and as in prior
Examples, the negative of the classifier outputs are taken so that
positive scores are associated with positive clinical tests for
pain and negative scores are associated with non-painful discs.
[0241] The partitioning of spectral acquisitions based on the
presence of lipid signal and on "clearly non-painful" spectral
attributes (PG/MAXLAAL>1.81), as taken from absorption spectra,
distinguish this classifier approach from previous efforts. The
logistic regression modeling of the resulting sub-groups also
provide different algorithms and varied specific factors as a
result. Group A contains spectra with sufficient lipid (lipid peak
at 1.3 PPM) that prevents the discrete characterization other
chemical components such as PG, LA and AL. It is noted that upon
evaluation of the absorption spectra results for this example, one
acquisition or n=1 of n=79 total overall discs initially to be
evaluated, was not considered to have sufficient signal quality
(e.g. SNR too low) for robust diagnostic processing and thus
excluded from that stage of processing, with resulting population
of n=78 evaluated diagnostically of n=79 attempted (e.g. 99%
procedural success, and <1% procedural failure rate due to low
SNR processed acquisition).
[0242] An example of a Group A spectra including suspected lipid
signal from an asymptomatic control L5-S1 disc is shown in FIG. 35.
Source of lipid in a given signal is not known, and may come from
several different sources. Lipid signal is often observed however
to result from capturing lipid-enriched vertebral body endplates by
the voxel prescribed, and often (though not always) in an oblique,
severely compromised (crushed) L51 disc. Another source of lipid
contamination may be due to patient movement during the MRS
acquisition, also involving end-plate artifact. It may also come as
out of voxel signal in some cases, and may in fact come
appropriately from within discs. Nevertheless, the prior grouping
of signals with and without lipid was successful in accurately
diagnosing most all discs, including all spectra with lipid across
all the Examples. In this Example 3, spectra of Group A (n=10) was
separated (100%) into painful and non-painful groups per test #1
and an associated probability of being painful is shown in FIG. 36.
It is also observed among these spectra in this Group A that the
presence of a sufficiently strong PG component in combination with
lipid signal is likely related to sufficiently correlating with
non-painful discs to provide the resulting reliable differentiation
between positive and negative control groups.
[0243] The second partition was applied to the disc population
without lipid contamination (n=68). By visual observation of
spectra across this population it was noted that all discs with
PG/MAXLAAL value exceeding about 2. In further analysis, a
threshold value of 1.85 was identified to partition only
non-painful negative control discs above the threshold, and
completely isolating the painful positive control disc population
below the threshold, but while also including other non-painful
negative control discs below this threshold. This PG/MAXLAAL
partition analysis is shown in FIG. 37A. The more statistically
robust linear regression model of test #2 derived and applied to
Group B (n=68) is shown in FIG. 37B. The painful vs. non-painful
segregations remain similar to the immediately previous analysis.
The % probability painful converted format of this data
distribution is shown in FIG. 37C, with threshold nearly
approaching 20% differentiating all the same negative control
discs, and none of the positive control discs in the group, below.
(Note: Probability of being non-painful=1-pain probability).
[0244] The sub-population of discs from Group B with a
PG/MAXLAAL<1.85 are partitioned into the third Group C (n=16),
with the linear regression test #3 derived from Group C resulting
in the data distribution shown in FIG. 38. There is 100% separation
between these remaining positive and negative control discs in this
final Group C.
[0245] The ultimate result of this applied step-wise partitioning
and logistic regression diagnostic algorithm approach was 100%
separation between known painful vs. non-painful results, across
all of the 78 discs evaluated diagnostically.
[0246] Nonetheless, it is to be appreciated that other specific
diagnostic algorithmic approaches may be applied and also achieve
significantly robust results. As one example, a hybrid linear
regression equation consisting of terms from test #2 and test #3
(derived from Groups B and C respectfully) is provided by algorithm
test #4 for Group B, shown partitioned in FIG. 39. This approach as
evaluated here also still retains all 78 test discs in the overall
population while resulting in 76/78 overall match to controls per
only 2 presumed false negative values (two PD+ discs indicated
instead as negative DDD-MRS tests as being pain free), and no false
positive results. The hybrid linear regression equation
coefficients range within two orders of magnitude of each other and
are fully normalized or proportional, characteristics that make for
a robust classifier. The hybrid equation mitigates the need to
perform the PG/MAXLAAL partition.
[0247] Group B, Test #4:
Score = - 6.94869 + 0.05035 * LAVVBMI - 0.028534 * ALVVBMI -
0.51761 * SQRT ( PGAUCVVBMI ) + 0.36976 * ALAUCVVBMI + 4.04875 * PG
/ MAXLAAL ; ##EQU00004##
where LAVVBMI=LA peak adjusted by voxel volume and BMI, ALVVBMI=LA
peak adjusted by voxel volume and BMI, PGAUCVVBMI is the PG area
under the curve (AUC) adjusted by voxel volume and BMI, ALAUCVVBMI
is the AL area under the curve (AUC) adjusted by voxel volume and
BMI, and PG/MAXLAAL is the ratio PG peak to the maximum peak of
either LA or AL.
[0248] According to the Examples 1-3 evaluating DDD-MRS diagnostic
processor aspects of the present disclosure across clinical
experience and data, features from in phase power and absorption
spectra may be used to develop diagnostic classifiers with a high
correlation to standard control measures for differentiating
painful from non-painful discs, including highly invasive, painful,
costly, and controversial needle-based provocative discography. The
Example 3 in particular, pursued according to the present DDD-MRS
embodiments of this disclosure, demonstrate that data from
absorption mode spectral acquisitions may be used to partition
spectra based on separating lipid from non-lipid signals and via a
relationship of PG/MAXLAAL prior to classification to achieve 100%
procedural success and 100% accurate diagnosis. While the intial
partition for lipid was done manually by visual signal quality
observation believed to indicate presence or absence of lipid
signal contribution, the recognition of this may be done
automatically using several techniques. For example, this may be
done by determining linewidth in the LAAL region (where lipid
co-exists, if present), LAAL peak amplitude exceeding a threshold,
LAAL peak/power (e.g. AUC), by the ability to detect a PG peak, or
by the combination of any of the aforementioned techniques, as may
be applied against thresholds determined empirically or otherwise
to represent a valid test for the signal differentiation.
[0249] It has also been shown herein that another statistically
robust hybrid linear regression equation may be used without the
PG/MAXLAAL partition, at the expense of only slightly increased
false negative scores (n=2).
Example 4
[0250] A DDD-MRS exam according to the DDD-MRS pulse sequence,
signal processing, and certain diagnostic algorithm aspects of the
present disclosure was conducted in a synthetic "phantom" spine
intended to simulate certain aspects of a lumbar spine with
controlled, known chemical environments with respect to aqueous
preparations of varying concentrations and relative ratios between
n-acetyl acetate (NAA) and lactic acid (LA) in simulated discs
providing regions of interest for voxel prescription and DDD-MRS
examinations for test validation purposes.
[0251] Materials and Methods
[0252] Sagittal plane MRI images from a GE Signa 3.0 T of the two
lumbar spine phantoms shown in FIGS. 40A-B included a longitudinal
series of simulated disc chambers along a column and floating in
mineral oil. The simulated disc chambers were filled with buffered
solutions of lithium lactate (LA) and n-acetyl aspartate (NAA) as
indicated in Table 6. Phantom "B" shown in FIG. 40A also had
alternating chambers that were also filled with mineral oil to
simulate vertebral bodies (VBs), whereas the Phantom "C" shown in
FIG. 40B had the discs in immediately adjacent succession without
intervening simulated VBs.
[0253] Voxels were prescribed within various discs among the
phantoms for varied range of target chemicals. DDD-MRS pulse
sequence acquisitions according to various of the present
embodiments were obtained from the Signa 3 T. Settings for these
exams included: TR/TE settings of 1000/28 ms, NSA=150, 3rd flip
angle=85, voxel dimension=5.times.20.times.20 mm, VSS bands were
default width, and sweep rate=2 Kh.
[0254] Metabolite signal (Smet) for NAA was measured by integrating
signal power over a range or "bin" centered on spectral peak with
width of +/-0.1 PPM. Lactate signal was measured by integrating
over bin ranging from 0.1 PPM on either side of observed doublet
peak. Unsuppressed water signal (Suw) measured over water peak
+/-0.5 PPM. Metabolite concentrations (CM) were then calculated
using the following formulaic relationship:
CM=(Smet/Suw).times.(Nw/Nmet).times.C water.times.K;
where Nw=2 H, Nmet=3 H (both NAA and LA), C water=55.5M, and
K=correction factor for each phantom based on relaxation, signal
measurement and acquisition factors. Factors underlying "K" were
not characterized, thus K was solved for each acquisition based on
known actual concentrations of each metabolite to derive an average
K value for each phantom which was then applied uniformly across
the phantom acquisitions to solve for each CM.
[0255] Results/Discussion
[0256] Results of measured/calculated concentration values per the
DDD-MRS exam were compared against known values for NAA and LA,
with comparison results shown in FIGS. 40C-D, respectively, and in
Table 6. DDD-MRS measured vs. known concentration comparisons
resulted in very high correlations of R2=0.95 for the LA comparison
and R2=0.93 for the NAA comparison (after removing one clearly
erroneous outlier--which resided at an exceptionally high test NAA
level which is well above the typical levels considered
physiologically relevant, at least for DDD pain diagnostic purposes
in the lumbar spinal discs). Ratios of NAA:LA were also
substantially accurate and significantly correlative, as shown in
Table 6.
[0257] According to this study featured in Example 4, operation of
the DDD-MRS system operation through respective modes of pulse
sequence spectral acquisition, signal processing, and data
extraction was verified to provide robust results with respect to
NAA and LA chemical concentrations, and ratios therebetween, in
this controlled simulated test environment. This provides some
degree of verification with respect to the accuracy and robust
operation of the DDD-MRS system in other applications for
performing similar operations in vivo.
[0258] Further Discussion and Additional Aspects of the
Disclosure
[0259] It is to be appreciated that the present disclosure,
including by reference to the Examples, provides various aspects
that can be highly beneficial, and represent new advancements that
enhance the ability to perform clinically relevant MRS-based
examinations of the lumbar spine, and/or of intervertebral discs,
and in particular indications for diagnosing DDD pain. Each of
these aspects, taken alone, is considered of independent value not
requiring combination with other aspects herein disclosed. However,
the combination of these aspects, and various sub-combinations
apparent to one of ordinary skill, represent still further aspects
of additional benefit and utility. The following are a few examples
of these aspects, in addition to others noted elsewhere herein or
otherwise apparent to one of ordinary skill, which aspects
nonetheless not intended to be limiting to other aspects disclosed
herein and are intended to be read in conjunction with the
remaining disclosures provided elsewhere herein:
[0260] Channel Selection for Data Processing and Diagnosis:
[0261] Conventional MRI systems use multi-channel acquisition coils
for spine detectors, which are pads that patients lye upon during a
scan. The GE Signa for example uses an 8 channel acquisition coil
array, of which 6 channels are typically activated for use for
lumbar spine imaging and diagnosis (including for MRS). However,
the system generally combines all data from these channels in
producing a single "averaged" curve. For single voxel MRS, this has
been determined to be highly inefficient and significant source of
error in the data, in particular reducing signal-to-noise ratio.
The channels vary in their geographical placement relative to
lumbar discs, and are believed to be at least one source of
variability between them regarding acquired signal quality for a
given disc. Of the six channels, most frequently at least one of
the channels is clearly "poor" data (e.g. poor signal-to-noise),
and often this can mean 2 to 5 of those channels being clearly
degraded vs. one or more "strong" channels. Accordingly, the
present disclosure contemplates that comparing the channels, and
using only the "strongest" channel(s), significantly improves
signal quality and thus data acquired and processed in performing a
diagnosis. This "channel isolation/selection" is considered
uniquely beneficial to the DDD pain application contemplated
herein, and can be done manually as contemplated herein, though the
present disclosure also includes automating this operation to
compare and choose amongst the channels for a given voxel scan via
an automated DDD-MRS signal processor disclosed.
[0262] "Coherent" Averaging Within and Between Channels:
[0263] During a single voxel scan, many repetitions are performed
that are later used for averaging in order to reduce noise and
increase signal-to-noise ratio in an acquired MRS spectrum. This
can range from about 100 repetitions to about 600 or more, though
more typically may be between about 200 to about 500, and still
more frequently between about 300 to about 400, and according to
one specific (though example) embodiment frequently included in the
physical embodiments evaluated in the clinical study of Example 1
may be about 384 repetitions. With a TR of 1 to 2 seconds for
example, this can range from less than 5 to 10 minutes time.
[0264] However, a "shift" in phase and frequency has been observed
among the acquired data over these repetitions. The current
standard MRI system configurations, via certain sequence routines,
do not completely correct for such shifts. Thus when these
repetitions are averaged the result becomes "blurred" with reduced
signal amplitude relative to noise, as well as possibility for
signal "broadening" or separation into multiple peaks from what
should be otherwise a single, more narrow band peak.
[0265] In addition or alternative to "strongest" channel selection
for processing, significant benefit and utility is contemplated
herein for correcting for one or both of these phase and/or
frequency "shifts" among the repetitions of an acquisition series
acquired at a channel during a single voxel scan. The observed
results of such processing have been higher signal quality, with
higher signal-to-noise ratio, and/or more narrow defined signals at
bands of interest to spectral regions associated with chemicals
believed (and correlated) to be relevant for diagnosing disc pain
(e.g., PG and/or LA and/or AL). It is noted, and relevant to
various of the detailed embodiments disclosed herein, that the
spectral peak region associated with water is typically the most
prominent and highest amplitude signal across the spectrum. This
peak and its location relative to a baseline is used according to
certain of the present embodiments to define a given shift in a
signal, and thus that shift at the water region is used to correct
the entire spectral signal back to a defined baseline. As water
peak shifts, or conversely is corrected, so does the rest of the
spectrum including the target chemical markers relevant to
conducting diagnoses.
[0266] This degree and location of the water peak may also be used
to determine and edit acquisition frames which are sufficiently
abnormally biased relative to the other acquisition frames to
adversely impact spectral data (or unable to "grab and shift"),
e.g. frame editing according to further embodiments.
[0267] Where water is not as prominent, e.g. highly desiccated
discs with over suppressed water in the sequence, other reliably
prominent and recognizable peaks maybe identified used for similar
purpose (e.g. peaks within the PG and/or LA and/or AL regions
themselves). However, due to its typical prominence and many
benefits of using the water peak for these various signal
processing purposes, novel approaches and settings for water
suppression are contemplated and disclosed herein. This provides
for a water signal, either manually or automatically, within an
amplitude range that is sufficient to locate and "grab" for
processing, but not so extensive to "washout" lower chemical
signatures in an inappropriate dynamic range built around the
higher water signal. The result of corrections contemplated herein
aligns the repetitions to phase and/or frequency coherence, and
thus the resulting averaging achieved is desirably more "coherent"
averaging. It is further contemplated that these shifts may be
observed and corrected in either time or frequency domain
(especially regarding frequency shift), and while certain
embodiments are described herein in detail corrections yielding
similarly improved results may be made in either domain (again esp.
re: frequency coherent correction).
[0268] DDD-MRS Factors, Criteria and Thresholds for Diagnostic
Results
[0269] The present disclosure provides an empirically derived
relationship between four weighted factors that involve data
derived from three regions of MRS spectra acquired from discs that
are generally associated with three different chemicals, namely PG,
LA, and AL. Other support exists to suspect these identified
chemicals may be active culprits in disc pain, e.g. reducing PG,
and increasing LA and AL, as factored in the diagnostic
relationship developed and applied herein. More directly, at least
a sub-set of these factors used in this diagnostic developed
relationship have been directly correlated to disc pain (e.g. PG/LA
ratio per prior 11 T studies performed ex vivo). These factors are
further addressed in view of further supporting literature and
disclosures, which are believed to support their correlation to
pain, as follows.
[0270] The normal intervertbral disc is avascular and disc cells
function under anaerobic conditions. (Ishihara and Urban 1999;
Grunhagen, Wilde et al. 2006) Anaerobic metabolism, such as in the
setting of oxygen deprivation and hypoxia, causes lactate
production. (Bartels, Fairbank et al. 1998; Urban, Smith et al.
2004) Disc pH is proportional to lactate concentration. (Diamant,
Karlsson et al. 1968) Lactic acid produces pain via acid sensing
ion channels on nociceptors. (Immke and McCleskey 2001; Sutherland,
Benson et al. 2001; Molliver, Immke et al. 2005; Naves and
McCleskey 2005; Rukwied, Chizh et al. 2007) Disc acidity has been
correlated with pre-operative back pain. (Diamant, Karlsson et al.
1968; Nachemson 1969; Keshari, Lotz et al. 2008)
[0271] Proteoglycan content within the nucleus pulposus, which is
the primary matrix which holds water in the disc nucleus, decreases
with disc degeneration, which is also associate with dehydration
e.g. via "darkened" disc nuclei seen on T2 MRI. (Roughley, Alini et
al. 2002; Keshari, Lotz et al. 2005; Keshari, Zektzer et al. 2005;
Roberts, Evans et al. 2006) ChondVOItin sulfate proteoglycans
inhibit nerve ingrowth. (Zuo, Hernandez et al. 1998; Zuo, Neubauer
et al. 1998; Jones, Sajed et al. 2003; Properzi, Asher et al. 2003;
Jain, Brady-Kalnay et al. 2004; Klapka and Muller 2006) Nerve
ingrowth is increased in degenerative painful discs. (Brown,
Hukkanen et al. 1997; Coppes, Marani et al. 1997; Freemont, Peacock
et al. 1997; Freemont, Watkins et al. 2002)
[0272] Discography is the current gold-standard of diagnostic care
for differentiating painful discs, but is controversial due to
being: invasive, painful, subjective, technique/operator dependent,
frequently challenged due to high false positive rates (principally
as indicated in studies with asymptomatic volunteers), and risky to
the patient. (Carragee and Alamin 2001; Guyer and Ohnmeiss 2003;
O'Neill and Kurgansky 2004; Cohen, Larkin et al. 2005; Carragee,
Alamin et al. 2006; Carragee, Lincoln et al. 2006; Buenaventura,
Shah et al. 2007; Wichman 2007; Derby, Baker et al. 2008; Scuderi,
Brusovanik et al. 2008; Wolfer et al., Pain Physician 2008;
11:513-538.cndot.ISSN 1533-3159, Derby et al., 2008) The prevailing
modern guidelines for performing discography generally require
concordant pain intensity scores equal to or above 6 (on increasing
scale of 0-10), provocation pressures of no more than 50 psi above
opening pressure, and another negative control disc in order to
determine a "positive discogram" result for a disc. This modern
technique has been most recently suggested to provide a higher
specificity (e.g. lower false positive) rates than previously
alleged in other studies. (Wolfer et al., Pain Physician 2008;
11:513-538.cndot.ISSN 1533-3159) However, notwithstanding this
potential improvement with modern techniques in the test's
accuracy, a more recent published study has shown the invasive
needle puncture of discography significantly increases disc
degeneration and herniations rates. Further to this disclosure,
these adverse affects of the discography needle puncture in the
"negative control discs" have been alleged as possible culprit in
adjacent level disc disease that often affects adverse outcomes
following surgical treatments removing the "positive discogram"
discs (e.g. fusion and/or disc arthroplasty).
[0273] Proteoglycan and lactate within discs have unique MR
signatures that can be identified and objectively measured using MR
Spectroscopy, and a calculated ratio based on these measures has
significantly differentiated painful from non-painful discs in ex
vivo studies of surgically removed discs. (Keshari, Lotz et al.
2008) In subsequent clinical evaluation and development, the
further inclusion of alanine--related to lactate to extent of both
providing biomarkers for hypoxia having reasonable suspected basis
in pain cascade--has resulted in similarly accurate predictive
values for the platform in vivo. In one Example, with only 6%
procedural failures to make a confident diagnosis, 99% accuracy
resulted and including 5/5 successes in prospective application.
DDD-MRS approaches, as disclosed herein, can thus non-invasively,
painlessly, and objectively measure and quantify proteoglycan and
lactate-related signatures (and for alanine spectral region) of
intervertebral discs in vivo using a novel software upgrade to
commercially available MRI systems, and a novel diagnostic
algorithm based at least in part upon these in vivo measures
reliably distinguishes painful vs. non-painful discs with a lower
false positive rate predicted versus discography.
[0274] The following publications are herein incorporated in their
entirety by reference thereto, and provide at least in part a
bibliography of certain disclosures referenced above and otherwise
elsewhere herein: [0275] Bartels, E. M., J. C. Fairbank, et al.
(1998). "Oxygen and lactate concentrations measured in vivo in the
intervertebral discs of patients with scoliosis and back pain."
Spine 23(1): 1-7; discussion 8. [0276] Brown, M. F., M. V.
Hukkanen, et al. (1997). "Sensory and sympathetic innervation of
the vertebral endplate in patients with degenerative disc disease."
J Bone Joint Surg Br 79(1): 147-53. [0277] Buenaventura, R. M., R.
V. Shah, et al. (2007). "Systematic review of discography as a
diagnostic test for spinal pain: an update." Pain Physician 10(1):
147-64. [0278] Carragee, E. J. and T. F. Alamin (2001).
"Discography. a review." Spine J 1(5): 364-72. [0279] Carragee, E.
J., T. F. Alamin, et al. (2006). "Low-pressure positive Discography
in subjects asymptomatic of significant low back pain illness."
Spine 31(5): 505-9. [0280] Carragee, E. J., T. Lincoln, et al.
(2006). "A gold standard evaluation of the "discogenic pain"
diagnosis as determined by provocative discography." Spine 31(18):
2115-23. [0281] Cohen, S. P., T. M. Larkin, et al. (2005). "Lumbar
discography: a comprehensive review of outcome studies, diagnostic
accuracy, and principles." Reg Anesth Pain Med 30(2): 163-83.
[0282] Coppes, M. H., E. Marani, et al. (1997). "Innervation of
"painful" lumbar discs." Spine 22(20): 2342-9; discussion 2349-50.
[0283] Derby, R., R. M. Baker, et al. (2008). "Analgesic
Discography: Can Analgesic Testing Identify a Painful Disc?"
SpineLine (November-December): 17-24. [0284] Diamant, B., J.
Karlsson, et al. (1968). "Correlation between lactate levels and pH
in discs of patients with lumbar rhizopathies." Experientia 24(12):
1195-6. [0285] Freemont, A. J., T. E. Peacock, et al. (1997).
"Nerve ingrowth into diseased intervertebral disc in chronic back
pain." Lancet 350(9072): 178-81. [0286] Freemont, A. J., A.
Watkins, et al. (2002). "Nerve growth factor expression and
innervation of the painful intervertebral disc." J Pathol 197(3):
286-92. [0287] Grunhagen, T., G. Wilde, et al. (2006). "Nutrient
supply and intervertebral disc metabolism." J Bone Joint Surg Am 88
Suppl 2: 30-5. [0288] Guyer, R. D. and D. D. Ohnmeiss (2003).
"Lumbar discography." Spine J 3(3 Suppl): 11S-27S. [0289] Immke, D.
C. and E. W. McCleskey (2001). "Lactate enhances the acid-sensing
Na+ channel on ischemia-sensing neurons." Nat Neurosci 4(9):
869-70. [0290] Ishihara, H. and J. P. Urban (1999). "Effects of low
oxygen concentrations and metabolic inhibitors on proteoglycan and
protein synthesis rates in the intervertebral disc." J Orthop Res
17(6): 829-35. [0291] Jain, A., S. M. Brady-Kalnay, et al. (2004).
"Modulation of Rho GTPase activity alleviates chondroitin sulfate
proteoglycan-dependent inhibition of neurite extension." J Neurosci
Res 77(2): 299-307. [0292] Jones, L. L., D. Sajed, et al. (2003).
"Axonal regeneration through regions of chondroitin sulfate
proteoglycan deposition after spinal cord injury: a balance of
permissiveness and inhibition." J Neurosci 23(28): 9276-88. [0293]
Keshari, K. R., J. C. Lotz, et al. (2005). "Correlation of HR-MAS
spectroscopy derived metabolite concentrations with collagen and
proteoglycan levels and Thompson grade in the degenerative disc."
Spine 30(23): 2683-8. [0294] Keshari, K. R., J. C. Lotz, et al.
(2008). "Lactic acid and proteoglycans as metabolic markers for
discogenic back pain." Spine 33(3): 312-317. [0295] Keshari, K. R.,
A. S. Zektzer, et al. (2005). "Characterization of intervertebral
disc degeneration by high-resolution magic angle spinning (HR-MAS)
spectroscopy." Magn Reson Med 53(3): 519-27. [0296] Klapka, N. and
H. W. Muller (2006). "Collagen matrix in spinal cord injury." J
Neurotrauma 23(3-4): 422-35. [0297] Molliver, D. C., D. C. Immke,
et al. (2005). "ASIC3, an acid-sensing ion channel, is expressed in
metaboreceptive sensory neurons." Mol Pain 1: 35. [0298] Nachemson,
A. (1969). "Intradiscal measurements of pH in patients with lumbar
rhizopathies." Acta Orthop Scand 40(1): 23-42.
[0299] Naves, L. A. and E. W. McCleskey (2005). "An acid-sensing
ion channel that detects ischemic pain." Braz J Med Biol Res
38(11): 1561-9. [0300] O'Neill, C. and M. Kurgansky (2004).
"Subgroups of positive discs on discography." Spine 29(19): 2134-9.
[0301] Properzi, F., R. A. Asher, et al. (2003). "Chondroitin
sulphate proteoglycans in the central nervous system: changes and
synthesis after injury." Biochem Soc Trans 31(2): 335-6. [0302]
Roberts, S., H. Evans, et al. (2006). "Histology and pathology of
the human intervertebral disc." J Bone Joint Surg Am 88 Suppl 2:
10-4. [0303] Roughley, P. J., M. Alini, et al. (2002). "The role of
proteoglycans in aging, degeneration and repair of the
intervertebral disc." Biochem Soc Trans 30(Pt 6): 869-74. [0304]
Rukwied, R., B. A. Chizh, et al. (2007). "Potentiation of
nociceptive responses to low pH injections in humans by
prostaglandin E2." J Pain 8(5): 443-51. [0305] Scuderi, G. J., G.
V. Brusovanik, et al. (2008). "A critical evaluation of discography
in patients with lumbar intervertebral disc disease." Spine J 8(4):
624-9. [0306] Sutherland, S. P., C. J. Benson, et al. (2001).
"Acid-sensing ion channel 3 matches the acid-gated current in
cardiac ischemia-sensing neurons." Proc Natl Acad Sci USA 98(2):
711-6. [0307] Urban, J. P., S. Smith, et al. (2004). "Nutrition of
the intervertebral disc." Spine 29(23): 2700-9. [0308] Wichman, H.
J. (2007). "Discography: over 50 years of controversy." Wmj 106(1):
27-9. [0309] Wolfer, L. R., R. Derby, et al. (2008). "Systematic
review of lumbar provocation discography in asymptomatic subjects
with a meta-analysis of false-positive rates." Pain Physician
11(4): 513-38. [0310] Zuo, J., Y. J. Hernandez, et al. (1998).
"Chondroitin sulfate proteoglycan with neurite-inhibiting activity
is up-regulated following peripheral nerve injury." J Neurobiol
34(1): 41-54. [0311] Zuo, J., D. Neubauer, et al. (1998).
"Degradation of chondroitin sulfate proteoglycan enhances the
neurite-promoting potential of spinal cord tissue." Exp Neurol
154(2): 654-62. Notwithstanding the foregoing, it is to be
appreciated that despite the support for suspecting these chemicals
as the cause of pain, and despite the belief that these chemicals
are measured and represented at least in part by the data derived
from the MRS data acquired, this correlation need not be accurate
in order for the data and diagnostic algorithm and approach
presented herein to remain valid and highly useful.
[0312] In particular regard to MRS data derived from regions
associated with LA and AL, these are quite narrowly defined ranges
closely adjacent to each other, and also overlap with a much
broader band associated with lipid. Accordingly, the data acquired
from these two "bins" may blur between the actual two chemical
sources. However, as they both relate to and are a product of
abnormal cellular metabolism and hypoxia, their combination may be
fairly considered a signature region more broadly for "abnormal
cellular metabolism/hypoxia." Furthermore, lipid contribution may
bias measurements in this region, and as lipid is a high molecular
weight molecule if present the signal is typically strong and often
may wash out resolution of either or both of LA or AL-based signal
in the region. However, in the current experience with DDD-MRS,
even where lipid signal is believed present, and even in
significant degree, the acquired data intended to represent LA and
AL as processed through the diagnostic algorithm and processor has
not produced a false result against controls (e.g. remains an
accurate result). When this happens, the diagnostic result is
consistently MRS+ indicating a positive result for pain in the
suspect disc. However, such lipid-related positive results occur
most frequently in L5-S1 discs that are associated with a
particular degenerative profile and morphology that is more
reliably diagnosed as painful on MRI alone (and consistently
confirmed as such via PD).
[0313] To the extent the measurements derived from the MRS
"regions" believed to be associated with these chemicals, and as
used in the weighted factor diagnostic algorithm developed, are
applied uniformly across the different control disc populations,
the diagnostic accuracy of the result prevails in the ultimate
comparison data--regardless of the source of the MRS data acquired.
Accordingly, the benefit and utility of the diagnostic approach is
defined ultimately by its diagnostic results, and not intended to
be necessarily limited and defined only by the theory as to what
the underlying sources of the measured signatures are.
[0314] Conversely, it is also further contemplated and to be
understood that the present disclosure provides a specific
diagnostic relationship algorithm that produces a particular range
of diagnostic results that compare with high correlation with
control measures for pain/non-pain in discs evaluated. However,
this is the result of statistically generated correlation and
retrospective approach to data fitting. While appropriate for
diagnostic algorithm development and the specific result disclosed
herein is considered highly beneficial, this may migrate to other
specific algorithms that may be more preferred though without
departing from the broad scope intended for the various aspects of
this disclosure. Such modifications may be the result of further
data processing across more samples, for example, and may affect
the "weighting" multipliers associated with each factor used in the
algorithm, or which factors are featured in the algorithm, or which
regions or features of the MRS spectra are even used as the
signatures from which data is derived and used in the algorithm.
This has been demonstrated by way of the Examples 1-3 provided
herein, and wherein three different specific diagnostically
relevant and viable approaches are presented and described for
similar data sets (e.g. in particular comparison between Examples 2
and 3 of the same clinical data set).
[0315] It is contemplated that while the DDD-MRS diagnostic
processor herein disclosed and diagnostic results provided
therefrom, as disclosed in context of clinical data presented under
Example 1 (and late by Examples 2 and 3), provide binary MRS+ and
MRS- results for severe pain and absence of severe pain in discs,
respectively. However, the results are also quantified along a
scaled range which may be appropriately interpreted by a
diagnostician as "levels" of relevance along the pain/non-pain
range. Such interpretation may impact the direction of pain
management decisions, such as which discs to treat, how to treat,
or not to treat at all. One example of such other way of presenting
DDD-MRS diagnostic information for utility to appropriate
clinicians is demonstrated by reference to the "% prediction
painful" presentation of data shown and discussed herein (which may
be instead or in combination also determined and presented as "%
prediction non-painful"). Moreover, while the current diagnostic
embodiments have been described by reference to site-specific
locations of pain sources at reference discs, diagnostic value may
be more generalized to confirmed presence or absence of any painful
disc at all. Such may impact more general management decision, such
as administration or avoidance of pain medication. Still further,
the current aspects may be used to assess aspects of the chemical
environments of discs, either in addition to or alternative to
specific diagnostic indications such as for pain or non-pain
determinations for given discs. This may be effectively utilitarian
for example by providing measures of chemical biomarkers, such as
PG, LA, AL, LAAL, etc., such as amounts or concentrations thereof
in the tissues (and/or ratios). This may be relevant for example in
other indications or applications, such as research purposes (e.g.
biologics or cell therapy approaches to treating or providing
prophylaxis to discs). This may be useful either prior to
treatment, and/or following treatment to assess certain aspects of
outcomes and progression of the treatment or underlying disease or
condition intended to be treated (as may relate to chemicals being
monitored).
[0316] Furthermore, in still further embodiments, the diagnostic
results may be provided in different forms than as described by the
specific embodiments disclosed by reference to a particular
example, such as Example 1 for example. For example, binary
definitive diagnoses of MRS+ and MRS- may be supplemented with
"indeterminate" as a third category. This may, for example,
represent a result of applying certain threshold criteria that must
be met in order to make a definitive +/- determination. Such
criteria may include, for example, SNR threshold of the underlying
post-processed DDD-MRS spectrum from which the diagnostic data is
extracted for performing the diagnoses. In another example, a
defined proximity of calculated diagnostic results from the DDD-MRS
diagnostic processor to the zero (0) median threshold between MRS+
and MRS- diagnoses may represent a threshold under which definitive
MRS+/- determination is not decidedly made by the processor.
[0317] It is also to be further appreciated that the pulse sequence
platform approach, and/or specific parameter settings, and/or
signal processing approaches (and/or parameter or threshold
criteria settings), may be modified. Such modifications may affect
resulting spectra (and data extracted therefrom) sufficiently to
redistribute the regional data used for diagnostic purposes, and
may thus motivate or necessitate a re-evaluation and re-formation
of the diagnostic algorithm that is appropriate for data acquired
and/or processed under those modified approaches. Accordingly,
while the present interactions between these component parts of an
overall DDD-MRS system, and results, are considered of particular
benefit for forward application in clinical use, such further
modifications are also considered to fall within the broad scope of
the aspects disclosed herein, and may represent for example a
consequence of further development and experience as would be
apparent to one of ordinary skill (though such further
modifications may also provide still further benefit).
[0318] L5-S1 and Novel Detection Coils:
[0319] The L5-S1 disc is typically oriented at an oblique angle
relative to other lumbar discs, and has unique shape that in many
circumstances challenges the ability to prescribe voxel for
adequate DDD-MRS data acquisition. The current voxelation plan for
MRS generally requires a three-dimensional "cube" of space to be
defined as the voxel (a pixel with volume), typically done by an
operator technician on overlay to MRI images of the region.
However, for this angled L5-S1 disc, the voxel volume may be
maximized by angling the voxel to match the angulated disc.
However, such angled voxels at this location have been observed to
relate to degraded data acquisition by existing spine detector
coils. Accordingly, a custom spine coil is further contemplated
that angles at least one coil channel to either a pre-determined
angle more representative of typical L5-S1 discs, or a range of
angles may be provided by multiple such coils in a kit, or the coil
channel may be given an "adjustable" angle to meet a given anatomy.
Furthermore, software may be adapted to identify an angled voxel
and modify the coordinate system assigned for sequence and/or
multi-channel acquisition in order better acquire data from an
angled voxel (e.g. where planar slices are taken through the voxel
as data acquired, the planar coordinates are revised into an
adjusted coordinate system that accounts for the angulation
relative to the data acquisition at the channel(s)). This uniquely
angled disc level is also associated with and located within a
radiused curvature at the small of the back, which may be more
extreme in some patients than others. While simply adjusting the
angle of lower detection channel coils may improve acquisition
here, further more dramatic variations are also contemplated. In
one such further aspect, a detector coil array is created with
smaller coils, and/or on a flexible platform that is adjusted to
more accurately fit against the lower back (vs. a planar array
currently used, but for curved lower spine with increasingly
angulated discs toward the lower lumbar and sacral regions).
Further to this approach, the relative locations and orientations
of the detector coils may be sensed, with proper coordinate system
assigned thereto for sequencing and acquisition during single voxel
MRS of the spine (especially intervertebral discs), and which also
may be adapted relative to coordinates of voxel orientation,
dimensions, and shape.
[0320] T1-Rho:
[0321] An additional MRI-based pulse sequence technology has been
previously disclosed called "T1-Rho". This is a sequence that has
been alleged for detecting, measuring, and indicating the amount
(e.g. concentration) of proteoglycan, via n-acetyl or n-acetyl
acetate, in tissue, and furthermore for using this information for
diagnostic benefit for some conditions. In one particular regard,
this has been alleged to be potentially useful for monitoring
degree of degeneration, in that reduced proteoglycan in discs may
correlate to advancing degree of degeneration. While pain
correlation with proteoglycan variability has not been determined,
the ration of PG to other metabolites, such as for example Lactate
(and/or alanine), is believed to be a consistent and potent
indicator for localized discogenic pain. Accordingly, the present
disclosure combines T1-Rho with other measurements, e.g. MRS
measurements, in evaluating tissue chemistry for purpose of
performing a diagnosis. In one particular mode contemplated herein,
the T1-Rho measurement of proteoglycan/n-acetyl content is used to
"normalize" or otherwise calibrate or compare an MRS measurement of
that related region. In doing so, other metabolites in the MRS
spectrum may be also calibrated for more accurately calculated
"concentration" measurement. This calibration may be done in
evaluating MRS signal quality, such as for example between channels
or within a channel itself, and MRS data is used for the diagnosis.
In a further mode, T1-Rho information related to PG may be used as
the data for that chemical constituent in tissue, and data for
another diagnostically relevant chemical, e.g. Lactate as measured
for example via MRS (or other modality), may be used in combination
with the PG measurement in an overall diagnostic algorithm or
evaluation. Such algorithms applied for diagnostic use may be
empirically driven based upon experimental data which may be
conducted and acquired by one of ordinary skill for such purpose
based upon this disclosure. For example, a database of sufficient
patient data based on T1-rho measurements (for proteoglycan) and
MRS measurements (such as for PG and/or Lactate, for example) may
be correlated in a multi-variate logistic regression analysis
against other pain indicators such as provocative discography or
treatment outcomes, resulting in a highly correlative algorithm
based upon the data fit. This may then be used prospectively in
predicting or assessing localized pain in newly evaluated patient
tissues. In one particular benefit, MRS techniques include
particular sequence parameters that emphasize lactate for improved
lactate-related data extraction, and decreasing lipid artifact
(which often overlays over lactate to confound lactate data
collection), but not considered as robust for other chemicals, such
as potentially PG/n-acetyl. One such technique extends the time
delay from magnetic activation to data collection, thus increasing
overall time for repetitive scans. However, T1-Rho is relatively
fast to perform relative to MRS. Accordingly, one particular
further embodiment uses T1-rho for PG measurement, and MRS as
enhanced for lactate measurement, and combines this data into an
empirically data-driven algorithm for performing a diagnosis.
Moreover, a further aspect contemplated herein uses T1-rho for PG
measurement, in combination with pH or p02 measurement (e.g. via a
sensor on a needle, such as a discography needle) to monitor local
acidity in the disc (also believed to relate to lactate
concentration).
[0322] Diagnostic Display "Enhancing" MRI Images
[0323] The various aspects, modes, and embodiments of the present
disclosure provide, among other beneficial advancements, a
significant enhancement and improvement to standard MRI for locally
diagnosing painful and/or non-painful discs. The utility of each of
these diagnoses--painful, and non-painful--is of independent value
on its own. While indicating a disc is definitively painful may
often augment other clinical or diagnostic indications for
directing treatment to the level, indicating a disc is definitively
not painful also provides valuable information to exclude a disc as
possible pain culprit and avoid unnecessary intervention to the
level (especially where other clinical or diagnostic indications
may indicate another level as painful, but not provide definitive
answer to the other level/s). This is for example often the case
with respect to L3-L4 and L4-L5 discs, where L5-S1 discs (most
prevalently painful among the levels) may often be already suspect
per MR1 and other indications, but the higher adjacent disc levels
are indeterminate.
[0324] The present aspects have been presented in terms of physical
embodiments evaluated in clinical study with highly accurate
results against controls. By providing a non-invasive alternative
to discography as presented by these present embodiments, even if
diagnostically equivalent, significant benefits are advanced by
avoiding morbidity, pain, and other inefficiencies and downsides
associated with that invasive test.
[0325] As an enhancement to MRI, further aspects of the present
disclosure provide useful diagnostic display to indicate the
results in overlay context onto the MRI image itself and providing
context to the structures revealed therein, such as for example as
shown in FIGS. 32A-B for two different patients receiving a DDD-MRS
diagnostic exam according to Example 1.
[0326] It is to be appreciated by one of ordinary skill that the
various aspects, modes, embodiments, features, and variations of
the present disclosure include, without limitation, the
following.
[0327] One aspect of the present disclosure is a MRS pulse sequence
configured to generate and acquire a diagnostically useful MRS
spectrum from a voxel located principally within an intervertebral
disc of a patient. According to one mode of this aspect, the pulse
sequence is configured to generate and acquire the MRS spectrum
from a single voxel principally located within the disc. According
to another mode of this aspect, the pulse sequence is configured to
generate and acquire the MRS spectrum from the voxel located
principally within a nucleus of the disc. According to another mode
of this aspect, the pulse sequence is configured to generate and
acquire the MRS spectrum with sufficient signal-to-noise ratio
(SNR) upon appropriate post-signal processing to perform at least
one of: detect and measure at least one chemical constituent within
the disc; and diagnose a medical condition based upon one or more
identifiable signal features along the spectrum. According to
another mode, the pulse sequence is configured to generate and
acquire the MRS spectrum from a single voxel principally located
within a nucleus of the disc. According to another mode, the pulse
sequence is configured to generate and acquire the MRS spectrum
from a voxel principally located within an intervertebral disc of
the lumbar spine. According to another mode, the pulse sequence is
configured to generate and acquire at least one MRS spectrum from
at least one voxel principally located within at least one of
L3-L4, L4-L5, and L5-S1 intervertebral discs. These discs are the
most predominant discs implicated by chronic, severe low back pain,
and are also characterized by typically larger disc spaces than
other higher disc levels and thus more conducive to single voxel
spectroscopy (though not necessarily so limited to only these discs
in all cases). According thus to another mode, however, the pulse
sequence is configured to generate and acquire multiple MRS spectra
from multiple voxels, respectively, principally located within each
of L3-L4, L4-L5, and L5-S1 intervertebral discs.
[0328] According to another mode, the pulse sequence is configured
to generate and acquire multiple MRS spectra from multiple voxels,
respectively, principally located within each of L3-L4, and L4-L5
intervertebral discs. These discs are typically less oblique than
L5-S1 disc, and thus represent different geometric, and perhaps in
certain circumstances different biomechanical and/or biochemical,
environments vs. typically more oblique L5-S1 disc, and thus may
represent unique optimal approaches for diagnostic application of
the present embodiments versus for the L5-S1 disc. According to one
embodiment of this mode, the discs are substantially non-oblique,
such as for example as may be relative to a relatively more oblique
L5-S1 adjacent thereto. According thus to yet another mode, the
pulse sequence is configured to generate and acquire the MRS
spectrum from the voxel located principally within the L5-S1
intervertebral disc. As stated above, this disc level may at times
present unique considerations relative to other lumbar discs that
are addressed with unique relative approaches versus other lumbar
discs. According to one embodiment of this mode, the disc is
substantially oblique, such as for example relative to adjacent
lumbar disc segments above this level. According to another mode,
the pulse sequence is configured to operate in a first mode for a
substantially non-oblique disc, and a second mode for a
substantially oblique disc.
[0329] The present disclosure is considered readily adaptable to
operate on and with multiple different specific MR systems,
including of different relative field strengths and as may be made
available and operate in relative custom formats from various
different manufacturers, though as may be custom developed by one
of ordinary skill for compatibility and optimal functionality for
intended use on and with any particular MR system or category (e.g.
field strength). According to another mode therefore, the pulse
sequence of the various aspects of the present disclosure is
configured to generate and acquire the MRS spectrum via an NMR
system of at least about 1.2 tesla (T) field strength. According to
another mode, the pulse sequence is configured to generate and
acquire the MRS spectrum via an NMR system of about 1.2 tesla (T)
field strength. According to another mode, the pulse sequence is
configured to generate and acquire the MRS spectrum via an NMR
system of at least about 1.5 tesla (T) field strength. According to
another mode, the pulse sequence is configured to generate and
acquire the MRS spectrum via an NMR system of about 1.5 tesla (T)
field strength. According to another mode, the pulse sequence is
configured to generate and acquire the MRS spectrum via an NMR
system of at least about 3.0 tesla (T) field strength. According to
another mode, the pulse sequence is configured to generate and
acquire the MRS spectrum via an NMR system of about 3.0 tesla (T)
field strength. According to another mode, the pulse sequence is
configured to generate and acquire the MRS spectrum via an NMR
system of about 7.0 tesla (T) field strength. According to another
mode, it is to be appreciated that the pulse sequence is configured
to generate and acquire the MRS spectrum via an NMR system in the
range of about 1.2 to about 7.0 tesla (T) field strength. According
to another mode, the pulse sequence is configured to generate and
acquire the MRS spectrum via an NMR system in the range of about
1.2 to about 3.0 tesla (T) field strength. According to another
mode, the pulse sequence is configured to generate and acquire the
MRS spectrum via an NMR system in the range of about 1.5 to about
3.0 tesla (T) field strength. While these ranges and specific field
strengths noted represent existing systems available on the market
today, or at least under investigation (e.g. 7.0 T), it is further
contemplated that other systems outside this range may also be
suitable. However, it is also to be appreciated that systems below
about 1.5 or 1.2 Tesla may be challenged with respect to
signal:noise ratio in many circumstances (though may nonetheless be
operable suitably as intended in others). Furthermore, current
experience has revealed that acquisitions following the DDD-MRS
aspects of the present disclosure may be similarly robust when
conducted with field strength as low as 1.5 T versus as acquired
via higher 3.0 T systems (such as used in the Examples). Moreover,
systems above about 3.0 T or 7.0 T may present significant safety
concerns for many applications (though again may nonetheless
suitable for others).
[0330] Certain pulse sequence modes of the present aspects of the
disclosure are also to be appreciated as providing particular
benefit for certain intended uses, including those featured
specifically herein such as via the Examples. According to one such
mode of the present aspects, the pulse sequence comprises a
chemical shift selective (CHESS) sequence. According to another
mode, the pulse sequence comprises a point resolved spectroscopy
(PRESS) sequence. According to another mode, the pulse sequence
comprises a combination CHESS-PRESS sequence. According to another
mode, the pulse sequence comprises a combination CHESS-VSS-PRESS
sequence. According to another mode, the pulse sequence comprises
at least one control variable (CV) parameter setting as disclosed
in Table 1. According to another mode, the pulse sequence comprises
all the control variable (CV) parameter settings disclosed in Table
1. According to another mode, the pulse sequence comprises an echo
time (TE) in the range of about 25 to about 40 milliseconds.
According to another mode, the pulse sequence comprises an echo
time of about 28 milliseconds. This specific setting, while not
intended to be necessarily limiting to broad intended scope of the
present aspects and modes, has been observed to provide
sufficiently robust results as intended for various uses, such as
according to the Examples. According to another mode, the pulse
sequence comprises a repetition time (TR) in the range of about 750
to about 2000 milliseconds (2 seconds). According to another mode,
the pulse sequence comprises a repetition time (TR) of about 1000
milliseconds. According to another mode, the pulse sequence
comprises a repetition time of about 750 milliseconds and is
configured to operate with an MR system with a field strength of
between about 1.2 T and about 1.5 T. This has been observed, for
example in one particular embodiment, to be particularly beneficial
for 1.5 T MR applications. According to another mode, the pulse
sequence comprises a repetition time of between about 1000 and
about 1500 milliseconds and is configured to operate with an MR
system with a field strength of between about 3 T and about 7 T.
According to another mode, the pulse sequence is configured to
adjust the repetition time (TR) with respect to the field strength
of the MR system, which may be automatic in one beneficial
variation, or manually set to adjust accordingly. It is to be
appreciated that these settings for TR present a certain trade off
with respect to time required to complete a pulse sequence
acquisition series, and thus sufficiently short times to provide
adequate signal quality may be optimized for time efficiency,
though longer times may be done if time is available or not of
essence. Time, however, may be a significant consideration in many
circumstances, such as for example for efficiency in conducing the
exam in MR imaging center setting, and also patient comfort, in
addition to longer times for exams increase opportunities for
patient motion artifact etc. that could compromise results (to
extent not countered by the various signal processing aspects of
the present disclosure).
[0331] According to another mode, the pulse sequence comprises an
acquisition matrix size setting of about 1 in each dimension, with
a number of spatial slices setting of 1.
[0332] Relative degree of water signal in DDD-MRS pulse sequence
acquisitions may be relevant to the ability to fully signal process
such signals as intended by various aspects of the present
disclosure, and thus certain aspects related to water suppression
and water signal control are disclosed herein and to be appreciated
with respect to the pulse sequence. According to another mode, the
pulse sequence is configured to generate and acquire a repetitive
frame MRS acquisition series from the voxel with signal-to-noise
ratio (SNR) in the water region along the spectrum of multiple said
frames that is sufficiently high to be identified, yet sufficiently
low to provide adequate dynamic range with sufficient
signal-to-noise ratio (SNR) along other chemical regions of
diagnostic interest along the spectral frames to allow the other
regions to be identified and evaluated, post-signal processing and
post-averaging of the frames, for diagnostic use. Suppressed water
signal, and control of it via the pulse sequence settings, varied
over time of development across the clinical data set featured
among the Examples 1-3 disclosed herein. However, as demonstrated
via the highly robust ultimate results these ranges of water
suppression control experienced were observed to provide
sufficiently adequate results in most cases. This experience ranged
between 45 and 125 degrees for 3.sup.rd CHESS flip angle, with an
average of about 120 degrees (plus/minus about 30 degrees standard
deviation). However, these settings for each acquisition are
discrete, and upon achieving sufficient results a chosen setting
was cast for a given acquisition. The majority of acquisitions are
believed sufficient at about 85 to about 100 degrees for this third
CHESS flip angle, though again may be custom set in iterative
experience or via automated feedback control based upon trial and
error in measured signal quality.
[0333] According nonetheless to another mode of the present
aspects, the pulse sequence comprises a third CHESS flip angle of
at least about 45 degrees. According to another mode, the pulse
sequence comprises a third CHESS flip angle of at least about 65
degrees. According to another mode, the pulse sequence comprises a
third CHESS flip angle of up to about 145 degrees. According to
another mode, the pulse sequence comprises a third CHESS flip angle
of up to about 125 degrees. According to another mode, the pulse
sequence comprises a third CHESS flip angle of between about 45 and
about 145 degrees. According to another mode, the pulse sequence
comprises a third CHESS flip angle between about 65 and about 125
degrees. According to another mode, the pulse sequence comprises a
third CHESS flip angle that is adjustable based upon a degree of
water observed in the region of interest. According to one
embodiment of this mode, the degree of water is observed according
to a prior test pulse sequence. According to one embodiment of this
mode, the pulse sequence is configured to operate in series
following the prior test pulse sequence in a common MR exam
session, and the third CHESS flip angle is automatically adjustable
based upon the observed degree of water in the prior test pulse
sequence. According to another embodiment, the third CHESS flip
angle is automatically adjustable based upon a T2-weighted
acquisition value for the region of interest. According to another
embodiment, the third CHESS flip angle is automatically adjustable
to a value determined based upon an empirical correlation between
third CHESS flip angle and T2-weighted acquisition value for the
region of interest according to a prior acquisition data set.
According to another mode, the pulse sequence comprises at least
one of the following CHESS flip angles: about 105 degrees (angle
1); about 80 degrees (angle 2); about 125 degrees (angle 3). In
some embodiments, the first CHESS flip angle can be between about
60 degrees and about 180 degrees, or between about 85 degrees and
about 125 degrees. In some embodiments, the second CHESS flip angle
can be between about 60 degrees and about 180 degrees, or between
about 65 degrees and about 105 degrees. In some embodiments, the
third CHESS flip angle can be between about 45 degrees and about
145 degrees, or between about 85 degrees and about 125 degrees, or
between about 105 degrees and about 145 degrees.
[0334] Certain aspects are also disclosed related to a PRESS mode
of operation. According to one such example mode, the pulse
sequence comprises PRESS correction settings of about 1.2 for each
of X, Y, and Z axes. Other PRESS correction settings can be used,
such as values greater than 1.0 and less than about 1.5. According
to another mode, the pulse sequence comprises at least one of the
following PRESS flip angles: about 90 (angle 1); about 180 (angle
2); about 180 (angle 3). According to another mode, either or both
of the second and third PRESS flip angles may be between about 150
and about 180 degrees, and in one particular embodiment may be for
example about 167 degrees. As flip angle generally correlates with
time required to conduct the exam, signal quality results may be
optimally determined empirically against different flip angles, and
it may also be the case that a setting (e.g. 180) may not be the
exact flip angle actually deployed (e.g. may actually be different,
e.g. about 167 for example).
[0335] According to another mode of the present MRS pulse sequence
aspects, the pulse sequence is provided in combination with an MRS
signal processor according to one or more of the various aspects,
modes, embodiments, variations, and or features thereof as
otherwise elsewhere herein provided.
[0336] Another aspect of the present disclosure is thus an MRS
signal processor configured to process spectral data from an MRS
pulse sequence.
[0337] According to one mode of this aspect, the MRS signal
processor comprises a channel selector that is configured to select
a sub-set of multiple channel acquisitions received
contemporaneously from multiple parallel acquisition channels,
respectively, of a multi-channel detector assembly during a
repetitive-frame MRS pulse sequence series conducted on a region of
interest within a body of a subject. According to one embodiment of
this mode, the channel selector of the MRS signal processor is
configured to select a sub-set of multiple channel acquisitions
received contemporaneously--from multiple parallel acquisition
channels, respectively, of a multi-channel detector assembly during
the repetitive-frame MRS pulse sequence series conducted on a voxel
principally located within an intervertebral disc within the body
of the subject. According to another embodiment, the channel
selector of the MRS signal processor is configured to automatically
differentiate relatively stronger from weaker channel acquisitions
received. According to another embodiment, the channel selector of
the MRS signal processor is configured to determine and select a
strongest single channel acquisition signal among the multiple
channel acquisitions. According to another embodiment, the channel
selector of the MRS signal processor is configured to determine and
select the strongest single channel acquisition based upon a
highest measured parameter of the single channel acquisition
spectral series comprising at least one of amplitude, power, or
signal-to-noise ratio (SNR) of water signal in the spectrum in the
selected channel relative to the other channel. According to one
highly beneficial variation of this embodiment, the channel
selector of the MRS signal processor is configured to determine and
select the strongest single acquisition channel with CHESS sequence
disabled. According to another beneficial variation of this
embodiment, the channel selector is configured to perform a channel
selection that is based upon a frame averaged spectrum of the
series acquired from the channel. According to one beneficial
alternative feature of this variation, the frame averaged spectrum
of the series is acquired with the CHESS disabled on unsuppressed
water frames. According to another variation of this embodiment,
the channel selector of the MRS signal processor is configured to
determine and select a sub-set of strongest channels based upon a
range threshold based from the highest measured parameter of the
strongest single channel. According to another embodiment, the
channel selector of the MRS signal processor is configured to
determine and select one or more "strongest" channels among the
series based upon a threshold criteria for a feature of the channel
acquisition data. According to one beneficial variation of this
embodiment, the one or more strongest channels is determined and
selected by averaging the first unsuppressed water frames for each
channel (with CHESS disabled) and measuring the signal to noise
ratio (SNR) of the unsuppressed water signal, determine which
channel has the strongest SNR and then selecting those additional
channels that fall within a threshold range, e.g. about 3 dB (or
may be for example a range of 1 to 6 dB) of the channel with the
strongest SNR. According to still further channel selector
embodiments, the channel selector is provided in combination with
one or more of the various other aspects, modes, embodiments,
variations, and features related to other MRS pulse sequence and/or
MRS signal processor disclosures provided herein.
[0338] Another mode of the MRS signal processor aspects of the
present disclosure comprises a phase shift corrector configured to
recognize and correct phase shifting within a repetitive
multi-frame acquisition series acquired by a multi-channel detector
assembly during an MRS pulse sequence series conducted on a region
of interest within a body of a subject. According to one embodiment
of this mode, the phase shift corrector is configured to recognize
and correct the phase shifting within a repetitive multi-frame
acquisition series acquired by a multi-channel detector assembly
during an MRS pulse sequence series conducted on a voxel within an
intervertebral disc in the body of the patient. According to
another embodiment, the phase shift corrector is configured to
recognize and correct the phase shifting in the time domain.
According to another embodiment, the phase shift corrector is
provided in combination with one or more of the various other
aspects, modes, embodiments, variations, and features related to
other MRS pulse sequence and/or MRS signal processor disclosures
provided herein.
[0339] Another mode of the MRS signal processor aspects of the
present disclosure comprises a frequency shift corrector configured
to recognize and correct relative frequency shifts between multiple
acquisition frames of a repetitive multi-frame acquisition series
acquired within an acquisition detector channel of a multi-channel
detector assembly during a MRS pulse sequence series conducted on a
region of interest within a body of a subject. According to one
embodiment of this mode, the frequency shift corrector is
configured to recognize and correct frequency shift error between
multiple acquisition frames of a repetitive multi-frame acquisition
series acquired within an acquisition detector channel of a
multi-channel detector assembly during a MRS pulse sequence series
conducted on a voxel within an intervertebral disc in the body of
the subject. According to another embodiment, the frequency shift
corrector is configured to recognize and correct the frequency
shift error in the time domain. According to one beneficial example
of this embodiment, the frequency shift is recognized and corrected
in the time domain by the application of the inverse of a 1.sup.st
order linear curve fit of the incremental phase estimate of time
domain information in the 16 frame average of unsuppressed water
frames (such as for example about 16 unsuppressed water frames of
the detailed illustrative embodiments and Examples disclosed
herein). According to another embodiment, the frequency shift
corrector is configured to recognize and correct the frequency
shift error in the frequency domain. According to one beneficial
example of this embodiment, the frequency shift is recognized and
corrected in the frequency domain by transforming the time domain
information in the unsuppressed water frames (e.g. n=16) into the
frequency domain to locate the water signal peak, determine the
frequency error of the water peak, and then shift the transformed
suppressed water frames by the negative of the frequency error.
According to another example, the frequency shift corrector is
configured to identify and locate a water signal in each of
multiple acquisition frames of the series, compare the location of
the located water signals against a reference baseline location to
determine a separation shift therebetween for each frame, and to
correct the shift to align the location to the baseline location by
applying an appropriate offset to all the spectral data of each
frame. According to one variation of this example, the location of
the water signal is estimated based upon a location range where the
water signal exceeds a threshold amplitude value. According to
another variation, the water signal identified and located
comprises a peak value of the water signal. According to one highly
beneficial feature that may be further embodied in this variation,
the threshold amplitude value is greater than about 0.6 and/or less
than about 0.9, and the threshold amplitude value can be 0.8 in
some cases. According to another embodiment of this mode, the
frequency shift corrector is provided in combination with one or
more of the various other aspects, modes, embodiments, variations,
and features related to other MRS pulse sequence and/or MRS signal
processor disclosures provided herein.
[0340] Another yet another mode of the MRS signal processor aspects
disclosed herein comprises a frame editor. According to one
embodiment of this mode, the frame editor is configured to
recognize at least one poor quality acquisition frame, as
determined against at least one threshold criterion, within an
acquisition channel of a repetitive multi-frame acquisition series
received from a multi-channel detector assembly during a MRS pulse
sequence series conducted on a region of interest within a body of
a subject. According to one example of this embodiment, the frame
editor is configured to edit out the poor quality frame from the
remainder of the MRS pulse sequence series otherwise retained for
further signal and/or diagnostic algorithm processing. According to
another embodiment, the frame editor is configured to recognize the
poor quality acquisition frame based upon a threshold value applied
to error in location of recognized water signal from an assigned
baseline location. According to another embodiment, the frame
editor is configured to recognize the poor quality acquisition
frame based upon a threshold confidence interval applied to the
ability to recognize the signal location of water signal in the
frame spectrum. According to one example of this embodiment, the
water signal location comprises a location of a peak of the water
signal. According to another example, a confidence level for the
location of the water signal peak of a frame is estimated and
compared to a confidence level threshold to qualify a frame for
subsequent frequency correction. According to another more detailed
example, a confidence level may be determined by the following
steps: (1) analyze the discrete amplitude spectrum in the range of
the center-tuned frequency plus and minus 40 Hz (in the case of a 3
T system, half that for a 1.5 T system); (2) locate the highest
peak and determine its width at the half-amplitude point; (3)
determine the total spectral width of all parts of the spectrum
which exceed the half-amplitude point of the highest peak; (4) form
the confidence estimate by taking the ratio of the spectral width
of the greatest peak divided by the total spectral width which
exceeds the threshold. By way of further illustration of this
example, if there is only a single peak above the threshold, the
confidence estimate will be 1.0, if there are many other peaks or
spectral components which could be confused with the greatest one,
then the estimate will reduce and ultimately approach zero (0). It
is believed that this provides a simple and robust estimate of the
randomness or dispersal of energy in the vicinity of the water
peak. Like an entropy measure, described elsewhere herein, it has
the desirable characteristic that its performance is generally
believed to be invariant with amplitude. According to still another
embodiment of the present mode, the frame editor is provided in
combination with one or more of the various other aspects, modes,
embodiments, variations, and features related to other MRS pulse
sequence and/or MRS signal processor disclosures provided
herein.
[0341] Another mode of the MRS signal processor aspects of the
present disclosure comprises an apodizer to reduce the truncation
effect on the sampled data. The apodizer according to certain
embodiments is configured to apodize an MRS acquisition frame in
the time domain otherwise generated and acquired via an MRS pulse
sequence aspect otherwise herein disclosed, and/or as also
otherwise signal processed by one or more of the various MRS signal
processor aspects also otherwise herein disclosed. The apodizer
according to various embodiments of this mode is provided in
combination with one or more of the various other aspects, modes,
embodiments, variations, and features related to other MRS pulse
sequence and/or MRS signal processor disclosures provided
herein.
[0342] It is to be further appreciated that the various MRS signal
processor, aspects, modes, features, variations, and examples
herein described may be configured according to further modes to
operate and/or provide diagnostic information related to a tissue
in a patient based upon an MRS spectrum in real-part squared
representation of the acquired spectral data or processed spectrum.
According to still further modes, such may be operated upon or
presented as complex absorption spectrum of the acquired or
processed data. Yet another mode contemplated operates and/or
presents processed results as complex absorption spectrum and also
as real part squared representation of the acquired and/or signal
processed data.
[0343] Another aspect of the present disclosure is an MRS
diagnostic processor configured to process information extracted
from an MRS spectrum for a region of interest in a body of a
subject, and to provide the processed information in a manner that
is useful for diagnosing a medical condition or chemical
environment associated with the region of interest.
[0344] According to one mode of this aspect, the MRS diagnostic
processor is configured to process the extracted information from
the MRS spectrum for a voxel principally located in an
intervertebral disc of the subject, and to provide the processed
information in a manner that is useful for diagnosing a medical
condition or chemical environment associated with the
intervertebral disc. According to one embodiment of this mode, the
MRS diagnostic processor is configured to process the extracted
information from the MRS spectrum for a voxel principally located
in a nucleus of the intervertebral disc, and to provide the
processed information in a manner that is useful for diagnosing a
medical condition or chemical environment associated with the
intervertebral disc. According to another embodiment, the MRS
diagnostic processor is configured to provide the processed
information in a manner that is useful for diagnosing the
intervertebral disc as painful. According to another embodiment,
the MRS diagnostic processor is configured to provide the processed
information in a manner that is useful for diagnosing the
intervertebral disc as severely painful. According to another
embodiment, the MRS diagnostic processor is configured to provide
the processed information in a manner that is useful for diagnosing
the intervertebral disc as not severely painful. According to
another embodiment, the MRS diagnostic processor is configured to
provide the processed information in a manner that is useful for
diagnosing the intervertebral disc as substantially non-painful.
According to another embodiment, the MRS diagnostic processor is
configured to diagnose the disc as painful. According to another
embodiment, the MRS diagnostic processor is configured to diagnose
the disc as severely painful. According to another embodiment, the
MRS diagnostic processor is configured to diagnose the disc as not
severely painful. According to another embodiment, the MRS
diagnostic processor is configured to diagnose the disc as
substantially non-painful. According to another embodiment, the MRS
diagnostic processor is configured to diagnose the disc with
respect to % probability the disc is painful. According to another
embodiment, the MRS diagnostic processor is configured diagnose the
disc with respect to % probability the disc is not painful.
According to one variation of the preceding embodiments, the MRS
diagnostic processor is configured to diagnose the disc with
respect to % probability the disc is painful or not painful based
upon a calculated value for the disc using acquired MRS spectral
information for the disc against an empirical prior test data set
of similarly calculated values for other sample discs correlated
with % predictive values against known or assumed classifications
for such other sample discs as painful vs. non-painful. According
to another embodiment, the MRS diagnostic processor is configured
to assign a value for the disc that is referenced against a range
for use in determining presence, absence, or level of pain.
According to another embodiment, the MRS diagnostic processor is
configured to provide the diagnostically useful information in a
display provided contextually with an MRI image of the respective
lumbar spine comprising the disc. According to another embodiment,
the MRS diagnostic processor is configured to provide the
diagnostically useful information in a display overlay onto an MRI
image of the respective lumbar spine comprising the disc. According
to one variation of this embodiment, the display overlay associates
the diagnostically useful information with one or more
intervertebral discs evaluated. According to another variation, the
display overlay comprises a scaled legend of values along a range,
and an indicator of a result referenced against the range in the
legend and associated with an intervertebral disc evaluated.
According to another variation, the display overlay comprises both
color coding and numerical coding of results in a legend and for at
least one indicator of processed information associated with at
least one intervertebral disc evaluated by the diagnostic
processor. According to another embodiment, the MRS diagnostic
processor comprises a diagnostic algorithm empirically created by
comparing acquired and processed MRS spectra for multiple
intervertebral discs against control measures for pain, and that is
configured to determine whether discs evaluated with the MRS
spectra are painful or non-painful. According to one variation, the
diagnostic algorithm comprises at least one factor related to
spectral information extracted from MRS spectral regions associated
with at least one of proteoglycan, lactate, and alanine chemicals.
According to one applicable feature of this variation, the spectral
information is extracted from an MRS spectral region associated
with n-acetyl resonance associated with proteoglycan. According to
one feature of this variation, the extracted information related to
at least one said region is adjusted according to an adjustment
factor related to voxel volume. According to one example, the
extracted information related to at least one said region is
divided by voxel volume. According to another feature of this
variation, the extracted information related to at least one said
region is adjusted according to an adjustment factor related to
body mass index (BMI). According to one example, the extracted
information related to at least one said region is multiplied by
body mass index (BMI) of the patient. According to another example,
the extracted information is multiplied by BMI of the patient
divided by a reference BMI. According to a further example, the
reference BMI is average BMI calculated across an empirical test
data set from which the diagnostic algorithm is statistically
developed for correlation to the classifications. According to
another feature of this variation, the extracted information
related to at least one said region comprises a peak value in the
region. According to another feature of this variation, the
extracted information related to at least one said region comprises
a power value in the region. According to another applicable
feature, the diagnostic algorithm comprises at least two factors
related to spectral information extracted from the MRS spectral
regions associated with at least two of said chemicals. According
to another applicable feature, the diagnostic algorithm comprises
three factors related to spectral information extracted from the
MRS spectral regions associated with all three of said chemicals.
According to one particularly beneficial example of this feature,
each of the three factors is related to one of the proteoglycan,
lactate, and alanine chemicals, respectively. According to another
applicable feature, the diagnostic algorithm comprises at least two
factors related to spectral information extracted from MRS spectral
regions associated with at least three said chemicals. According to
one particularly beneficial example of this feature, a first factor
is related to spectral information extracted from the MRS spectral
region associated with proteoglycan (e.g. n-acetyl peak region),
and a second factor is related to spectral information extracted
from MRS spectral regions associated with lactate and alanine in
combination. According to another particularly beneficial feature,
the diagnostic algorithm comprises a factor related to spectral
information extracted from MRS spectral regions associated with
each of lactate and alanine chemicals in combination. According to
one highly beneficial example of this feature, the factor comprises
maximum peak value across the combination of the lactate and
alanine spectral regions. According to another highly beneficial
example, the factor comprises a power value across the combination
of the lactate and alanine spectral regions. According to another
applicable feature, the diagnostic algorithm comprises at least two
said factors related to spectral information extracted from the MRS
spectral regions associated with all three of said chemicals.
According to still another applicable feature, at least one said
factor is weighted by a constant. According to another applicable
feature, at least one said factor comprises a ratio of at least two
values associated with information extracted from the MRS spectra
at regions associated with at least two of proteoglycan, lactate,
and alanine chemicals. According to still a further variation, the
algorithm comprises four factors associated with MRS spectral data
associated with proteoglycan region, lactate region,
proteoglycan:lactate region ratio, and proteoglycan:alanine region
ratio. According to one applicable feature of this variation, the
algorithm comprises four factors associated with MRS spectral data
associated with proteoglycan region divided by voxel volume,
lactate region divided by voxel volume, proteoglycan:lactate region
ratio, and proteoglycan:alanine region ratio. According to still
another applicable feature, the four factors are weighted by
constants. According to still a further variation, the algorithm is
configured to calculate a diagnostically useful value based upon
PG/LA, PG/AL, PG/vol, and LA/vol factors, wherein PG=peak
measurement in proteoglycan spectral region, AL=peak measurement in
alanine region, LA=peak measurement in LA region, and vol=volume of
prescribed voxel in the disc used for MRS data acquisition.
According to still a further variation, the algorithm is configured
to calculate a diagnostically useful value as follows:
Value=-[log(PG/LA*(0.6390061)+PG/AL*(1.45108778)+PG/vol*(1.34213514)+LA/-
VOL*(-0.5945179)-2.8750366)];
wherein PG=peak measurement in proteoglycan spectral region,
AL=peak measurement in alanine region, LA=peak measurement in LA
region, and vol=volume of prescribed voxel in disc used for MRS
data acquisition. Further to this algorithm, however, it is to be
appreciated that, though considered highly beneficial, the specific
constants may be slightly varied, and aspects such as the negative
and log multipliers of the overall remaining functions may not be
absolutely necessary and the removal of these aspects may still
provide sufficiently robust results (e.g. the negative multiplier
inverts negative values, otherwise corresponding with painful
results to positive numbers as more colloquially corresponding with
"positive" test results indicating pain condition is present, and
visa versa for negative test results; and the log function provides
collapse of data distribution spread not necessary for all
applications and not necessarily altering ultimate results).
According to still a further applicable feature, the calculated
diagnostically useful value is compared against a threshold value
of zero (0) to determine pain diagnosis. According to still a
further applicable feature, positive calculated values are
considered painful and negative calculated values are considered
non-painful diagnoses. According to another variation, the
diagnostic algorithm is based at least in part upon a feature
associated with a combined spectral region associated with lactate
and alanine chemicals. According to another variation, the
diagnostic algorithm is based at least in part upon a power
measurement taken along an MRS spectral region that combines
regions associated with lactate and alanine chemicals.
[0345] According to another mode of the MRS diagnostic processor
aspects of the disclosure, the diagnostic processor is provided in
combination with one or more of the various other aspects, modes,
embodiments, variations, and features related to other MRS pulse
sequence and/or MRS signal processor disclosures also provided
herein.
[0346] According to another mode of the present aspect, the MRS
diagnostic processor may be configured to implement the following
equation:
Score = - 4.6010405 + 1.58785166 ( BLA ) - 0.081991 ( VBLAAL -
29.3125 ) * ( VBLAAL - 29.3125 ) + 0.01483355 ( PG / MAXLAAL -
7.14499 ) * ( PG / MAXLAAL - 7.14499 ) * ( PG / MAXLAAL - 7.14499 )
+ 0.1442603 ( MAXLAAL / vol - 16.1202 ) * ( VBLAAL - 29.3125 ) -
0.0008879 ( VBLAAL - 29.3125 ) 2 ( MAXLAAL / VOL - 16.1202 )
##EQU00005##
where BLA is the BMI corrected LA spectral peak, VBLAAL is the ROI
volume and BMI normalized sum of the LA and AL spectral peaks,
MAXLAAL is the maximum of either the LA or AL peaks, and PG is the
n-acetyl spectral peak.
[0347] According to another mode of the present aspect, the MRS
diagnostic processor may be configured to implement one or more of
the following equations:
[0348] High Lipid Classifier
Score = - ( - 335.51971 + 0.00010632 * ( LAVVBMI ) 2 + 873.744714 *
( PG / ( LAALVVBMI ) ) ) ; ##EQU00006##
where LAVVBMI equals the voxel volume and BMI adjusted LA peak
value.
[0349] PG/MAXLAAL>1.85, Non-Lipid, Classifier
Score=-(-1.4959544+1.72223147*(PG(MAXLAAL)));
where PG/MAXLAAL equals the PG peak value divided by the maximum
peak value of the LAAL region.
[0350] PG/MAXLAAL<1.85, Non-Lipid, Classifier
Score = - 1 * ( - 134.40909800961 + 3.96992556918043 * LAVVBMI -
2.6198628365642 * ALVVBMI + 113.683315467568 * ALAUCVVBMI -
149.65896624348 * SQRT ( PGAUCVVBMI ) ) ; ##EQU00007##
where LAVVBMI is the voxel volume and BMI adjusted LA peak value,
ALVVBMI is the voxel volume and BMI adjusted AL peak value,
ALAUCVVBMI is the AL region area under the curve as voxel volume
and BMI adjusted, and PGAUCVVBMI is the PG region area under the
curve as voxel volume and BMI adjusted.
[0351] It is to be appreciated that these formulaic relationships
shown above, and elsewhere herein, are examples of highly accurate
results that have been enjoyed with the present embodiments when
put into practice. However, the examples are also provided in fine
detail for full disclosure and understanding. These finer details
are not intended to be necessarily limiting in all cases. For
example, many of the constants disclosed herein are shown to many
decimal points, which is the format generated by the engineering
platforms employed to generate them. It would be readily apparent
to one of ordinary skill that these likely could be significantly
truncated or rounded without significant degradation or departing
from the scope of the present disclosure. In addition, in order to
provide abundance of understanding and disclosure, certain theories
and explanations may be put forth and postulated herein, which may
not be fully accurate, and are not necessary in order to fully
embrace and enjoy the present embodiments and invention. The
novelty and beneficial utility of the present embodiments may be
fully appreciated and enjoyed without being bound by theory, as
should be appreciated by one of ordinary skill.
[0352] Another aspect of the present disclosure comprises a
diagnostic system configured to generate information useful for
diagnosing a medical condition or chemical environment in a tissue
of a subject based at least in part upon a combination of
lactate-related and alanine-related factors measured or estimated
in the tissue. According to one mode of this aspect, the diagnostic
system is configured to generate the useful information based at
least in part upon one combination lactate-alanine (LAAL)-related
diagnostic factor related to a combination of lactate-related and
alanine-related factors measured or estimated in the tissue.
According to one embodiment of this mode, the combination LAAL
factor provides useful information as a LAAL biomarker for hypoxia
in the tissue. According to another embodiment of this mode, the
diagnostic system is further configured to provide the useful
information based on the LAAL factor in combination with a second
factor related to a third chemical-related factor measured or
estimated in the tissue. According to one variation of this
embodiment, the third chemical-related factor comprises a biomarker
associated with enervation of the tissue. According to another
variation of this embodiment, the third chemical-related factor is
associated with proteoglycan content in the tissue. According to
another mode of this aspect, the diagnostic system comprises an MRS
diagnostic processor, and the lactate-related and alanine-related
factors comprise features associated with lactate-related and
alanine-related regions of an MRS spectrum of a region of interest
in the tissue. According to one embodiment of this mode, the MRS
diagnostic processor is further configured to generate the useful
information based at least in part upon one combination
lactate-alanine (LAAL)-related diagnostic factor related to a
combination of the lactate-related and alanine-related factors
measured or estimated in the tissue. According to one variation of
this embodiment, the combination LAAL factor comprises a maximum
peak spectral value in the combined LAAL region of the MRS
spectrum. According to another variation of this embodiment, the
combination LAAL factor comprises a measured or estimated overall
power value in the combined LAAL region of the MRS spectrum.
[0353] Another aspect of the present disclosure is an MRS system
comprising an MRS pulse sequence, MRS signal processor, and MRS
diagnostic processor, and which is configured to generate, acquire,
and process an MRS spectrum for providing diagnostically useful
information associated with a region of interest in a body of a
patient. According to one mode of this aspect, the MRS system
comprising the MRS pulse sequence, MRS signal processor, and MRS
diagnostic processor, is configured to generate, acquire, and
process the MRS spectrum for a voxel principally located in an
intervertebral disc in the body of the patient and to provide
diagnostically useful information associated with the disc.
According to one embodiment of this mode, the voxel is principally
located in a nucleus of the disc. According to another embodiment
of this mode, the diagnostically useful information is useful for
diagnosing pain or absence of pain associated with the disc.
Various further modes of this aspect are contemplated that comprise
one or more of the various aspects, modes, embodiments, variations,
and features of the MRS pulse sequence, MRS signal processor, and
MRS diagnostic processor as elsewhere described herein. According
to one such further mode, for example, the MRS pulse sequence
comprises a combination CHESS-PRESS sequence. According to another
example of such a mode, the MRS pulse sequence comprises a
combination CHESS-VSS-PRESS sequence. According to another such
further mode, the MRS pulse sequence comprises a TE of about 28 ms
and a TR of about 1000 ms, whereas TE according to further
embodiments can range from between about 25 to about 40 ms and TR
can typically range from between about 750 to about 2000 ms.
According to another such further mode, the MRS signal processor
comprises at least one of a channel selector, a phase shift
corrector, an apodizer, a frame editor, a frequency shift
corrector, and a frame averaging combiner. According to another
mode, the MRS diagnostic processor is configured to calculate and
provide diagnostically useful information for diagnosing pain
associated with at least one intervertebral disc based upon at
least one MRS spectral region associated with at least one of
proteoglycan, lactate, and alanine chemicals. According to one
embodiment of this mode, information associated with each of the
MRS spectral regions associated with each of these chemicals is
used by the MRS diagnostic processor in providing the
diagnostically useful information. According to another embodiment,
a combination LAAL factor associated with a combination of the
lactate-related and alanine-related MRS spectral regions is used.
According to one variation of this embodiment, the combination LAAL
factor is used in further combination with a second factor
associated with a proteoglycan-related (such as for example
n-acetyl) MRS spectral region for an overall diagnostic
algorithm.
[0354] According to another mode of the various aspects above, each
or all of the respective MRS system components described is
provided as user or controller operable software in a
non-transitory computer readable storage medium configured to be
installed and operated by one or more processors. According to one
embodiment of this mode, a non-transitory computer operable storage
medium is provided and stores the operable software.
[0355] The following issued US patents are also herein incorporated
in their entirety by reference thereto: U.S. Pat. Nos. 5,617,861;
5,903,149; 6,617,169; 6,835,572; 6,836,114; 6,943,033; 7,042,214;
7,319,784.
[0356] The following pending US Patent Application Publication is
herein incorporated in its entirety by reference thereto:
US2007/0253910.
[0357] The following PCT Patent Application Publication is also
herein incorporated in its entirety by reference thereto:
WO2009/058915.
[0358] Some aspects of the systems and methods described herein can
advantageously be implemented using, for example, computer
software, hardware, firmware, or any combination of computer
software, hardware, and firmware. Computer software can comprise
computer executable code stored in a computer readable medium that,
when executed, performs the functions described herein. In some
embodiments, computer-executable code is executed by one or more
general purpose computer processors. A skilled artisan will
appreciate, in light of this disclosure, that any feature or
function that can be implemented using software to be executed on a
general purpose computer can also be implemented using a different
combination of hardware, software, or firmware. For example, such a
module can be implemented completely in hardware using a
combination of integrated circuits. Alternatively or additionally,
such a feature or function can be implemented completely or
partially using specialized computers designed to perform the
particular functions described herein rather than by general
purpose computers.
[0359] A skilled artisan will also appreciate, in light of this
disclosure, that multiple distributed computing devices can be
substituted for any one computing device illustrated herein. In
such distributed embodiments, the functions of the one computing
device are distributed (e.g., over a network) such that some
functions are performed on each of the distributed computing
devices.
[0360] Some embodiments of the present invention may be described
with reference to equations, algorithms, and/or flowchart
illustrations of methods according to embodiments of the invention.
These methods may be implemented using computer program
instructions executable on one or more computers. These methods may
also be implemented as computer program products either separately,
or as a component of an apparatus or system. In this regard, each
equation, algorithm, or block or step of a flowchart, and
combinations thereof, may be implemented by hardware, firmware,
and/or software including one or more computer program instructions
embodied in computer-readable program code logic. As will be
appreciated, any such computer program instructions may be loaded
onto one or more computers, including without limitation a general
purpose computer or special purpose computer, or other programmable
processing apparatus to produce a machine, such that the computer
program instructions which execute on the computer(s) or other
programmable processing device(s) implement the functions specified
in the equations, algorithms, and/or flowcharts. It will also be
understood that each equation, algorithm, and/or block in flowchart
illustrations, and combinations thereof, may be implemented by
special purpose hardware-based computer systems which perform the
specified functions or steps, or combinations of special purpose
hardware and computer-readable program code logic means.
[0361] Furthermore, computer program instructions, such as embodied
in computer-readable program code logic, may also be stored in a
computer readable memory (e.g., a non-transitory computer readable
medium) that can direct one or more computers or other programmable
processing devices to function in a particular manner, such that
the instructions stored in the computer-readable memory implement
the function(s) specified in the block(s) of the flowchart(s). The
computer program instructions may also be loaded onto one or more
computers or other programmable computing devices to cause a series
of operational steps to be performed on the one or more computers
or other programmable computing devices to produce a
computer-implemented process such that the instructions which
execute on the computer or other programmable processing apparatus
provide steps for implementing the functions specified in the
equation(s), algorithm(s), and/or block(s) of the flowchart(s).
[0362] While various alternative modalities may be employed as
stated, one particular example of an overall diagnostic system 200
and various related functional interfacing components are shown in
FIGS. 41A-B and referenced with respect to schematic flow of an
exam and related steps post-DDD-MRS pulse sequence acquisition as
follows. FIG. 41A shows a DDD-MRS pulse sequence acquisition and
output communication flow diagram. An MRI exam is first conducted
on the patient 202 who typically is slid supine into MR system 210
while lying on a spine detector coil 220 that acquires the MR and
MRS signals. This is followed by the DDD-MRS pulse sequence, as
also conducted via the same MR system 210 and by a trained
operator/technician 204. Data representative of the anatomy of the
patient 202 is generated 258 (e.g., data representative of the
chemical makeup of an area of interest inside the intervertebral
disc of the patient's spine 252). The results are then packaged in
a data archive folder 250 that includes information related to the
MRI image 258 (if taken and retained for this purpose), voxel
prescription in various relevant planes 252, pre-packaged output
spectra 254 (if desired for any further use, or not), complex data
files for the acquired series 256. This is sent via PACS 260 for
storage and/or further communication either as push or pull for
further processing. In some embodiments, the MR system 210 may be
operated by a computer system or terminal 230 that can be located
remotely or can be integrated into the MR system 210, to allow one
or more operators 240 (or the technician 204) to provide
instructions or other information to the MR system 210.
[0363] As shown in FIG. 41B, this data package 250 may then be
accessed or pushed from the PACS 260 to another local DDD-MRS
engine 261, which may be a local computer 262 (and related
peripheral devices such as display 264 and keyboard 266), work
station, or other modality, or terminal (e.g., terminal 230), which
may conduct the DDD-MRS signal processing and/or diagnostic
processing and for packaged display of results as appropriate. This
may be monitored via other remote device 290, such as via the
internet 270 as shown schematically in FIG. 41B--and this may
include for example license monitoring such as on a "per click" or
"volume"-related use license fee basis or other such use monitoring
purposes (e.g. data collection and analysis purposes, e.g. for
trials, studies, registries, etc.). The more remote processors may
be a central server 292 providing certain SAAS support to the
system, or again for more monitoring. These files, at any stage,
can be configured to push or be pulled electronically, such as to a
remote DDD-MRS station 280 with engine components including a
computer 282, monitor 284, keyboard 286, where diagnostic results
such as overlaid images 288 may be seen and analyzed for example
and the various processors may be stored and employed for
functional use in a variety of single or multiple coordinated
locations and controllers or computers, with ultimate flexibility
re: specific modality for operation and storage 294 and/or
communication of results.
[0364] While certain embodiments of the disclosure have been
described, these embodiments have been presented by way of example
only, and are not intended to limit the scope of the broader
aspects of the disclosure. Indeed, the novel methods, systems, and
devices described herein may be embodied in a variety of other
forms. For example, embodiments of one illustrated or described
DDD-MRS system component may be combined with embodiments of
another illustrated or described DDD-MRS system component.
Moreover, the DDD-MRS system components described above, e.g. pulse
sequence, signal processor, or diagnostic processor, may be
utilized for other purposes. For example, an MRS system (or
component sequence, signal processor, or diagnostic processor
useful therewith or therein), may be configured and used in manners
consistent with one or more broad aspects of this disclosure for
diagnosing other tissue environments or conditions than pain within
an intervertebral disc. Or, such may be usefully employed for
diagnosing pain or other tissue environments or conditions in other
regions of interest within the body. Such further applications are
considered within the broad scope of disclosure contemplated
herein, with or without further modifications, omissions, or
additions that may be made by one of ordinary skill for a
particular purpose. Furthermore, various omissions, substitutions
and changes in the form of the methods, systems, and devices
described herein may be made without departing from the spirit of
the disclosure. Components and elements may be altered, added,
removed, or rearranged. Additionally, processing steps may be
altered, added, removed, or reordered. While certain embodiments
have been explicitly described, other embodiments will also be
apparent to those of ordinary skill in the art based on this
disclosure.
[0365] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
magnetic resonance spectroscopy (MRS) processing system configured
to process a repetitive frame MRS spectral acquisition series
generated and acquired for a voxel principally located within an
intervertebral disc via an MRS pulse sequence, and acquired at
multiple parallel acquisition channels of a multi-coil spine
detector assembly, in order to provide diagnostic information
associated with the disc, comprising: an MRS signal processor
comprising a channel selector, a phase shift corrector, a frequency
shift corrector, a frame editor, and a channel combiner, and
configured to receive and process the MRS spectral acquisition
series for the disc and to generate a processed MRS spectrum for
the series with sufficient signal-to-noise ratio (SNR) to acquire
information associated with identifiable features along MRS
spectral regions associated with unique chemical constituents in
the disc; and an MRS diagnostic processor configured to extract
data from identifiable chemical regions in the processed MRS
spectrum in a manner that provides diagnostic information for
diagnosing a medical condition or chemical environment associated
with the disc.
[0366] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a magnetic resonance spectroscopy (MRS) processing system
configured to process a repetitive frame MRS spectral acquisition
series generated and acquired for a voxel principally located
within an intervertebral disc via an MRS pulse sequence, and
acquired at multiple parallel acquisition channels of a multi-coil
spine detector assembly, in order to provide diagnostic information
associated with the disc, comprising: an MRS signal processor
comprising a channel selector, a phase shift corrector, a frequency
shift corrector, a frame editor, and a channel combiner, and
configured to receive and process the MRS spectral acquisition
series for the disc and to generate a processed MRS spectrum for
the series with sufficient signal-to-noise ratio (SNR) to acquire
information associated with identifiable features along MRS
spectral regions associated with unique chemical constituents in
the disc; and an MRS diagnostic processor configured to extract
data from identifiable chemical regions in the processed MRS
spectrum in a manner that provides diagnostic information for
diagnosing a medical condition or chemical environment associated
with the disc.
[0367] In one embodiment, a magnetic resonance spectroscopy (MRS)
processing method is used for processing a repetitive frame MRS
spectral acquisition series generated and acquired for a voxel
principally located within an intervertebral disc via an MRS pulse
sequence, and acquired at multiple parallel acquisition channels of
a multi-coil spine detector assembly, and for providing diagnostic
information associated with the disc, the method comprising:
receiving the MRS spectral acquisition series from the multiple
acquisition channels; signal processing the MRS acquisition series,
comprising selecting one or more channels among the parallel
channels based upon a predetermined criteria, recognizing and
correcting phase shift error among multiple frames within the
series of a channel acquisition, recognizing and correcting a
frequency shift error between multiple frames within the series of
the channel acquisition, recognizing and editing out frames from
the series based upon a predetermined criteria, combining selected
and corrected channels for a combined average processed MRS
spectrum; and diagnostically processing the processed MRS spectrum
by extracting data from identifiable chemical regions in the
processed MRS spectrum and processing the extracted data in a
manner that provides MRS-based diagnostic information for
diagnosing a medical condition or chemical environment associated
with the disc.
[0368] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a signal processor
configured to signal process a repetitive multi-frame MRS pulse
sequence acquisition series of MRS spectra frames received from
multiple acquisition channels of a detector assembly during a MRS
pulse sequence series conducted on a region of interest (ROI)
within a tissue in a body of a subject; and wherein the signal
processor comprises a channel selector configured to measure a
parameter related to MRS spectral signal quality for the acquired
MRS spectral series from each acquisition channel, compare the
measured parameters for the respective channels against at least
one threshold criteria for channel selection, identify a number of
selected channels which meet or exceed the threshold criteria and a
number of other failed channels which fail to meet the threshold
criteria, and retain the selected channels and discard the failed
channels from the acquisition series.
[0369] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a signal
processor configured to signal process a repetitive multi-frame MRS
pulse sequence acquisition series of MRS spectra frames received
from multiple acquisition channels of a detector assembly during a
MRS pulse sequence series conducted on a region of interest (ROI)
within a tissue in a body of a subject; and wherein the signal
processor comprises a channel selector configured to measure a
parameter related to MRS spectral signal quality for the acquired
MRS spectral series from each acquisition channel, compare the
measured parameters for the respective channels against at least
one threshold criteria for channel selection, identify a number of
selected channels which meet or exceed the threshold criteria and a
number of other failed channels which fail to meet the threshold
criteria, and retain the selected channels and discard the failed
channels from the acquisition series.
[0370] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a signal processor
configured to process a repetitive multi-frame MRS pulse sequence
acquisition series of MRS spectra frames received from an
acquisition channel of a detector assembly during a MRS pulse
sequence series conducted on a region of interest (ROI) within a
tissue in a body of a subject; wherein the signal processor
comprises a frame editor configured to measure a parameter related
to signal quality for the MRS spectrum for each acquired frame of
the acquisition series, compare the measured values for the
parameter for the respective frames against a threshold criteria,
and designate a number of successful frames that meet the threshold
criteria and a number of failed frames that fail to meet the
threshold criteria; and wherein the frame editor is further
configured to retain successful frames in the acquisition series,
and edit out the failed frames from the acquisition series if
number of successful frames meets or exceeds a minimum frame number
threshold, but to retain at least some of the failed frames in the
acquisition series if the number of successful frames is below the
minimum frame number threshold.
[0371] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a signal
processor configured to process a repetitive multi-frame MRS pulse
sequence acquisition series of MRS spectra frames received from an
acquisition channel of a detector assembly during a MRS pulse
sequence series conducted on a region of interest (ROI) within a
tissue in a body of a subject; wherein the signal processor
comprises a frame editor configured to measure a parameter related
to signal quality for the MRS spectrum for each acquired frame of
the acquisition series, compare the measured values for the
parameter for the respective frames against a threshold criteria,
and designate a number of successful frames that meet the threshold
criteria and a number of failed frames that fail to meet the
threshold criteria; and wherein the frame editor is further
configured to retain successful frames in the acquisition series,
and edit out the failed frames from the acquisition series if
number of successful frames meets or exceeds a minimum frame number
threshold, but to retain at least some of the failed frames in the
acquisition series if the number of successful frames is below the
minimum frame number threshold.
[0372] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a signal processor
configured to process a repetitive multi-frame MRS pulse sequence
acquisition series of MRS spectra frames received from an
acquisition channel of a detector assembly during a MRS pulse
sequence series conducted on a region of interest (ROI) within a
tissue in a body of a subject; wherein the signal processor
comprises a frequency error corrector configured to calculate a
confidence level in an ability to estimate frequency shift error
for the MRS spectra of each frame of the series, compare each
calculated confidence level for each frame against at least one
threshold criteria, and determine a number of successful frames
that meet or exceed the threshold criteria and a number of other
failed frames that fail to meet the threshold criteria; and wherein
the signal processor is further configured to automatically
determine whether to (a) edit out the failed frames from the
acquisition series and perform frequency shift error correction via
the frequency error corrector in a manner to at least in part
reverse the frequency shift error estimate on each of the
successful frames, if the number of successful frames meets or
exceeds a minimum threshold number, or (b) retain at least some of
the failed frames and not perform frequency error correction to the
series via the frequency error corrector if the number of
successful frames is below the minimum threshold.
[0373] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a signal
processor configured to process a repetitive multi-frame MRS pulse
sequence acquisition series of MRS spectra frames received from an
acquisition channel of a detector assembly during a MRS pulse
sequence series conducted on a region of interest (ROI) within a
tissue in a body of a subject; wherein the signal processor
comprises a frequency error corrector configured to calculate a
confidence level in an ability to estimate frequency shift error
for the MRS spectra of each frame of the series, compare each
calculated confidence level for each frame against at least one
threshold criteria, and determine a number of successful frames
that meet or exceed the threshold criteria and a number of other
failed frames that fail to meet the threshold criteria; and wherein
the signal processor is further configured to automatically
determine whether to (a) edit out the failed frames from the
acquisition series and perform frequency shift error correction via
the frequency error corrector in a manner to at least in part
reverse the frequency shift error estimate on each of the
successful frames, if the number of successful frames meets or
exceeds a minimum threshold number, or (b) retain at least some of
the failed frames and not perform frequency error correction to the
series via the frequency error corrector if the number of
successful frames is below the minimum threshold.
[0374] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a signal quality evaluator
configured to automatically determine whether or not an MRS
spectrum acquired from a region of interest (ROI) in a tissue in a
body of a subject via an MRS pulse sequence series exam of the ROI
comprises a regional signature signal along the MRS spectrum that
is characteristic of lipid.
[0375] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a signal quality
evaluator configured to automatically determine whether or not an
MRS spectrum acquired from a region of interest (ROI) in a tissue
in a body of a subject via an MRS pulse sequence series exam of the
ROI comprises a regional signature signal along the MRS spectrum
that is characteristic of lipid.
[0376] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a diagnostic processor
configured to provide diagnostic information for diagnosing a
medical condition or chemical environment associated with a region
of interest (ROI) in a tissue in a body of a subject based at least
in part upon at least one chemical factor related to information
extracted from the ROI and associated with lactate (LA) and alanine
(AL) chemicals.
[0377] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a diagnostic
processor configured to provide diagnostic information for
diagnosing a medical condition or chemical environment associated
with a region of interest (ROI) in a tissue in a body of a subject
based at least in part upon at least one chemical factor related to
information extracted from the ROI and associated with lactate (LA)
and alanine (AL) chemicals.
[0378] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a diagnostic processor
configured to provide diagnostic information for diagnosing a
medical condition or chemical environment associated with a region
of interest (ROI) in a tissue in a body of a subject based at least
in part upon a chemical factor related to information extracted
from the ROI and associated with a chemical and as adjusted by an
adjustment factor that comprises at least one of a voxel-related
adjustment factor associated with a voxel prescribed to correspond
with the ROI and related to the information extracted, and a
subject-dependent variable-related adjustment factor associated
with the subject.
[0379] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a diagnostic
processor configured to provide diagnostic information for
diagnosing a medical condition or chemical environment associated
with a region of interest (ROI) in a tissue in a body of a subject
based at least in part upon a chemical factor related to
information extracted from the ROI and associated with a chemical
and as adjusted by an adjustment factor that comprises at least one
of a voxel-related adjustment factor associated with a voxel
prescribed to correspond with the ROI and related to the
information extracted, and a subject-dependent variable-related
adjustment factor associated with the subject.
[0380] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a diagnostic processor
configured to provide diagnostic processed information for
diagnosing a medical condition or chemical environment associated
with a region of interest (ROI) in a tissue in a body of a subject
based at least in part upon taking a first MRS measurement for a
chemical factor taken at a region of an MRS spectrum acquired from
the ROI and associated with a chemical and comparing the first MRS
measurement against a value derived from a different second
measurement and that is associated with an amount of the chemical
in the ROI.
[0381] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a diagnostic
processor configured to provide diagnostic processed information
for diagnosing a medical condition or chemical environment
associated with a region of interest (ROI) in a tissue in a body of
a subject based at least in part upon taking a first MRS
measurement for a chemical factor taken at a region of an MRS
spectrum acquired from the ROI and associated with a chemical and
comparing the first MRS measurement against a value derived from a
different second measurement and that is associated with an amount
of the chemical in the ROI.
[0382] In one embodiment, a computing system, comprising one or
more microprocessors receiving at least one signal responsive to
data collected in an MR scanner, is configured to implement a
medical diagnostic system, comprising: a diagnostic processor
configured to provide diagnostic information for diagnosing a
medical condition or chemical environment associated with a region
of interest (ROI) of a tissue in a body of a subject based at least
in part upon a chemical factor related to information extracted
from the ROI and associated with a chemical; and wherein said
diagnostic information comprises a probability value assigned to a
likelihood that the medical condition or chemical environment meets
certain criteria in the ROI.
[0383] In one embodiment, a physical computer readable medium
stores computer executable code that causes a computing system to
implement a medical diagnostic system, comprising: a diagnostic
processor configured to provide diagnostic information for
diagnosing a medical condition or chemical environment associated
with a region of interest (ROI) of a tissue in a body of a subject
based at least in part upon a chemical factor related to
information extracted from the ROI and associated with a chemical;
and wherein said diagnostic information comprises a probability
value assigned to a likelihood that the medical condition or
chemical environment meets certain criteria in the ROI.
[0384] In one embodiment, a medical diagnostic method comprises:
using a computing system to implement a signal processor for signal
processing a repetitive multi-frame MRS pulse sequence acquisition
series of MRS spectra frames received from multiple acquisition
channels of a detector assembly during a MRS pulse sequence series
conducted on a region of interest (ROI) within a tissue in a body
of a subject; and wherein the signal processing further comprises
using one or more microprocessors to operate a channel selector for
measuring a parameter related to MRS spectral signal quality for
the acquired MRS spectral series from each acquisition channel,
comparing the measured parameters for the respective channels
against at least one threshold criteria for channel selection,
identifying a number of selected channels which meet or exceed the
threshold criteria and a number of other failed channels which fail
to meet the threshold criteria, and retaining the selected channels
and discarding the failed channels from the acquisition series.
[0385] In one embodiment, a medical diagnostic method comprises:
using a computing system for signal processing a repetitive
multi-frame MRS pulse sequence acquisition series of MRS spectra
frames received from an acquisition channel of a detector assembly
during a MRS pulse sequence series conducted on a region of
interest (ROI) within a tissue in a body of a subject; wherein the
signal processing further comprises using a computing system for
implementing a frame editor for measuring, using one or more
microprocessors, a parameter related to signal quality for the MRS
spectrum for each acquired frame of the acquisition series,
comparing, using the one or more microprocessors, the measured
values for the parameter for the respective frames against a
threshold criteria, and designating, using the one or more
microprocessors, a number of successful frames that meet the
threshold criteria and a number of failed frames that fail to meet
the threshold criteria; and wherein the frame editing further
comprises retaining, using the one or more microprocessors,
successful frames in the acquisition series, and editing out the
failed frames from the acquisition series if the number of
successful frames meets or exceeds a minimum frame number
threshold, but retaining at least some of the failed frames in the
acquisition series if the number of successful frames is below the
minimum frame number threshold.
[0386] In one embodiment, a medical diagnostic method, comprising:
using a computing system to execute executable code for
implementing a signal processor for processing a repetitive
multi-frame MRS pulse sequence acquisition series of MRS spectra
frames received from an acquisition channel of a detector assembly
during a MRS pulse sequence series conducted on a region of
interest (ROI) within a tissue in a body of a subject; wherein the
signal processing further comprises using a computing system to
execute executable code for operating a frequency error corrector
for calculating, using one or more microprocessors, a confidence
level in an ability to estimate frequency shift error for the MRS
spectra of each frame of the series, comparing, using the one or
more microprocessors, each calculated confidence level for each
frame against at least one threshold criteria, and determining,
using the one or more microprocessors, a number of successful
frames that meet or exceed the threshold criteria and a number of
other failed frames that fail to meet the threshold criteria; and
wherein the signal processing further comprises using a computing
system to execute executable code for automatically determining
whether to (a) edit out the failed frames from the acquisition
series and perform frequency shift error correction via the
frequency error corrector in a manner to at least in part reverse
the frequency shift error estimate on each of the successful
frames, if the number of successful frames meets or exceeds a
minimum threshold number, or (b) retaining at least some of the
failed frames and not performing frequency error correction to the
series via the frequency error corrector if the number of
successful frames is below the minimum threshold.
[0387] In one embodiment, a medical diagnostic method comprises:
using a computing system to execute executable code to implement a
signal quality evaluator for automatically determining, using one
or more microprocessors, whether or not an MRS spectrum acquired
from a region of interest (ROI) in a tissue in a body of a subject
via an MRS pulse sequence series exam of the ROI comprises a
regional signature signal along the MRS spectrum that is
characteristic of lipid.
[0388] In one embodiment, a medical diagnostic method comprises:
using a computing system to execute executable code to implement a
diagnostic processor for providing, using one or more
microprocessors, diagnostic information for diagnosing a medical
condition or chemical environment associated with a region of
interest (ROI) in a tissue in a body of a subject based at least in
part upon at least one chemical factor related to information
extracted from the ROI and associated with lactate (LA) and alanine
(AL) chemicals.
[0389] In one embodiment, a medical diagnostic method, comprising:
using a computing system to execute executable code to implement a
diagnostic processor for providing diagnostic information for
diagnosing a medical condition or chemical environment associated
with a region of interest (ROI) in a tissue in a body of a subject
based at least in part upon a chemical factor related to
information extracted from the ROI and associated with a chemical
and comprising adjusting, using one or more microprocessors, the
chemical factor by an adjustment factor that comprises at least one
of a voxel-related adjustment factor associated with a voxel
prescribed to correspond with the ROI and related to the
information extracted, and a subject-dependent variable-related
adjustment factor associated with the subject.
[0390] In one embodiment, a medical diagnostic method, comprising:
using a computing system to execute executable code to implement a
diagnostic processor for providing processed diagnostic information
for diagnosing a medical condition or chemical environment
associated with a region of interest (ROI) in a tissue in a body of
a subject based at least in part upon taking a first MRS
measurement for a chemical factor taken at a region of an MRS
spectrum acquired from the ROI and associated with a chemical, and
comparing, using one or more microprocessors, the first MRS
measurement against a value derived from a different second
measurement and that is associated with an amount of the chemical
in the ROI.
[0391] In one embodiment, a medical diagnostic method comprises:
using a computing system to execute executable code to implement a
diagnostic processor for providing diagnostic information for
diagnosing a medical condition or chemical environment associated
with a region of interest (ROI) of a tissue in a body of a subject
based at least in part upon a chemical factor related to
information extracted from the ROI and associated with a chemical;
and using a computing system to execute executable code for
providing, using one or more microprocessors, the diagnostic
information that comprises a probability value assigned to a
likelihood that the medical condition or chemical environment meets
certain criteria in the ROI.
TABLE-US-00001 TABLE 1 Examples of CV Variables for DDD-MRS
CHESS-VSS-PRESS pulse sequence for generating MRS spectra useful
for post-processing and diagnosing DDD pain of lumbar
intervertebral discs (e.g. in a 3.0 Tesla MRI system) CV Variable
Value TE (usec) 28000 TR (usec) 1000000 Acquisition Matrix Size 1
Acquisition Matrix Size 1 Number of spatial slices 1 Water
Suppression Method 1 CHESS Flip Angle 1 1050 CHESS Flip Angle 2 800
CHESS Flip Angle 3 125 VSS Band Configuration 7 PRESS Correction -X
axis 1.2 PRESS Correction -Yaxis 1.2 PRESS Correction -Z axis 1.2
Number of Frames 300 PRESS Flip Angle 1 90 PRESS Flip Angle 2 167
PRESS Flip Angle 3 167 PRESS Correction Function 0
TABLE-US-00002 TABLE 2 Example 1, DDD-MRS Clinical Study Group
Demographics and Comparison DDD-MRS Clinical Study - Group
Demographics Pain Patients Asymptomatics p value By SUBJECT (n =
31) n= 12 19 Male 7 (58%) 9 (47%) Female 5 (42%) 10 (53%) Age 46.6
.+-. 9.4 32.4 .+-. 11.3 ** 0.0006 Height 68.3 .+-. 4.1 66.8 .+-.
4.5 0.1805 Weight 172.5 .+-. 38.5 .sup. 151 .+-. 36.3 0.0639 BMI
25.9 .+-. 4.4 23.7 .+-. 3.99 0.0824 By DISCS (n = 52) n= 25 27 Male
16 (64%) 16 (59%) Female 9 (36%) 11 (41%) Age 46.2 .+-. 9.04 35.2
.+-. 14.6 ** 0.0010 Height 68.7 .+-. 4.03 67.9 .+-. 4.5 0.2584
Weight 177.4 .+-. 39.3 157.6 .+-. 39.5 * 0.0381 BMI 26.2 .+-. 4.4
23.8 .+-. 4.3 * 0.0280 By DISCS (n = 52) Pos. Controls Neg.
Controls p value n= 13 39 Male 8 (62%) 24 (62%) Female 5 (38%) 15
(38%) Age .sup. 46 .+-. 9.7 38.7 .+-. 13.9 * 0.0445 Height 68.9
.+-. 3.7 68.1 .+-. 4.4 0.2661 Weight 182.4 .+-. 35.9 .sup. 162 .+-.
40.8 0.0570 BMI 26.9 .+-. 4.2 24.4 .+-. 4.5 * 0.0402
TABLE-US-00003 TABLE 3 Example 1, Comparison of Clinical DDD-MRS
Results (MRS+/-) vs. Positive & Negative Controls, per Disc
DDD-MRS Results DDD-MRS Results % Presumed TRUE Presumed FALSE
Match 3T Pain 23 2 92.0% (All Disco) 3T Pos Control 12 1 92.3%
(Pain, PD+) 3T Neg Control 11 1 91.7% (Pain, PD-) 3T Neg Control 27
0 100.0% (Asymptomatic) 3T Neg Control 38 1 97.4% (All, PD- +
Asymptomatics) 3T All 50 2 96.2%
TABLE-US-00004 TABLE 4 Example 1, Comparison of Clinical DDD-MRS
Results (MRS+/-) vs. Positive & Negative Controls, per
Conventional Diagnostic Performance Measures: Sensitivity,
Specificity, Positive Predictive Value (PPV), Negative Predictive
Value (NPV), Global Performance Accuracy (GPA). DDD-MRS Diagnostic
Performance Sensitivity 92.3% Specificity 97.4% PPV 92.3% NPV 97.4%
GPA 96.2%
TABLE-US-00005 TABLE 5 Example 2, Clinical Data Set (retrospective
and prospective combined) By Subject (pain) (volunteer) mean .+-.
St. Dev. mean .+-. St. Dev. p value Age (yrs) 45.7 .+-. 8.9 36 .+-.
12.9 p = 0.0005 Height (in) 67.8 .+-. 4.sup. 67.2 .+-. 4.4 p =
0.251 Weight (lbs) 166.4 .+-. 39.1 154.3 .+-. 32.7 p = 0.126 BMI
25.2 .+-. 4.4 23.9 .+-. 3.5 p = 0.147 n= 14 28 Male 7 14 Female 7
14 By Disc (per Subject Group) (pain) (volunteer) mean .+-. St.
Dev. mean .+-. St. Dev. p value Age (yrs) 45.9 .+-. 8.8 35.2 .+-.
14.6 p = 0.001 Height (in) 68.1 .+-. 3.92 .sup. 68 .+-. 4.4 p =
0.358 Weight (lbs) 170 .+-. 40 160.7 .+-. 32.1 p = 0.087 BMI 25.5
.+-. 4.4 24.3 .+-. 3.4 p = 0.063 n= 30 49 Male 16 28 Female 14 21
By Disc (per +/-Control Group) (+control) (-control) mean .+-. St.
Dev. mean .+-. St. Dev. p value Age (yrs) 45.3 .+-. 9.2 41.7 .+-.
13.2 p = 0.05 Height (in) 68.4 .+-. 3.7 .sup. 68 .+-. 4.3 p = 0.398
Weight (lbs) 175.4 .+-. 38.5 161.6 .+-. 34.4 p = 0.138 BMI 26.2
.+-. 4.4 24.4 .+-. 3.7 p = 0.093 n= 15 64 Male 8 36 Female 7 28
TABLE-US-00006 TABLE 6 DDD-MRS Disc Phantom: Expected vs. Measured
(Example 4) PG Concentration (mM) LA Concentration (mM) PG/LA Ratio
Phantom/Disc Expected Measured % Diff Expected Measured % Diff
Expected Measured % Diff C/1 7 7.7 9% 7 7.0 0 1 1.09 9% C/2 14 12.4
-11% 14 11.9 -15% 1 1.04 4% C/3 21 21.9 4% 21 25.4 21% 1 0.86 -14%
B/1 28 30.3 8% 28 29.4 5% 1 1.03 3% B/2 42 57.9 38% 14 16.4 17% 3
3.54 18% B/3 14 14.6 4% 42 51.4 22% 0.33 0.28 -14% B/4 28 23.9 -14%
28 25.0 -11% 1 0.96 -4% B/5 42 34.3 -18% 14 11.2 -20% 3 3.07 2%
* * * * *