U.S. patent application number 14/863770 was filed with the patent office on 2016-03-17 for methods and apparatus for accurate characterization of signal coil receiver sensitivity in magnetic resonance imaging (mri).
The applicant listed for this patent is Ohio State Innovation Foundation. Invention is credited to Zhong-Lin Lu, Jinghua Wang.
Application Number | 20160077182 14/863770 |
Document ID | / |
Family ID | 49211849 |
Filed Date | 2016-03-17 |
United States Patent
Application |
20160077182 |
Kind Code |
A1 |
Wang; Jinghua ; et
al. |
March 17, 2016 |
METHODS AND APPARATUS FOR ACCURATE CHARACTERIZATION OF SIGNAL COIL
RECEIVER SENSITIVITY IN MAGNETIC RESONANCE IMAGING (MRI)
Abstract
A method for estimating receiver sensitivity in a magnetic
resonance (MR) system. The method includes steps of acquiring an
image SI(x); determining a transmission function T(x) including
using Bloch's equation with an estimated transmit field
B.sub.1.sup.transmit; generating an estimate of the bias field
B(x); and combining the estimated bias field B(x) and the
transmission function T(x) to determine a receiver sensitivity
S(x).
Inventors: |
Wang; Jinghua; (Columbus,
OH) ; Lu; Zhong-Lin; (Dublin, OH) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Ohio State Innovation Foundation |
Columbus |
OH |
US |
|
|
Family ID: |
49211849 |
Appl. No.: |
14/863770 |
Filed: |
September 24, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13778226 |
Feb 27, 2013 |
9146293 |
|
|
14863770 |
|
|
|
|
61603542 |
Feb 27, 2012 |
|
|
|
61691040 |
Aug 20, 2012 |
|
|
|
Current U.S.
Class: |
702/57 |
Current CPC
Class: |
G01R 33/58 20130101;
G01R 33/44 20130101; G01R 33/5611 20130101; G01R 33/246 20130101;
G01R 33/5612 20130101; G01R 35/00 20130101; G01R 33/36 20130101;
G06T 7/00 20130101; G01R 33/5659 20130101 |
International
Class: |
G01R 33/58 20060101
G01R033/58; G01R 33/36 20060101 G01R033/36 |
Claims
1. A method for estimating receiver sensitivity in a magnetic
resonance (MR) system, the method comprising: acquiring an image
SI(x); determining a transmission function T(x) comprising using
Bloch's equation with an estimated transmit field
B.sub.1.sup.transmit; generating an estimate of the bias field
B(x); and combining the estimated bias field B(x) and the
transmission function T(x) to determine a receiver sensitivity
S(x).
2. The method of claim 1, wherein acquiring an image SI(x)
comprises acquiring a plurality of signal intensity images using
different imaging parameters.
3. The method of claim 2, wherein the estimated transmit field
B.sub.1.sup.transmit is determined by estimating a relative flip
angle map using the image SI(x), for the plurality of signal
intensity images, each acquired using different imaging parameters,
estimating a relative transmit field map, for each nominal flip
angle, determining a calibration factor based on a linear
relationship between the nominal flip angles and the measured flip
angles, and for each nominal flip angle, calculating an absolute
flip angle corresponding to the nominal flip angles.
4. The method of claim 3, wherein determining a transmission
function T(x) comprising using Bloch's equation with an estimated
transmit field B.sub.1.sup.transmit is performed for each nominal
flip angle.
5. The method of claim 1, wherein generating an estimate of the
bias field B(x) comprises generating an estimate using an image
SI(x) acquired using a given sequence with minimized TE and
geometric distortions.
6. The method of claim 1, wherein generating an estimate of the
bias field B(x) comprises generating an estimate using an image
SI(x) acquired while minimizing at least one of B.sub.0
inhomogeneity, chemical shift, and magnetic susceptibility
artifacts.
7. The method of claim 1, wherein generating an estimate of the
bias field B(x) comprises generating an estimate using at least one
algorithm selected from an N3, FSL-FMRIB, and SPM algorithms.
8. The method of claim 1, wherein combining the estimated bias
field B(x) and the transmission function T(x) to determine a
receiver sensitivity S(x) comprises determining a ratio between the
estimated bias field B(x) and the transmission function T(x).
9. The method of claim 3, wherein the different imaging parameters
for estimating a relative transmit field map include at least one
of nominal flip angle, repetition time (TR), number of RF pulses,
and excitation frequency.
10. The method of claim 3, wherein estimating a relative flip angle
map comprises estimating a relative flip angle map using at least
one method selected from a double flip angle method, and an actual
flip angle imaging method.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is a division of U.S. patent application
Ser. No. 13/778,226, filed Feb. 27, 2013, which claims the benefit
of U.S. Provisional Application Nos. 61/603,542 filed Feb. 27, 2012
and 61/691,040 filed Aug. 20, 2012, each of which is incorporated
herein by reference in its entirety.
BACKGROUND
[0002] The present invention relates to nuclear magnetic resonance
(NMR) systems and methods for measuring in vivo receiver
sensitivity maps.
[0003] Accurate characterization of receiver sensitivity is
necessary for performing magnetic resonance (MR), particularly fast
MR imaging (parallel MR imaging) with multi-channel coils composed
of a plurality of coil elements. During parallel magnetic resonance
imaging, different receive coils sample different portions of the
k-space concurrently. The data from each coil is transformed or
"unfolded" in accordance with its sensitivity in reconstructing the
final image. The accuracy of the final image depends on accurate
characterization of the sensitivity of the coil. Accurate
characterization of coil sensitivity is also critical for accurate
and precise quantification of various MRI parameters, such as
macromolecule, perfusion, and molecule concentration.
SUMMARY OF THE INVENTION
[0004] Three methods for estimating in vivo receiver sensitivity
are disclosed. The electromagnetic field method is only available
for estimating the receiver sensitivity of transceiver coils. The
method is based on measurements of the transmit field of
transceiver coils. The sensitivity of a receive-only coil can be
obtained using a reference method with the measured transceiver
coil receive sensitivity. The other two methods, the bias field
method and the uniform transmit field method, are based on
post-processing of MR signal intensity. The bias field method
includes estimations of the transmit field and bias field. Receiver
sensitivity is determined using the estimated transmit field and
bias field. The uniform transmit field method is realized using
either transmit field shimming or adiabatic pulse methods. Because
transit fields are assumed to be uniform after applying either
transmit field shimming or adiabatic pulses, receiver sensitivity
is determined by estimating the bias field of the acquired image
using the uniform transmit field. These approaches can be used to
reduce wrapping or aliasing artifacts in parallel MR image
reconstruction, to reduce artifacts and variability in quantitative
MRI, and to improve the accuracy and precision of molecule
concentration estimation in quantitative MRS or MRSI.
[0005] The general objective of the invention is to quantitatively
estimate receiver sensitivity in any coil configuration. According
to embodiments of the invention, three methods (the bias field
method, uniform transmit field method, and electromagnetic method)
are disclosed which can be used to achieve this and other objects
of the invention.
[0006] An object of the invention is to provide methods which would
facilitate the characterization of receiver sensitivity in NMR
systems. The present invention contains procedures which may be
executed by the NMR system to calibrate, check, or diagnose image
and signal quality. It is particularly useful in evaluating the
performance of RF receiver coils used in NMR systems and how their
performance is affected by the subject of the NMR scan and by each
other.
[0007] An object of the invention is to provide an exact estimate
of receiver sensitivity for improving the performance of image
reconstruction in parallel imaging, particularly for SENSE-based
parallel imaging.
[0008] An object of the invention is to provide an exact estimate
of receiver sensitivity for improving the accuracy and precision of
absolutely quantitative MRI using internal and external references.
As a result, it will greatly reduce the variability of MRI acquired
across sites and in longitudinal studies.
[0009] An object of the invention is to provide an exact estimate
of receiver sensitivity for exploration of new biomarkers (such as
electric properties) for MRI. It can be applied in functional MRI,
diagnosis of diseases, electromagnetic therapy, and human safety in
electromagnetic environment.
[0010] A more specific object of the invention is to provide
information for RF dynamic shimming by adjusting the phase and
magnitude of each coil element. As a result, after RF dynamic
shimming, the inhomogeneous signal intensities from non-object
characteristics can be reduced or eliminated.
[0011] Another general object of the invention is to provide
methods for correcting the influence of inhomogeneous receiver
sensitivity on quantification of various MR parameters, such as
signal intensity, Ti, and perfusion.
[0012] The present invention relates to methods for mapping
receiver sensitivity of RF receiver coils. More particularly,
embodiments of the invention include methods and systems for:
[0013] (1) Producing a set of signal intensity images by an MR scan
conducted with different imaging parameters for estimating the
transmit field of a transmit coil and receiver sensitivity of a
receive coil. In order to reduce scan time for estimating receiver
sensitivity, in various embodiments images are acquired in low
resolution using fast acquisition methods (such as segmented echo
planar imaging sequences, parallel imaging acquisition, and k-space
under-sampling);
[0014] (2) Estimating the relative flip angle map, and then
relating the transmit field maps with the images acquired at
different flip angles;
[0015] (3) Determining the calibration factor of the transmit RF
field or flip angles with the assumption of a linear relationship
between the nominal and measured flip angles;
[0016] (3) Calculating absolute flip angles corresponding to given
nominal flip angles;
[0017] (4) Estimating the bias field of an image acquired using a
given sequence with minimized TE and geometric distortion;
[0018] (5) Estimating transmit function according to Bloch's
equation for a given nominal flip angle;
[0019] (6) Determining the relative receiver sensitivity for
receiver coils with the ratio between the bias field of an image
and the transmit function of the image.
[0020] Accordingly, in one embodiment the invention provides a
method for estimating receiver sensitivity in a magnetic resonance
(MR) system. The method includes steps of acquiring an image SI(x);
determining a transmission function T(x) including using Bloch's
equation with an estimated transmit field B.sub.1.sup.transmit;
generating an estimate of the bias field B(x); and combining the
estimated bias field B(x) and the transmission function T(x) to
determine a receiver sensitivity S(x).
[0021] In another embodiment the invention provides a method for
estimating receiver coil sensitivity in a magnetic resonance (MR)
system. The method includes steps of acquiring an image SI(x) of a
receiver coil, wherein acquiring includes obtaining the image SI(x)
using various combined RF pulses and RF shimming methods, such that
the image SI(x) has a uniform transmission function T(x); and
estimating a bias field B(x) using the image SI(x), wherein the
receiver coil sensitivity includes the estimated bias field
B(x).
[0022] In still another embodiment, the invention includes a method
for estimating receiver coil sensitivity in a magnetic resonance
(MR) system. The method includes steps of estimating a transmit
field of a transceiver coil; determining a transverse
electromagnetic field B.sub.xy of the transceiver coil using at
least one phase image; determining a receiver sensitivity of the
transceiver coil using the transmit field and transverse
electromagnetic field; using the transceiver coil as a reference
coil for estimating a coil sensitivity of a second receiver coil;
and estimating a receiver coil sensitivity using a reference method
with the estimated receive sensitivity of the transceiver coil.
[0023] In yet another embodiment, the invention includes a method
for estimating receiver coil sensitivity in a parallel magnetic
resonance (MR) system, where the parallel MR system has a plurality
of coils in a coil array including a reference coil. The method
includes steps of estimating receiver coil sensitivity in each of
the plurality of coils in the coil array of the parallel MR system
using one of an electromagnetic field method, a bias field method,
and a uniform transmit field method. Each of the plurality of coils
in the coil array includes one of a surface coil, a volume coil, a
set of coil arrays, or an antenna element of a coil array. The
method further includes steps of obtaining a reference image using
the reference coil; generating a measured sensitivity map using the
reference coil; and for each of the plurality of coils of the coil
array, calculating a receiver coil sensitivity map using a ratio of
the reference image and an image received by the respective coil of
the coil array, with the result of the division being multiplied by
the measured sensitivity map of the reference coil.
[0024] Other aspects of the invention will become apparent by
consideration of the detailed description and accompanying
drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0025] FIG. 1 is a flowchart employed in a magnetic resonance
imaging system suitable for estimating in vivo receiver sensitivity
using the bias field method.
[0026] FIG. 2 shows a transmit field (a, c) and receive sensitivity
(b, d) of a body coil with a uniform phantom and in vivo human
brain at 3.0 T.
[0027] FIG. 3 shows a simulated left-handed circularly
B.sub.1.sup.transmit (a) and right-handed circularly
B.sub.1.sup.transmit (b) rotating components of B.sub.1 and their
difference (c) from a linear birdcage coil at 10, 20, 64, and 128
MHz.
[0028] FIG. 4 shows an image of a uniform phantom (a) and corrected
images using N3 (b), SPM (c), and FSL-FMRIB (d).
[0029] FIG. 5 Shows a receiver sensitivity of a uniform phantom
using the pre-scan method (a), no contrast image method (b), and
proposed BF method (c).
[0030] FIG. 6 shows a receiver sensitivity of a human subject using
the pre-scan method (a), minimal contrast image method (b), and
proposed BF method (c).
[0031] FIG. 7 shows distributions of computer simulated magnitudes
of transverse RF field (a), transmit field (b), and receiver
sensitivity (c) of a 10-cm diameter single-loop transceiver surface
coil for a 16-cm diameter spherical phantom.
[0032] FIG. 8 shows calculated (a) and experimental (b) axial
gradient echo images at the nominal flip angles of 90.degree.,
respectively.
DETAILED DESCRIPTION
[0033] In the following detailed description, numerous specific
details are set forth in order to provide a thorough understanding
of embodiments. However it will be understood by those of ordinary
skill in the art that the embodiments may be practiced without
these specific details. In other instances, well-known methods,
procedures, components and circuits have not been described in
detail so as not to obscure the embodiments.
[0034] Three methods for estimating in vivo MR receiver coil
sensitivity are disclosed herein: an electromagnetic field method,
a bias field method, and a uniform transmit field method. The
electromagnetic field method is only available for estimating the
receiver sensitivity of transceiver coils. The method is based on
measurements of the transmit field of transceiver coils. The
sensitivity of a receive-only coil can be obtained using a
reference method with the measured transceiver coil receive
sensitivity. The other two methods, the bias field method and the
uniform transmit field method, are based on post-processing of MR
signal intensity. The bias field method includes estimations of the
transmit field and bias field. Receiver sensitivity is determined
using the estimated transmit field and bias field. The uniform
transmit field method is realized using either transmit field
shimming or adiabatic pulse methods. Because transit fields are
assumed to be uniform after applying either transmit field shimming
or adiabatic pulses, receiver sensitivity is determined by
estimating the bias field of the acquired image using the uniform
transmit field. These approaches can be used to reduce wrapping or
aliasing artifacts in parallel MR image reconstruction, to reduce
artifacts and variability in quantitative MRI, and to improve the
accuracy and precision of molecule concentration estimation in
quantitative MRS or MRSI.
[0035] In order to obtain a detectable NMR signal, the net
magnetization of nuclear spins is rotated away from the static
magnetic field. The rotation is performed using a radio frequency
excitation field with the same frequency as the Larmor frequency of
the nucleus. The angle through which the net magnetization is
rotated is determined by the field strength of the radio frequency
excitation signal and its duration. In the end of the radio
frequency excitation pulse, the nuclei, in relaxing to their normal
spin conditions, generate a decaying signal at the same radio
frequency as that used for excitation. The NMR signal is picked up
by a receive coil, amplified and processed by the NMR system to
produce MR spectroscopy and/or MR images.
[0036] Before the commencement of each NMR scan, it is common
practice to adjust the strength of the transmitted RF excitation
field and the gain of the RF receiver to insure that the RF
excitation pulses have the optimal frequency, strength and duration
to evoke the desired NMR signal. However, this does not necessarily
mean that the expected RF excitation field will be produced
uniformly throughout the region of interest, or that the resulting
NMR signals will be received uniformly from all locations in the
region of interest. RF fields produced by most transmit coils
loaded with a study subject are not homogeneous and the receive
fields of most receive coils are similarly not homogeneous. This is
particularly true of imperfect coil configurations, such as surface
coils and phase array coils. Even if the transmit and receive
fields are homogeneous in free space or in the unloaded condition,
wave behavior and penetration of the RF field into the subject may
give rise to a non-uniform transmit field and receiver sensitivity
throughout the region of interest. Moreover, incorrect calibration
of the RF pulse amplitude, instability or drift of the RF amplifier
or other RF electronics, can lead to a non-uniform transmit field.
Also, mutual inductance between transmit and receive coils may
cause further inhomogeneities in the transmit field and receiver
sensitivity. Either inhomogeneous transmit field or inhomogeneous
receiver sensitivity or both can gives rise to ghost artifacts in
signal intensity and introduce variations of MR data over time in
longitudinal studies and across different sites. The variations
become apparent in parallel acquisition, high field strengths, and
new scanner designs such as open access scanners. Recently,
quantitative MRI using internal and external references for MRI/MRS
has been used to reduce such variations. However, their
applications are limited by various factors, particularly by
receiver sensitivity.
[0037] Parallel imaging techniques have been developed to
accelerate MR data acquisition. In parallel imaging, multiple
receive coils are used to accelerate data acquisition in one or
more dimensions by exploiting the spatial dependency of the
sensitivity of phased array coils. Parallel imaging has been shown
to be successful in reducing scan time, image blur, and geometric
distortions. Moreover, parallel imaging can be used to improve
spatial and/or temporal resolution as well as volumetric coverage.
There are several types of parallel imaging reconstruction methods.
These methods can generally be divided into two categories based on
how they treat the reconstruction problem: 1) SENSE-based
techniques (Sensitivity Encoding) estimate coil sensitivity
profiles from low-resolution calibration images, which can then be
used to unwrap aliased pixels in image space using a direct
inversion algorithm; and 2) autocalibrating parallel imaging-based
methods, such as GRAPPA (Generalized Auto-calibrating Partially
Parallel Acquisition) and ARC (Autocalibrating Reconstruction for
Cartesian sampling), that calculate reconstruction weights
necessary to synthesize signals from acquired data using an
algorithm that does not require coil sensitivity estimates.
Although both SENSE- and autocalibrating parallel imaging-based
approaches have been successful in practice, accurate
characterization of coil sensitivity is a prerequisite for
successful application of SENSE-based techniques.
[0038] Various in vitro and in vivo methods have been proposed to
estimate receiver sensitivity. The in vitro methods include the
simulation method and phantom pre-scan method. The simulation
method estimates the receive sensitivity based on the configuration
of the receiver coil and the electromagnetic properties of the
loaded subject. It is time consuming and not accurate. The phantom
pre-scan method assumes that the receiver sensitivity of the loaded
condition is identical to that of a uniform phantom. The method is
apparently not accurate. Generally, receiver sensitivity depends on
both coil configuration and properties of the loaded subject.
[0039] The in vivo methods include intensity-based methods,
field-based methods, and k-space calibration methods. The
intensity-based methods include the pre-scan method, the minimal
contrast method (MCI), and the uniform magnetization method. In the
pre-scan method, the sensitivity distributions of the respective RF
receiving coils are calculated based on the assumption that the
body coil has a relatively uniform sensitivity distribution, which
is however only approximately true at low field strengths when the
wave behavior and RF penetration of the RF magnetic field can be
ignored. When the field strength is high, the assumption is no
longer valid. Moreover, sometimes an MRI system may not have any
body coil or volume coil, in which case the pre-scan method cannot
be used to estimate receiver sensitivity. The minimal contrast
image method and uniform magnetization method can estimate receiver
sensitivity in vivo and greatly improve the accuracy of receiver
sensitivity estimation. But these methods are limited in their
practical applications, particularly in configurations involving
multiple tissue types.
[0040] The field-based methods include the reciprocity principle
method, the rotating-object method, and calibration based on
transmit field. At low field strengths, the difference between the
transmit field and receiver sensitivity for a transceiver coil is
small and can be ignored because the phase is constant inside an
object. The reciprocity principle works very well. In that case,
quasi-static approximation (Biot-Savart's Law) can depict the
transmission and reception fields very well. Thus, receiver
sensitivity can be replaced with the transmit field at low field
strengths for transceiver coils. Various methods have been proposed
for estimating transmit fields in vivo. These methods can be
categorized into MR amplitude based and MR phase based methods. MR
amplitude based methods include double flip angle method, dual
pulse spin echo, actual flip angle imaging, and stead state method.
MR phase based methods include Bloch Siegert shift method, and
phase method. However, the wavelength of the RF magnetic field in
high fields becomes less than 1/10 of the size of the object. The
phase distribution of the magnetic field inside the object must
then be considered. The difference between the transmit field and
receiver sensitivity becomes significant at high field strengths
and the reciprocity principle no longer works. Replacing receiver
sensitivity with transmit field will introduce significant errors
in quantitative MRI and large artifacts in parallel imaging
reconstruction. It is known that the measured transmit field is a
left-handed circularly-polarized component of a transmit/receive
coil. When the relative direction between equilibrium magnetization
and the main magnetic field is reversed by either rotating the
object or inverting the main magnetic field, the measured
pseudo-transmit field is a right-handed circularly polarized
component of the coil before rotating the object or inverting the
field. The right-handed component is the receiver sensitivity of
the transmit/receive coil. A receiver sensitivity map can be
calculated by the ratio map .alpha.(x) and transmit field. The
ratio map .alpha.(x) is the ratio between transmit field and
receiver sensitivity, and is obtained in 3D spaces inside a volume
coil, although obtaining .alpha.(x) is time consuming. This method
is only applicable to transceiver coils and cannot be applied to a
transceiver array coil when the RF phase and amplitude of each coil
element is adjusted for each subject by RF shimming.
[0041] Coil sensitivity calibration includes auto- or
self-calibrating methods. The idea is to record coil sensitivity
information directly during the actual scan by adding a small
number of additionally acquired fully Fourier-encoded
auto-calibration lines. Direct sensitivity calibration for each
image is beneficial in combination with flexible coil arrays or for
imaging of uncooperative patients. Various auto-calibration methods
have also been used to estimate receiver sensitivity for parallel
imaging reconstruction because coil sensitivity varies slowly and
smoothly over space. The auto-calibration methods also have some
drawbacks, although they are very good for parallel imaging
reconstruction. Receiver sensitivity is a property of the receiver
coil, independent of transmit coil, acquisition sequence, and
imaging parameters. Nevertheless, receiver sensitivity estimated
using auto-calibration methods depends to a high degree on these
factors, indicating the inaccuracy of these methods. This is
because all calibration methods introduce a virtual receiver coil
whose receiver sensitivity is equal to the square root of the sum
of the squares of receiver sensitivity of each coil element and
assume that the sensitivity of the virtual coil is uniform. In
practice, this assumption is problematic, particularly at high
field strengths and if the imaged object is of a large size. The
errors in estimated receiver sensitivity using auto-calibration
methods are not problematic for qualitative parallel image
reconstruction because artifacts of reconstruction are dominated by
relative receiver sensitivity. However, they are very problematic
for quantitative parallel image reconstruction or quantitative MRI
using internal and external references. In addition, the
auto-calibration methods also require additional scan time to
obtain additional reference k-space line(s).
[0042] Based on the assumption that receiver sensitivity of coils
changes slowly and smoothly in MR experiments, the sensitivity maps
for unfolding can be provided by low-pass filtering. However, the
peripheral portions of the image frequently show abnormal diverging
data, making it difficult to obtain a smooth and stable sensitivity
map.
[0043] 1. Methods for Receiver Sensitivity Estimation
[0044] a. The Bias Field (BF) Method for Receiver Sensitivity
[0045] The definition of the bias field is given by:
SI(x)=B(x)I(x)+n (1)
where SI(x) is the measured signal intensity of a corrupted image,
I(x) is the true intensity of the image, B(x) is the corresponding
bias field of the image, and n is additive noise.
[0046] The disclosed BF method for receiver sensitivity assumes
that the bias field mostly results from transmit field and receiver
sensitivity:
B(x)=S(x)T(x), (2)
where S(x) is the receiver sensitivity, and T(x) is the
transmission function that is calculated using Bloch's Equation
with the estimated transmit field B.sub.1.sup.transmit, which can
be estimated using a number of methods. The methods include,
without limitation, the double flip angle method, Simulated echo
method, actual flip angle imaging, and steady state method,
although other methods are also possible. The other contributions,
such as B.sub.0 inhomogeneity, chemical shift, and magnetic
susceptibility artifacts, can be minimized using certain sequences
with optimized imaging parameters. For example, they can be
minimized using a gradient echo sequence with a very short TE.
[0047] Various methods, including the nonparametric nonuniform
intensity normalization (N3) algorithm, SPM, and FSL-FMRIB, can
also be used to estimate the bias field or signal intensity
inhomogeneity of acquired images. In the N3 method, bias field is
estimated by sharpening the intensity histogram using Gaussian
devolution. In FSL-FMRIB, estimation of the bias field is based on
a hidden Markov random field model and an associated
Expectation-Maximization (EM) algorithm. SPM is based on a Gaussian
mixture tissue model, EM algorithm and Levenberg-Marquardt
optimization. Additionally, many other algorithms for estimating
bias field have been developed. In the invention, at least one
algorithm is used to estimate the bias field and generate the
receiver sensitivity S(x). It can be obtained from the estimated
B(x) and T(x):
S(x)=B(x)/T(x). (3)
[0048] Both transmit field and receiver sensitivity of coils
changes slowly and smoothly in MR experiments. That is, there is
small error between low resolution and high resolution transmit
field and receiver sensitivity. Thus, receive sensitivity can be
estimated using the images acquired with low resolution in order to
save the scan time and cost.
[0049] The receive sensitivity is not dependent of imaging
parameters, acquisition sequences, and k-space sampling strategies.
Various fast acquisition techniques, such as segmented echo planar
imaging sequences, parallel imaging acquisition, and k-space
under-sampling can be used to reduce the scan time and cost for
estimating the receive sensitivity.
[0050] b. The Uniform Transit Field Method
[0051] For a uniform transmit field, the transmit function is a
constant. The relative receiver sensitivity of a coil whose
transmit function is normalized to 1 is equal to its estimated bias
field.
[0052] For a single channel coil or volume coil, a uniform transmit
field can be obtained using one or more methods including, without
limitation, parallel transmit, RF shimming, adiabatic pulses and
composite RF pulses. The bias field of the images acquired with a
uniform transmit field can be estimated using various methods
(Belaroussi B, et al., Med Image Anal. 10:234; 2006, incorporated
by reference in its entirety). The relative receiver sensitivity of
the coil is equal to the estimated bias field. For multi-channel
coils, the bias field of each channel can be estimated using
various methods. The relative receive sensitivity of each channel
is equal to the corresponding estimated bias field.
[0053] Compared with the bias field method, the uniform transmit
field method reduces the time needed to estimate the transmit field
and the errors resulted from the transmit field. However,
performing RF shimming requires additional time. Combined RF pulses
may also introduce SAR problems. This limits its application at
higher field strengths.
[0054] c. The Electromagnetic Method:
[0055] A novel method, the electromagnetic method, can be developed
to estimate the receiver sensitivity of a transceiver coil,
particularly for body coils. When a transceiver coil with a linear
driven is used for the simultaneous transmit coil and receive coil,
computer simulation (FIG. 7) and indirect MR experiment (FIG. 8)
have confirmed that the transmit field B.sub.1.sup.transmit and
receiver sensitivity B.sub.1.sup.receive has a relationship as
follows:
|B.sub.1.sup.transmit.parallel.B.sub.1.sup.receive|=|B.sub.x.sup.2+B.sub-
.y.sup.2|
where B.sub.x and B.sub.y are component of the transceiver RF
electromagnetic field along x direction and y direction,
respectively. As mentioned before, the transmit field
B.sub.1.sup.transmit of a transceiver coil can be estimated using
various methods. If a transverse electromagnetic field
B.sub.xy=(B.sub.x.sup.2+B.sub.y.sup.2).sup.1/2 of the transceiver
coil can be determined, The receiver sensitivity
|B.sub.1.sup.receive| of the transceiver coil can be estimated
using
B x 2 + B y 2 / B 1 transmit . ##EQU00001##
[0056] For most efficient RF transceiver coils, for example a body
coil, it follows B.sub.x.sup.2+B.sub.y.sup.2>>B.sub.z.sup.2.
The radiofrequency field of the transceiver coil can be
approximated as B.sub.1.sup.2.apprxeq.B.sub.x.sup.2+B.sub.y.sup.2.
That is, B.sub.x.sup.2+B.sub.y.sup.2 can be obtained from B.sub.1,
which can be estimated from phase images using various MRI methods
(Morrell G R, Magn Reson Med. 60:889-894, 2008; Sacolick L I et.
al., Magn Reson Med. 63:1315-1322, 2010, each of which is
incorporated by reference in its entirety). The receiver
sensitivity of any coil can be estimated using a reference method
with the estimated receiver sensitivity of the transceiver
coil.
[0057] For some transceiver coil configurations, for example
surface coil and cardiac phased array coil,
B.sub.x.sup.2+B.sub.y.sup.2 is comparable to B.sub.z.sup.2.
B.sub.xy have to be estimated using special sequences and/or
imaging parameters. Then the receiver sensitivity
|B.sub.1.sup.receive| of these transceiver coils can be estimated
using
B x 2 + B y 2 / B 1 transmit . ##EQU00002##
[0058] For quadrature driven transceiver coil, the phases used as a
transmit coil are totally different from those used as a receive
coil. For quadrature driven transmit coil, phase distribution will
maximize the efficiency of the transmit. If this phase distribution
is used for a receive, it will minimize the efficiency of receive.
But for quadrature driven receive coil, phase distribution will
maximize the efficiency of the receive. If the phase is used for a
transmit, it will minimize the efficiency of the transmit.
Therefore, a quadrature transceiver coil used as a quadrature
transmit coil is different from that used as a quadrature receive
coil. That is, they should be regarded as two different coils. Only
for linear drive transceiver coil, the coil used as a transmit coil
is totally identical to that used as a receive coil. In this case,
its receive sensitivity can be determined using the electromagnetic
method mentioned above. For quadrature driven transceiver coil, it
must be converted from quadrature to linear transmit mode so that
receive sensitivity of the coil could be estimated using the
electromagnetic field method.
[0059] The radiofrequency transverse electromagnetic field of the
transceiver coil can be obtained using dual-coil methods. The first
phase image can be acquired at the on-resonance excitation of the
transceiver coil. The second phase image can be acquired at the
on-resonance excitation of the transceiver coil and off-resonance
excitation of a transceiver coil being measured. The difference of
the two phase images can provide the electromagnetic field of the
transceiver coil being measured.
[0060] Compared with the bias field method, this method will be
more accurate. It does not require extra time to get the bias
field, and can reduce the error of the estimated receive
sensitivity resulted from the bias field.
[0061] 2. Receiver Sensitivity Estimation
[0062] Similarity metrics (SM), introduced to quantitatively
describe or compare the difference between two MR images (image A
and image B), is expressed mathematically as:
SM ( x ) = SI A ( x ) - SI B ( x ) SI A ( x ) + SI B ( x ) , ( 4 )
##EQU00003##
where SI.sub.A (x) and SI.sub.B (x) are the magnitudes of images A
and B at location x. As defined in Eq. (4), SM ranges from 0 to 1,
with 1 indicating that image A and image B are completely
different, and 0 indicating perfect match.
[0063] The coefficient of variation (CV) is used to evaluate the
uniformity of MR signal intensity and estimated parameters (such as
receiver sensitivity). CV is defined as
CV = .delta. .mu. , ( 5 ) ##EQU00004##
where .mu. and .delta. are the mean and standard deviation of the
signal or parameter.
[0064] FIG. 1 is a simplified flowchart employed in a magnetic
resonance imaging system suitable for correcting inhomogeneous
signal intensities with estimated transmit field and receiver
sensitivity. The procedure includes: (1) acquire multiple images
for transmit field estimation; (2) acquire an image with minimized
artifacts (susceptibility artifacts and geometry distortions) using
the shortest TE and other optimized parameters; (3) estimate the
transmit function of the image with minimized artifacts according
to Bloch's equation; (4) estimate the bias field of the image with
minimized artifacts; and (5) estimate relative receiver sensitivity
from the bias field and the transmit function.
[0065] FIG. 2 shows the transmit field (FIGS. 2a, c) and receiver
sensitivity (FIGS. 2b, d) of a body coil with a uniform phantom and
an in vivo human brain at 3.0 T. Their transmit fields differ
significantly from their respective receiver sensitivities. Their
differences can be up to 15% for the uniform phantom, and 20% for
the human brain. The estimated receiver sensitivity is noisy
because the noise in the MR images is amplified during the
estimation of receiver sensitivity. But low pass filtering can
reduce this noise and will not affect the accuracy of the estimated
receiver sensitivity. Moreover, the CV of the body coil receiver
sensitivity is about 14.5% for the phantom, and 22.6% for the human
brain. Thus, replacing receiver sensitivity of the body coil with
its transmit field can lead to inaccurate parallel imaging
reconstruction and quantitative MRI or MRS.
[0066] In early MRI development, receiver sensitivity was replaced
with transmit field. But FIG. 2 show that the replacement can lead
to large errors at 3.0 T. In order to quantify the error, we
simulated the transmit field and receiver sensitivity at different
frequencies or field strengths (FIG. 3). At less than 20 MHz, the
SM between the transmit field and receiver sensitivity is 0.01 and
the two sensitivities can be used interchangeably. However, the SM
can be more than 0.1 in some regions and become significant at
Larmor frequencies greater than 20 MHz. Transmit field and receiver
sensitivity cannot replace each other. This is because the
definitions of transmit field and receive sensitivity are totally
different in electromagnetic field theory and MRI: The relationship
between electromagnetic transmitter and receiver is described using
Maxwell's equations, but the relationship between MRI transmitter
and receiver is described by Bloch's equation and Maxwell's
equations. Both the transmit field and receiver sensitivity are
B.sub.1 in electromagnetic field, but they are B.sub.1.sup.transmit
and B.sub.1.sup.receiver in MRI experiments. At low fields, the
difference between the transmit field and receiver sensitivity for
a transceiver is small and can be neglected because the phase of
the magnetic field is constant inside an object. However, when the
size of the object becomes larger than one-tenth of the wavelength
of the magnetic field, phase distribution inside the object must be
considered. In this case, if transmit field is used to replace
receiver sensitivity, the maximum error can be as high as 10% in
some regions of the object. Therefore, transmit field and receiver
sensitivity cannot replace each other.
[0067] There is no gold standard to evaluate the quality of
estimated bias field, particularly for in vivo MR images. A uniform
phantom is used to evaluate the performance of several popular bias
field estimation procedures. For the uniform phantom, signal
intensity inhomogeneity only results from bias field. The original
MR image of a uniform phantom and its corrected signal
distributions using N3, SPM, and FSL-FMRIB are shown in FIG. 4. All
these correction methods greatly improved the severely
inhomogeneous signal intensities in the raw image. The CV of the
raw image is 11.9%. The CV of the corrected image using the N3,
SPM, and FSL-FMRIB are 7.3, 8.0 and 7.4%, respectively. It is noted
that the mean of the corrected signal intensity using these methods
is almost identical (their difference of mean is less than 3%)
after signal intensity inhomogeneity corrections. Because it
slightly outperformed the other methods, N3 is used to estimate
bias field for determining receiver sensitivity in this
invention.
[0068] In FIG. 5, the receiver sensitivity of a phased-array coil
for a uniform phantom is estimated using different methods. The
result using the no contrast image (NCI) method is considered as
the ground truth (FIG. 5a). The results from the proposed BF method
and the NCI method are very similar, while higher sensitivity
estimations are obtained at both the top and bottom parts of the
image using the pre-scan method (FIG. 5b). The observable
differences between FIGS. 5a and 5c are due to the induced
artifacts at the center and boundary of the phantom when bias field
is estimated using the N3 method in FIG. 4b. Commonly admitted
reason for the failure of the pre-scan method is the incorrect
assumption that the receiver sensitivity of a body coil is uniform.
In this study, the CV of the estimated body coil receiver
sensitivity is around 14.5% for the phantom (FIG. 2b).
[0069] FIG. 6 shows the receiver sensitivity for a human subject
using different methods. The estimated receiver sensitivity from
the BF method is more similar to that of the MCI method than the
pre-scan method. But it is difficult to compare the performance of
the proposed BF method and the MCI method for in vivo human brain.
If the region of interest (ROI) contains one or two tissue types,
the MCI method has an advantage over the BF in the accuracy of
estimating receiver sensitivity. But if the ROI contains more than
two type tissues, the proposed BF method has advantages over MCI in
terms of time efficiency and the accuracy of estimated receiver
sensitivity.
[0070] The pre-scan method assumes that the receiver sensitivity of
a body coil is uniform. At low field MRI or when imaging objects
with low permittivity, this is a good approximation. But wave
behavior becomes apparent with high field strength or high
permittivity of imaged objects. The wave behavior effect will lead
to big variability of receiver sensitivity of the body coil. For
example, the size of the human brain, around 18 cm in diameter, is
comparable to the wavelength of the RF in the brain, about 24 cm at
3.0 Tesla. As a result, coherent addition of electromagnetic waves
gives rise to the maximal receiver sensitivity at the center of the
brain and up to 7% non-uniformity or variability of sensitivity
across the brain. Thus, the assumption that receiver sensitivity of
the body coil is uniform does not hold in this case. The estimated
receiver sensitivity using pre-scan method would not be accurate at
high field strengths. Our quantitative analyses indicate that the
CV is about 14.5% for the phantom, and 22.6% for the human
brain.
[0071] The MCI method has been regarded as a unique way to estimate
in vivo receiver sensitivity. However, the MCI method needs
additional time to obtain the minimal contrast images. In addition,
residual contrasts in the minimal contrast image can lead to errors
in estimated receiver sensitivity (or residual tissue structure) as
shown in FIG. 5b. Particularly in regions containing multiple
tissue types and multiple pathological changes for the same tissue
type, the MCI method needs more time to obtain the minimal contrast
image, and the residual contrasts may increase, limiting the
application of this method.
[0072] Our proposed BF method assumes that the bias field mainly
results from transmit field and receiver sensitivity. The other
factors, such as B0 inhomogeneity and susceptibility artifacts, can
be minimized using optimized imaging parameters. For example,
minimizing the TE of a gradient echo sequence can reduce the effect
of B0 inhomogeneity and susceptibility artifacts on brain signal
inhomogeneity. The major error of the proposed method comes from
the estimation of the bias field. As shown in FIG. 2, the CV of the
corrected image using N3 is still 7.3%. The SNR of the raw image is
50. The variation of signal intensity from noise should be about 2%
(1/SNR). The CV of the corrected image using N3 is much higher than
the variation from noise. The reason is that artifacts caused by
the N3 bias field correction are around 4% at the center and 5% at
the boundary of the corrected image. With improvement of the
accuracy of estimated bias field, the accuracy of estimated
receiver sensitivity using the proposed BF method will be greatly
improved. This method offers an alternative and improved technique
to estimate in vivo receiver sensitivity. It has a significant
advantage over the MCI method in terms of time efficiency, and can
be easily extended for any tissue and coil configurations.
[0073] The other major error of estimated receiver sensitivity
using the BF method comes from the estimate of the transmit field.
Since transmit field and receive sensitivity change slowly and
smoothly in MRI experiments, they can be estimated using low
resolution images and interpolated for MR images at higher
resolution. In most cases, the errors in estimated transmit field
and receiver sensitivity are smaller than those of MR images with
high resolution. Thus, the effect of the error of estimated
transmit field on receiver sensitivity is ignorable in practice.
Another major drawback of the BF method is that it requires extra
scan time for estimating transmit field. This drawback can be
solved using the uniform transmit field which is generated by RF
shimming and combined RF pulses (such as the adiabatic pulse and
other RF pulses). In that case, the bias field of the images
acquired with a uniform transmit field can be estimated using
various methods, and the receiver sensitivity is equal to the
estimated bias field. Compared with the bias field method, the
uniform transmit field method reduces the time needed to estimate
the transmit field and the errors resulted from the transmit field.
However, performing RF shimming requires additional time and
combined RF pulses may introduce SAR problems. These drawbacks may
limit the applications of uniform transmit field in estimating
receiver sensitivity at high field strengths.
[0074] FIG. 7 illustrates the distributions of computer simulated
magnitudes of transverse RF field (a), transmit field (b), and
receiver sensitivity (c) of a 10-cm diameter single-loop
transceiver surface coil for a 16-cm diameter spherical phantom.
The transverse RF field is calculated using the square root of the
product of the transmit field B.sub.1.sup.+ and receiver
sensitivity B.sub.1.sup.- (Collins, et al, Magn Reson Med.
47:1026-1028, 2002, incorporated by reference in its entirety). The
relationship between the transmit field and receiver sensitivity is
confirmed by MRI experiment at 7.0 Telsa (Collins et al, 2002; Wang
et al, Magn Reson Med. 48:362-369, 2002, incorporated by reference
in its entirety). It provides the basic principle for applying the
electromagnetic method to estimate receiver sensitivity in this
invention.
[0075] To validate and confirm the definition of transmit field and
receiver sensitivity and their relationship in FIG. 7, the
calculated images using the transmit field and receiver sensitivity
are compared with the experimental results under the same
conditions. A circular surface coil with a diameter of 10 cm was
constructed of copper tape with four capacitive (4.7 pF) as a
transceiver coil (Collins, et al 2000). A spherical phantom with a
radius of 8 cm and containing 20 mM saline solution was used as a
phantom to confirm the relationship between transmit field and
receiver sensitivity of the transceiver coil. The phantom was
modeled with an identical geometry, a relative permittivity of 78
and conductivity of 0.26 S/m. The conductivity of this phantom is
about halfway between those of white matter and fat at 300 MHz. It
was used because it produced a characteristic image intensity
distribution at 7.0 T. The specific image distribution provides a
stringent test of our computer modeling method and serves as an
excellent example for the complexity of the RF field polarization
behavior. The finite difference time domain (FDTD) numerical method
for electromagnetics was used to calculate all electrical and
magnetic fields throughout the model (Collins et al, 2002; Wang et
al, 2002). Our methods for modeling the MR experiment with the FDTD
method and then relating calculated results to the MR experiment
are presented. All FDTD calculations were performed with a
commercially available software package ("xfdtd"; Remcom, State
College, Pa.). The details about calculation and experiments are
shown in our previous publication (Collins et al, 2002; Wang et al,
2002). FIG. 8 shows calculated (a) and experimental (b) axial
gradient echo images at the nominal flip angles of 90.degree.,
respectively. The flip angles for the calculated images were
obtained by adjusting parameter to match the experimental image
intensity distribution with 90.degree. flip angle. Some subtle
features such as the two "dark holes" near the center of the
phantom in the images are exactly reproduced in the calculated
images. The agreement indirectly confirms that the definition of
transmit field and receiver sensitivity of the surface coil is
right, and their relationship
|B.sub.xy|.sup.2=|B.sub.1.sup.transmit||B.sub.1.sup.receive| is
correct. That is, the result confirms the basic principle of
electromagnetic method to estimate the receiver sensitivity.
[0076] Various applications of the method and apparatus of the
invention are possible:
[0077] The first application is for correcting inhomogeneous signal
intensities from non-object characteristics using receiver
sensitivity. MR images after correcting inhomogeneous signal
intensities caused by inhomogeneous RF and inhomogeneous receive
sensitivity can be used to improve accuracy and precision of tissue
segmentation and classification. Accurate segmentation of magnetic
resonance images is very crucial in many biomedical applications:
(a) quantifying functional cortex; (b) performing quantitative
analysis for diagnosis; (c) performing multimodality fusion and
registration; (d) correcting partial volume effects and refine the
quantitative analysis of magnetic resonance spectroscopy (MRS), PET
and other parameters; (e) detecting the lesion as well as lesion
volume.
[0078] The second application is for estimating absolute parameters
using an internal and external reference in MRI and/or magnetic
resonance spectroscopy (MRS).
[0079] Reducing scan time is a very important objective in MRI. In
addition to improved patient comfort, shorter scan times can free
up the imaging system for more patients and reduces image artifacts
caused by patient motion. Parallel imaging techniques have been
widely used in MRI and MRS. Coil sensitivity information is very
important for parallel imaging reconstruction, particularly for
SENSE-based parallel imaging techniques. The third application of
the invention is that the obtained receiver sensitivity map is used
to reconstruct images acquired with parallel imaging techniques The
parallel MR system has a plurality of coils in a coil array
including a reference coil, the coil sensitivity of the reference
coil can be determined using at least one of bias field method,
uniform field method and the electromagnetic field method. For each
element of the coil array, its receiver coil sensitivity map is
determined using a ratio of the reference image and an image
received by the respective coil of the coil array, with the result
of the division being multiplied by the measured sensitivity map of
the reference coil. The obtained receive sensitivity can be used
for reconstruct images acquired from each element using various
algorithms such as SENSE, PARS, SMASH, and GRAPPA, and various
k-space trajectories. Accurate receive sensitivity estimation can
reduce the artifacts of reconstructed images, and improve the image
quality.
[0080] Since the receive sensitivity strongly depend on the
electrical properties a sample being imaged, the estimated receive
sensitivity can be used to determine conductivity and permeability
of a sample, electromagnetic field distribution inside the sample.
As a result, a new biomarker about electrical properties and
electromagnetic field can be provides for functional MRI, diagnosis
of diseases, electromagnetic therapy, and human safety in
electromagnetic environment. The fourth application is to use for a
new bio marker.
[0081] The fifth application is to use parallel transmit techniques
and/or dynamic shimming of RF field to eliminate inhomogeneous
signal intensities caused by receiver coil.
[0082] In various embodiments, the disclosed methods may be
implemented on one or more computer systems. Each computer system
may be in wired or wireless communication with an MR system and/or
with one another through a combination of local and global networks
including the Internet. Each computer system may include one or
more input device, output device, storage medium, and processor.
Possible input devices include a keyboard, a computer mouse, a
touch pad, a touch screen, a digital tablet, a microphone, a track
ball, and the like. Output devices include a cathode-ray tube (CRT)
computer monitor, a liquid-crystal display (LCD) or LED computer
monitor, touch screen, speaker, and the like. Storage media include
various types of local or remote memory devices such as a hard
disk, RAM, flash memory, and other magnetic, optical, physical, or
electronic memory devices. The processor may be any typical
computer processor for performing calculations and directing other
functions for performing input, output, calculation, and display of
data in accordance with the disclosed methods. In various
embodiments, implementation of the disclosed methods includes
generating sets of instructions and data (e.g. including image data
and numerical data) that are stored on one or more of the storage
media and operated on by a controller.
[0083] In some embodiments, implementation of the disclosed methods
may include generating one or more web pages for facilitating
input, output, control, analysis, and other functions. In other
embodiments, the methods may be implemented as a locally-controlled
program on a local computer system which may or may not be
accessible to other computer systems. In still other embodiments,
implementation of the methods may include generating and/or
operating modules which provide access to portable devices such as
laptops, tablet computers, digitizers, digital tablets, smart
phones, and other devices.
[0084] Thus, the invention provides, among other things, a method
for estimating receiver sensitivity in a magnetic resonance (MR)
system. Various features and advantages of the invention are set
forth in the following claims.
* * * * *