U.S. patent application number 14/805285 was filed with the patent office on 2015-11-12 for non-linear projections of 3-d medical imaging data.
This patent application is currently assigned to CARL ZEISS MEDITEC, INC.. The applicant listed for this patent is Carl Zeiss Meditec, Inc.. Invention is credited to Paul F. STETSON.
Application Number | 20150320304 14/805285 |
Document ID | / |
Family ID | 42735407 |
Filed Date | 2015-11-12 |
United States Patent
Application |
20150320304 |
Kind Code |
A1 |
STETSON; Paul F. |
November 12, 2015 |
NON-LINEAR PROJECTIONS OF 3-D MEDICAL IMAGING DATA
Abstract
The present invention improves projection displays of volume
data. Using the Minimum Intensity Projection (MinIP), fluid filled
regions or other regions of hyporeflective tissue are displayed. By
limiting the projection to partial volumes within the volume,
differences in the scattering intensity within specific regions are
isolated. In this way, hyperreflectivity of weakly scattering
tissue can be assessed.
Inventors: |
STETSON; Paul F.; (Piedmont,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Carl Zeiss Meditec, Inc. |
Dublin |
CA |
US |
|
|
Assignee: |
CARL ZEISS MEDITEC, INC.
Dublin
CA
|
Family ID: |
42735407 |
Appl. No.: |
14/805285 |
Filed: |
July 21, 2015 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13692781 |
Dec 3, 2012 |
9105123 |
|
|
14805285 |
|
|
|
|
12535046 |
Aug 4, 2009 |
8332016 |
|
|
13692781 |
|
|
|
|
Current U.S.
Class: |
351/206 ;
351/246 |
Current CPC
Class: |
G06T 2207/10101
20130101; G06T 11/003 20130101; A61B 3/102 20130101; A61B 3/14
20130101; G06T 15/08 20130101; A61B 3/0025 20130101; G06T 7/0012
20130101; A61B 2560/0475 20130101; A61B 3/0041 20130101; G06T
2207/30041 20130101; G06T 7/12 20170101 |
International
Class: |
A61B 3/00 20060101
A61B003/00; A61B 3/10 20060101 A61B003/10 |
Claims
1. A method of generating an image of an anatomical region within
an eye, said eye being examined by an optical coherence tomography
(OCT) system, said OCT system including a light source for
generating a beam of light, a sample arm, a reference arm, and a
detector for measuring light combined from the sample and reference
arms, said method comprising the steps of: a. scanning the beam of
light over a region of the eye via the sample arm; b. combining the
light from the sample and reference arms; c. measuring the combined
light; d. generating three dimensional volume image data of the eye
from the combined light; e. identifying a first surface within the
image data; f. identifying a subvolume of the image data, said
subvolume having less than all the generated image data, said
subvolume being identified based upon the identification of the
first surface, said subvolume being either bounded by or including
the first surface, said subvolume including the anatomical region;
g. selecting a set of image points within the subvolume, said image
points being selected by evaluating a plurality of ray projections
extending through the subvolume and identifying one image point in
each projection, wherein each of the identified image points has a
common intensity attribute; h. determining a smooth reference
surface using the locations of the set of identified image points;
i. generating an image based on the smooth reference surface; and
j. storing or displaying the image.
2. A method as recited in claim 1, wherein the smooth reference
surface is determined by fitting.
3. A method as recited in claim 1, further comprising using the
intensity of image pixels within the volume to determine
intensities at the locations on the smooth surface.
4. A method as recited in claim 1, wherein the image is generated
by integrating a certain margin above and below the smooth
reference surface.
5. A method of generating an image of an anatomical region within
an eye, said eye being examined by an optical coherence tomography
(OCT) system, said OCT system including a light source for
generating a beam of light, a sample arm, a reference arm, and a
detector for measuring light combined from the sample and reference
arms, said method comprising the steps of: a. scanning the beam of
light over a region of the eye via the sample arm; b. combining the
light from the sample and reference arms; c. measuring the combined
light; d. generating three dimensional volume image data of the eye
from the combined light; e. identifying a first surface within the
image data; f. identifying a subvolume of the image data, said
subvolume having less than all the generated image data, said
subvolume being identified based upon the identification of the
first surface, said subvolume being either bounded by or including
the first surface, said subvolume including the anatomical region;
g. selecting a set of image points within the subvolume, said image
points being selected by evaluating a plurality of ray projections
extending through the subvolume and identifying one image point in
each projection, wherein each of the identified image points has a
common intensity attribute; h. generating an image based on
information associated with the selected set of image points; and
i. displaying the image.
6. A method as recited in claim 5, wherein the subvolume contains
the outer nuclear layer (ONL) of the eye, wherein the common
intensity attribute corresponds to the minimum intensity of the
associated ray projections, and further comprising identifying
regions of retinal disruptions based on the minimum intensity.
7. A method as recited in claim 6, wherein the minimum intensity is
used to identify areas at risk of future pathology.
8. A method as recited in claim 5, wherein the image generation is
based on location information associated with each of the selected
set of image points within the associated ray projection and
wherein the generated image is a height map of the maximum
intensity locations relative to a reference surface.
9. A method as recited in claim 5, wherein the generated image is a
height map of the common intensity attribute locations and the
height is encoded by color or brightness.
10. A method as recited in claim 5, wherein one or more of the ray
projections are along axes different from the propagation axis of
the beam.
Description
PRIORITY CLAIM
[0001] This is a continuation of U.S. patent application Ser. No.
13/692,781, filed Dec. 3, 2012 which is in turn a continuation of
U.S. patent application Ser. No. 12/535,046, filed Aug. 4, 2009 the
disclosures of which are incorporated herein by reference.
TECHNICAL FIELD
[0002] The present invention relates to non-linear projections of
3-D medical imaging data, and in particular to optical medical
imaging such as Optical Coherence Tomography ("OCT") for use in
diagnosis and monitoring of tissue health.
BACKGROUND
[0003] Volume imaging data has become prevalent in medical imaging
modalities such as ultrasound imaging, magnetic resonance imaging,
computed tomography, and Optical Coherence Tomography (OCT). Volume
data has brought innovation to the medical field, making it
possible to safely and noninvasively directly observe the internal
structure of a body, in particular of a human body. In recent
years, volume rendering image processing techniques have been used
to reduce the volume data to images displayed for medical
diagnosis. Volume rendering enables visualization of
three-dimensional structures by displaying a 2D projection of the
3D structure on commercially available 2D display.
[0004] Optical Coherence Tomography (OCT) is a technology for
performing high-resolution real time optical imaging in situ. OCT
is an optical measurement and imaging technique using either
low-coherent light from a broadband source or sweeping light from a
tunable laser to create interference signals measuring backscatter
intensity at depths along a sample path. This depth profile is
commonly called an "A-scan". Laterally scanning the sample beam
over a series of adjacent A-scans synthesizes cross-sectional
images, creating a 2-D tomogram commonly called a B-scan.
Typically, volumes are acquired by laterally scanning the sample
beam over a series of B-scans; however alternative scan patterns,
such as a spiral scan pattern, have been suggested to acquire
volume data.
[0005] Evaluation of biological materials using OCT was first
disclosed in the early 1990's (see U.S. Pat. No. 5,321,501,
Swanson, et al.). More recently it has been demonstrated that
frequency domain OCT (FD-OCT) has significant advantages in speed
and signal to noise ratio compared to time domain OCT (Leitgeb, R.
A., et al., (2003) Optics Express 11:889-894). In Spectral Domain
OCT (SD-OCT), sometimes referred to as Frequency Domain OCT
(FD-OCT) and sometimes also referred to as Spectral Radar (Hausler,
G and Linder, M W, Journal of Biomedical Optics Vol. 3 No. 1 (1998)
21-31), the measurement is achieved by examining the spectral
content of the interference pattern out of the interferometer.
[0006] Volume data is reduced from 3-D to 2-D for display on a
monitor or when printed on paper. In U.S. Pat. No. 7,301,644,
Knighton et al. disclosed a method for reducing 3-D data for 2-D
display called en-face imaging. This technique takes the volume
data (or an interesting volume subset of the 3-D dataset) and
integrates the data along each A-line, creating a 2-D image from
the 3-D dataset. This technique has proven to be a very useful tool
for viewing the OCT volume data. In light of the above, there is a
need in the art for additional methods for viewing 2-D images of
3-D volume OCT data. The present invention provides such additional
methods for creating 2-D images from 3-D volume data sets to
display medically relevant information.
SUMMARY
[0007] The present invention is defined by the claims and nothing
in this section should be taken as a limitation on those claims.
Advantageously, embodiments of the present invention provide
additional means for obtaining and displaying medically relevant
information extracted from volume data.
[0008] In accordance with one aspect of the present invention, a
Minimum Intensity Projection (MinIP) is used to find a fluid filled
region or region of retinal disruption within a volume of data.
[0009] In accordance with another aspect of the present invention,
volume information has low-intensity speckle artifacts reduced
prior to projection through a partial volume.
[0010] In accordance with yet another aspect of the present
invention, the projection information is compensated for variations
in illumination prior to display.
[0011] In accordance with yet another aspect of the present
invention, the MinIP is found along a collection of A-lines and the
coordinate locations of the minimal intensity points are stored.
These locations are smoothed before the smoothed locations are
displayed as a minimum intensity height map.
[0012] In accordance with yet another aspect of the present
invention, the partial volume over which the MinIP is computed is
determined at least partially as a function of at least one special
location. The special location may be user defined, or
automatically computed from the volume image data.
BRIEF DESCRIPTION OF THE DRAWINGS
[0013] FIG. 1 is a schematic illustration of a MinIP display using
the partial volume between the ILM and the RPE.
[0014] FIG. 2 is a schematic illustration of a MinIP display using
the partial volume between the OPL and the RPE.
[0015] FIG. 3 is a flowchart of an embodied MinIP display
application.
[0016] FIG. 4 is a flowchart of an embodied display application in
which the location of the minimum intensity is used to determine a
surface from which the displayed intensity is determined.
[0017] FIG. 5 is an image showing hyperreflective tissue.
[0018] FIG. 6 is an image showing both hyperreflective and
hyporeflective tissue.
[0019] FIG. 7 illustrates one design of an optical coherence
tomography system.
DETAILED DESCRIPTION
[0020] The embodiments expressed herein are examples and the
descriptions presented have been chosen to illustrate the
principles of the invention and its practical applications and not
as a definition of the invention. Modifications and variations of
the invention will be apparent to those skilled in the art. The
scope of the invention is defined by the claims, which includes
known equivalents and unforeseeable equivalents at the time of
filing of this application.
[0021] Herein, an intensity projection of an image volume is a two
dimensional intensity image derived from the image volume by
casting projection rays through the volume and determining a single
intensity to represent each ray. A maximum intensity projection
(MIP) is created by first identifying the pixel in each ray that
has the highest intensity. An image is then created by mapping the
locations of the rays to locations in the image and setting the
intensity of the pixel at each location to the highest intensity
associated with that ray. Similarly, a minimum intensity projection
(MinIP) is the intensity projection with the result chosen to be
the minimum intensity along the ray. In general MIP and MinIP can
be formed casting projection rays along the x-, y-, or z-axis, or
in any direction whatsoever. In MIP, each selected point has the
attribute that its intensity is greater than or equal to the
intensity of any other point along the projection ray. Similarly,
MinIP selects points according to the minimum attribute. Many other
attributes may be ascribed to some point along the projection
array. For example, the location nearest the upper surface of the
volume through which the projection ray is cast with intensity
equal to the median intensity along the projection ray provides
another attribute by which a single image point may be selected for
each projection ray. The maximum, minimum, and median are nonlinear
functionals--they do not satisfy the additivity property of linear
functionals. In contrast, summation and integration are linear
functionals. For simplicity of exposition, MinIP will be the
primary example used throughout this description and rays will
normally be cast parallel to the z-axis.
[0022] In an imaging modality where a fluid filled region is less
reflective than the surrounding tissue, the minimum intensity
projection (MinIP) offers a unique display technique. The MinIP
provides a 2-D projection of the boundary of the fluid filled
region. In the early 1990s, minimum intensity projections were used
in Nuclear Magnetic Resonance Imaging to find fluid filled regions
(U.S. Pat. No. 5,189,369, Takane et al.). Ophthalmologic images
present a unique challenge, since the eye itself is fluid filled
and OCT images contain speckle, which is an interference pattern in
the backscattered light that causes intensity variations in the
image. Speckle can create a very low intensity reflection in the
midst of highly reflective tissue. By the late 90's, systems for
ray-casting projections through speckle reduced ultrasound volumes
were disclosed (U.S. Pat. No. 5,779,641, Hatfield, et al.) Speckle
can be reduced by spatial or frequency compounding, low pass
filtering image data, or other means. These systems, however, fail
to recognize the importance of casting the projection through a
partial volume (i.e., through a limited volume region.)
[0023] In OCT, the preferred method for visualizing fluid has been
to measure the overall macular thickness. This method may delineate
pockets of fluid, but it does so imprecisely. The invention
disclosed herein, using MinIP, is sensitive to extremely small
pockets of fluid. In prior art efforts, automatic segmentation of
fluid pockets was difficult to perform reliably, and manual
intervention to correct errors in automatic segmentation is time
consuming and operator dependent. While the embodiments enclosed
herein do not provide volume information, they enable the
visualization of fluid filled regions, which is most important for
determining drug treatment. For laser treatment, knowing the
lateral location and extent of fluid lesions is critical for
precisely targeting treatment where it is needed.
[0024] The Minimum Intensity Projection (MinIP) is a projection
algorithm that finds the minimum brightness along the projection
path in a volume of data. MinIP solves the problem of viewing the
volume image when looking for weak reflections or deep shadows. For
imaging systems that image backscattered radiation, i.e., imaging
systems like OCT, hyporeflective regions surrounded by more
reflective regions are readily detectable. Not all anatomy of
interest is strongly reflecting. Often anatomy of interest, such as
hyporeflective pockets of fluid within the retina are weakly
reflecting. Such anatomy may be best visualized using minimum
intensity projections.
[0025] The Outer Nuclear Layer (ONL) is normally not as reflective
as other retinal tissue. Tissue that backscatters more light than
normal is hyperreflective. In converse, tissue that backscatters
less light than normal is hyporeflective. Even in the presence of
diseased or disrupted retina which may cause hyperreflectivity in
the ONL, the ONL often remains less reflective than neighboring
retinal tissue. Thus, even a hyperreflective ONL may be the least
reflective tissue found along a projection line and thus the
hyperreflective ONL will still appear in the MinIP, but it will be
brighter than normal. Hyperreflectivity in the Outer Nuclear Layer
and other normally dark regions of the retina also can appear over
and around areas of retinal disruption.
[0026] As with most modern image processing procedures, the
preferred MinIP implementation is by means of an electronic
computational processing unit or CPU. The CPU might be a general
purpose computer like a PC or workstation or a specialized CPU such
as a digital signal processor (DSP), application specific
integrated circuit (ASIC), field programmable gate array (FPGA) or
any other integrated circuit, whether fully contained in a single
package or distributed across multiple chips.
[0027] In one embodiment of the invention, an OCT image volume and
two segmented surfaces of the retina (e.g., ILM and RPE, or ILM and
IS/OS) are required. The two segmented surfaces bound the partial
volume through which the minimum projection is cast. The choice of
surfaces may significantly impact the resulting MinIP display. The
data depicted in FIG. 1 illustrates an example where the MinIP 160
is generated from the partial volume between the inner limiting
membrane (ILM) 110 and the retinal pigment epithelium (RPE) 150.
The illustrated strip 160 representing the MinIP has been widened
for illustrative purposes. The projection cast along a ray is a
single point or pixel. The wider band depicted here makes the MinIP
intensity values easier to see.
[0028] For the example data depicted in FIG. 1, the MinIP 160 is
the same whether it is generated from the partial volume between
the ILM 110 and the RPE 160 or it is generated from the partial
volume between the ILM and the boundary between the inner segment
and outer segments of photoreceptors (IS/OS) 140. This is because
the intensity depicted for the IS/OS is greater than the
intensities depicted for neighboring tissue on either side of the
IS/OS. Hence, the IS/OS intensity will not be the selected minimum.
Furthermore, the depicted data has the intensity of the scattering
tissue between the IS/OS and the RPE that is no less than the
tissue above the IS/OS. Hence, the data from the IS/OS to the RPE
in this example, do not change the intensity of the minimum
intensity along the projection ray. This data was chosen for
illustrative purposes and in no way implies that backscatter
returned from below the IS/OS cannot have intensities less than the
intensities derived from backscatter above the IS/OS. It can, but
in this example, does not.
[0029] FIG. 2 illustrates how the MinIP can change by changing the
partial volume through which the projection ray is cast. FIG. 2
illustrates the same theoretical eye as FIG. 1. That is, the data
illustrated in FIGS. 1 and 2 are the same. However, the MinIP 260
is generated from the partial volume between the outer plexiform
layer (OPL) 230 and the RPE 250 whereas the MinIP 160 in FIG. 1 was
generated from the partial volume between the ILM 110 and the RPE
150. The fluid filled region 235 is within the outer nuclear layer
(ONL) 240. Because the MinIP in this example was cast between the
OPL and the RPE, the fluid filled region 235 is readily apparent in
the MinIP strip 260. However, the hyporeflective regions 225, which
appeared in MinIP strip 160 are not visible in the MinIP strip 260
because they are not within the partial volume used to generate the
display.
[0030] In an alternate embodiment of the invention, an OCT image
volume and one segmentation of the retina (e.g., ILM, RPE, ILM, or
IS/OS) is required. In this case, the segmented surface and a
predefined or computed parameter or surface can define the partial
volume through which the minimum projection is performed. The
volume could be all points within a given distance of the surface;
or it could be all points within a fixed depth below (or above) the
surface, or all points within a range from a fixed offset to the
surface, or other volume defined with respect to the surface.
Alternatively, a surface can be defined by fitting a smooth surface
to the segmented surface. The volume can be defined as the space
between the surfaces or the space between a surface offset a fixed
distance from the segmented surface to a surface offset a fixed
distance from the computed surface, or other volume defined with
respect to the two surfaces. In yet another embodiment of the
invention, an image volume and an a priori defined subset of the
volume are required. In this case, the partial volume through which
the minimum projection is performed is chosen to be the a priori
defined subset of the complete image volume.
[0031] FIG. 3 is a flowchart of an embodied MinIP display
application. A volume data 310 is input into the application. A
partial volume is selected or chosen 320, either by one of the
methods described above or by some other means for choosing a
subvolume of the actual volume. The speckle is reduced in the
partial volume 330, in one embodiment by a smoothing filter that is
preferably larger than a speckle cell and smaller than the body to
be imaged, or in other embodiments, by other image processing
means. Rays are cast 340 through the partial volume and the
location and intensity of the minimal intensity along each ray are
determined 350. An image is formed from the minimal intensities and
that image is displayed 360. The displayed image can be comprised
of the minimum intensity values, or normalized minimum intensity
values that are compensated for variations in illumination. The
intensities within the ray are affected by what the light has
passed through before entering the partial volume. Hence, when rays
are cast in the direction of the illumination light (along the
z-axis), a local compensation may be applied to the projection ray
before computing the minimum, or, in particular if the compensation
is linear, simply applied to the minimum selected from the ray. For
example, the displayed values can be minimum intensity values
normalized by the mean intensity along the projection ray within
the partial volume. This normalization factor may be computed from
points within the partial volume that are within a local
neighborhood of the projection ray, or even points within the
entire volume that are within a local neighborhood of the
projection ray. Additionally, the display pixels can be smoothed to
present a more appealing image. This is perhaps the simplest
embodiment of the invention disclosed herein.
[0032] Minimum intensity projections determined according by the
process of the preceding paragraph can have an "unnatural" look, in
part because the depths from which the intensity displayed in the
image is taken can be highly irregular, jumping greatly in depth
even over small changes in X-Y. While determining the minimum
intensity, we can save and then use the location within the image
volume from which the minimum intensity was attained. We denote the
depth at which the minimal intensity is attained by z.sub.min(x,y).
That is, z.sub.min(x,y) is the depth of the location of the minimal
intensity to be displayed at the (x,y) pixel of the MinIP image. We
can improve the visual impression of the displayed image and get a
more natural looking en-face image by fitting a smooth surface to
the collection of points (x, y, z.sub.min(x,y)) and then
integrating the volume image intensities a certain margin above and
below a smoothed surface. A two-dimensional quadratic or quartic
surface generally gives a sufficiently smooth contour, although in
some cases the fitting may not follow the desired anatomy because
it is too smooth or the areas of abnormal anatomy were not properly
excluded from the fitting. Technically, the en-face image generated
by integrating in the neighborhood of a smoothly fit surface is no
longer a minimal intensity projection display image. It does,
however, require finding the minimal intensity projection voxels in
order to compute the displayed image. In the limit where the
en-face is only over a single pixel, the intensity values on the
smooth surface are displayed.
[0033] Alternatively, the visual impression of the displayed image
will appear more natural if the locations of the depths
z.sub.min(x,y) are constrained to be within a local neighborhood.
For example, the smoothed surface fitted to the collection of
points (x, y, z.sub.min(x,y)) can be used to define a new partial
volume and the MinIP can be recomputed over this volume.
Alternatively, it can be accomplished by first choosing a
collection of MinIP points as good and recomputing the remaining
points by constraining the search for the minimum intensity point
at a neighboring location further within the partial volume so that
neighboring MinIP is chosen from a set of points within a fixed
distance of the location of the chosen neighbor.
[0034] In yet another alternative, an intensity other than the
minimum intensity along the projection ray can be chosen. Grayscale
image data intensities are distributed according to a Gaussian-like
distribution in that the number of voxels with intensities near the
mean is far greater than the number of voxels with intensities in
the tails of the distribution. The minimum values of the image data
along the rays in some regions will fluctuate statistically from
ray to ray since the minima occur out in the sparsely populated
tail of the intensity distribution. Selecting an intensity level
slightly higher than the minimum, such as the 5th-percentile
intensity, selects a portion of the intensity distribution that is
more highly populated and thus less susceptible to statistical
fluctuation, yet it preserves most of the characteristics of the
minimum. Thus choosing the 5th percentile intensity value along the
projection ray produces a smoother, less noisy image than the
minimum intensity projection, although it may fail to depict fluid
in locations where the total height of fluid-filled regions is less
than 5% of the retinal thickness.
[0035] FIG. 4 depicts an alternative embodiment of the disclosed
invention. The volume data is acquired 410 and a partial volume is
selected 420. The speckle in the data is reduced 430 before rays
are cast 440 and the location and intensity of the minimum
intensity along the cast ray is determined 450. In this embodiment,
the display intensities are indirectly determined from minimum
values along the cast rays, so only those rays that are needed to
implement the method are cast 460. In most cases, the more rays
cast up to the resolution of the image volume, the better and more
accurate the final display will be. The locations where the minimum
intensities along the cast rays were found are points within the
partial volume. A surface is determined from these locations 470.
This surface may be the collection of location points themselves,
or it may be a surface fit to those locations, or any surface
otherwise determined from the location of the minimum intensity
points. A smooth surface is determined 480 either directly from the
points so that the smoothed surface of 480 is the same as the
surface of 470, or the smooth surface is determined from the
surface of 470. This smooth surface should be within the volume.
The intensity of image pixels within the volume are used to
determine intensities at locations on the smooth surface 490 and
the intensities on the smooth are used to create an image for
display 495. As in the previous example, the displayed image can be
comprised of the computed intensity values, or normalized computed
intensity values that are compensated for variations in
illumination. Additionally, the display pixels can be smoothed to
present a more appealing image.
[0036] Alternatively, the locations of the minimum intensity pixels
can be used to create a minimum intensity height map. That is, a
map of the height of the minimum intensity locations can be
determined. When the projection ray is cast parallel to the Z-axis,
the height is just the location along the ray (or a constant offset
from that location, depending on the origin from which the height
is calculated). If the projection ray is not parallel to the
Z-axis, but is rather at an angle .theta. to the Z-axis, then the
actual height is proportional to the location of the minimal
intensity along the ray for the appropriate choice of origin. The
proportionality is constant for all rays. The minimum intensity
height map can be displayed as a surface rendered from a 3-D volume
or as a 2-D image, with the height encoded by color or brightness.
This height may be usefully displayed relative to the location of
the RPE or some other segmented or fitted surface from the image
volume.
[0037] In light of FIG. 4, instead of the locations of minimum
intensity along the projection ray depicted in 470, we can choose
the location along the projection ray where the intensity attains,
for example, the 5.sup.th percentile value. We can take the
locations where these values were attained and use them (instead of
the minimum locations) to determine the surface of 470. Of course,
statistically as we move closer to the 50th percentile, the
likelihood of finding more than one location with the same
intensity increases. In this case, when implementing methods
following the style of FIG. 4, a choice must be made, choosing one
of the locations. This choice can simply be the first such point
found, with the obvious bias attached, or we can choose locations
based on the nearness of local clustering. Other means of choosing
one location from the collection of locations will be obvious to
those versed in the art.
[0038] Many forms of acquired image data contain an image artifact
called speckle. Image speckle is a phenomenon generally associated
with wave imaging and occurs as a result of wave interference at
the detector. OCT and ultrasound image data (either 2-D or volume)
generally contain speckle. The interference of various waveforms
causes fluctuations of the detected intensity within the image
volume. Volume image data with speckle should have at least a mild
speckle reduction process performed to reduce the null intensities
created in the neighborhood of high intensity reflections before
performing a minimum intensity projection. Mild speckle reduction
is preferred because it is generally sufficient to remove deep
nulls in the neighborhood of strong reflections and stronger
speckle reduction techniques generally require more data, take
longer, and/or smear the data. Generally averaging across twice the
speckle diameter is sufficient. After speckle reduction, each axial
scan is analyzed for the minimum intensity, i.e., the darkest gray
level, between the segmentation boundaries or through the partial
volume.
[0039] Alternatively, a projection through the partial volume in a
direction different from the axial scan direction can be performed.
This generally requires interpolation and incurs inefficiencies in
computational speed. Hence, projection along the axial direction is
preferred.
[0040] As discussed previously, one approach to removing the
"unnatural" look of the MinIP display is to integrate an en-face
about a smooth surface derived from the collection of locations
where the minimum intensities were attained. Another alternative
for removing this "unnatural" look of the display is to globally
define the minimum. That is, we can define a metric over a
collection of intensity projections and choose the intensity
projection of minimal intensity to display.
[0041] For example, if z.sub.ILM(x,y) represents the ILM
segmentation and z.sub.RPE(x,y) represents the RPE segmentation,
then a family of surfaces between the ILM and RPE is represented
by:
z.sub.mix(f,x,y)=f*z.sub.ILM(x,y)+(1-f)*z.sub.RPE(x,y)
[0042] where f is a desired fraction between 0 and 1, chosen to
produce a dark image generally falling in the ONL. For example, a
set of images for f=0.1, 0.2, . . . , 0.9 can be generated and the
image with the minimum median intensity can be selected.
[0043] Any family of surface projections of the volume can be
chosen whereby a metric can be applied by which a minimum surface
can be chosen. The metric can be the median intensity of a
projection or the average intensity of a surface or the absolute
minimum intensity of a surface or any other metric. The metrics of
the family of projections are compared and that projection with the
minimum metric is chosen for display. If more than one surface has
the same minimal value as its metric, a system for choosing one of
these must be chosen. We can choose the first one with the minimum
value or the last one with the minimum value. Alternatively, the
choices can be displayed and one can be chosen by an operator.
Alternatively, when the family is parameterized, the surface with
parameter closest to a fixed value may be chosen.
[0044] Whether we smooth the collection of locations where the
minimal intensity is found and integrate an en-face about that
smoothed surface, or we fit a smooth surface to the collection of
locations where the minimal intensity is found and integrate an
en-face about that smooth surface, or we choose a collection of
smooth surfaces and find the minimum amongst them, the projection
provides an image that is like a MinIP but has less discontinuity
from changing depths of the displayed data. They will be different
from the MinIP, though, in that they may not depict small pockets
of fluid or other small dark areas because of the constraints on
the smoothness of the imaged surface.
[0045] The Outer Nuclear Layer (ONL) 240 has had little attention
paid to it in OCT research, in part because it is difficult to
segment. The minimum intensity in the healthy macula usually lies
within this layer. The ONL is significantly more reflective where
retinal disruptions occur, e.g., in the vicinity of retinal fluid
pockets or over sub-RPE fluid. Longitudinal clinical studies may
show whether this hyperreflectivity predicts future locations of
pathology. The axial location of the minimum intensity relative to
the RPE may also indicate focal disruptions in the retina, with
increased height indicating hyperreflectivity in the outer
retina.
[0046] Some of the interesting information in Minimum Intensity
Projections (MinIPs) comes from looking at the hyperreflectivity of
the ONL that can be caused by disruption of the retina. For this
type of imaging, the partial volume through which the minimum
intensity is projected must contain the ONL. The minimum intensity
lies in the ONL about 99% of the time in healthy areas of the
retina. If we do not have a good reliable ONL segmentation, MinIPs
are an alternative way to examine the hyperreflectivity of the ONL
(although bright regions in a MinIP can also mean the ONL has
atrophied away completely). To ensure that you are looking at the
ONL, it is best to limit the partial volume to the region between
the OPL surface and the IS/OS surface. Since the IS/OS may
sometimes be absent or invisible in the OCT images, limiting the
partial volume between the OPL and RPE is more reliable. Even the
OPL is difficult to segment and the results remain interesting when
the partial volume is taken between the ILM and the RPE.
Alternatively, the partial volume can be taken to be a volume above
the RPE to a fixed distance above the RPE or the volume above the
RPE to a surface that is a percentage of the distance from the RPE
to the ILM. These partial volumes can generally be used to
incorporate a large portion if not all of the ONL. Though the
partial volume may contain other tissue, the surrounding tissue is
generally brighter and thus will only appear in the MinIP when the
ONL is severely disrupted and is even more hyperreflective than the
MinIP indicates at that point.
[0047] FIG. 5 is a MinIP image using a partial volume taken between
the ILM and the RPE and containing hyperreflective tissue in the
ONL. An area of hyperreflectivity is found within the circle 510.
This region is determined to be hyperreflective because the average
intensity within the region 510 is significantly higher than the
average intensity of a normal region 520. Hyperreflectivity can be
determined using statistics other than the mean, such as the
median. Any statistic which provides a statistically meaningful
distinction between normally reflective tissue and abnormally
reflective tissue will suffice, though the quality of the results
may vary.
[0048] FIG. 6 is another MinIP image, again using a partial volume
taken between the ILM and the RPE and containing hyperreflective
and hyporeflective tissue in the ONL. An area of hyperreflectivity
is found within the circle 610. This region is determined to be
hyperreflective because the average intensity within the region 610
is significantly higher than the average intensity of a normal
region 630. An area of hyporeflectivity is found within the circle
620. This region is determined to be hyporeflective because the
average intensity within the region 620 is significantly lower than
the average intensity of a normal region 630. Any statistic which
provides a statistically meaningful distinction between normally
reflective tissue and abnormally reflective tissue will suffice,
though the quality of the results may vary.
[0049] FIG. 7 illustrates an OCT device which can be used to
implement the subject invention. Further information about this
type of OCT device is disclosed in U.S. Patent Publication No.
2007/0291277, incorporated herein by reference. A low coherence
light source 700, typically a superluminescent diode (SLD), is
coupled to source fiber 705 that routes light to directional
coupler 710. The optimal directional strength of the coupling
depends on system design choices and may be 90/10 (as shown in FIG.
7) or 70/30 or other choices depending on SLD back reflection
tolerance, the source illumination required to image the sample and
other system design parameters. Directional coupler 710 splits the
light into sample fiber 715 and reference fiber 735. The sample
path may include a delay apparatus (not shown) to adjust the length
of the sample path. The transverse scanner 720 deflects the OCT
beam and preferably creates a focus in the beam near the region of
interest in sample 730. The transverse scanner laterally scans the
optical beam across the sample in order to image a volume of the
sample.
[0050] Some light scattered from sample 730 returns through the
scanner and delay apparatus to sample fiber 715. Coupler 710 routes
this light through loop 760 to fiber coupler 750, where it
interferes with the reference light. The combining coupler 750
provides two outputs. These outputs can be used for balanced
detection (see U.S. Pat. No. 5,321,501, FIG. 10). Alternatively,
the coupling ratio of coupler 750 can be adjusted to send most of
the interfered light to a single OCT detector 770. Each OCT
detector can be a single photodetector for use in time-domain OCT
or swept-source OCT, or a spectrometer for use in spectral domain
OCT.
[0051] Optional tap 740 diverts a fraction of the reference light
to detector 745, which may be used to monitor the source power.
Monitoring may be included to monitor the safety of the sample or
to detect a degradation in the source 700. Alternatively,
monitoring may not be included at all in the system. The tap
removes some fraction of optical power from the reference fiber
735, reducing the power that reaches coupler 750. Sensitivity in
OCT can reach the shot-noise limit if the reference power is large
enough to bring the interference signal above receiver noise, but
not so large as to bring intensity noise or beat noise above the
level of shot noise.
[0052] The coupling ratios in directional couplers 710, 740 and 750
are chosen to set a safe level of illumination to the sample, and
to set the appropriate reference power at the detector or
detectors. For example, in the case of ophthalmic OCT of the retina
using light with wavelengths near 850 nm, the safe exposure level
is approximately 0.5 mW, and the optimum reference level at the
detector is approximately 0.005 mW. Sources are available in this
wavelength range having output power of approximately 5 mW. For
these conditions one would use a coupling ratio near 90%/10% in the
splitting coupler 710 so that 10% of the source power reaches the
sample. 90% of the backscattered light will then be routed to loop
760. In the case where there is a single OCT detector 770, the
combining coupler 750 preferably routes most of the sample light to
that detector. The splitting coupler 710 routes 90% of source
light, 4.5 mW, to reference fiber 735, while only 0.005 mW is
required at the detector. One could use a combining coupler 750
that couples 0.1% of the reference light into the single OCT
detector 770, but in manufacture it is difficult to control the
0.1% coupling factor. A preferred solution is to use a 99%/1% split
ratio in combining coupler 750, and take advantage of the
additional degree of freedom in tap 740 to adjust the reference
power. Nominally, tapping 89% of the power form reference fiber 735
will provide an appropriate reference level of 0.005 mW at OCT
detector 770, in this example.
[0053] As an alternative to adjusting the tap ratio of optional tap
740, one can adjust the reference level by including attenuating
fiber (U.S. Pat. No. 5,633,974) in the reference path.
[0054] The output of the detector 770 is routed to processor 780.
This processor may be a single device or a plurality of devices,
preferentially optimized for their portion of processing. The
processor 780 is connected to one or more peripherals providing a
user interface devices, such as display 790. The processor might
also be connected to other user interface devices (such as a
keyboard, mouse, joystick, and others), and one or more external
communication devices (such as a USB or network connector, optical
storage media, printer, internet, and others), as well as possibly
connecting to other imaging hardware (such as cameras, fixation
targets, fundus viewers, and others) or peripheral patient devices
(such as head support, height adjustment, and others) which are not
shown. The processor 780 provides the computational power (in one
or more modules) processing functions such as image formation,
volume rendering, segmentation, registration, evaluation of cost
functions, and/or other computational tasks required for medical
imaging and analysis. Processed results may be displayed on display
790 or stored locally on a local storage device (not shown) or
stored or displayed externally using one or more of the processor's
external communication devices.
[0055] It should be understood that the embodiments, examples and
descriptions have been chosen and described in order to illustrate
the principles of the invention and its practical applications and
is not intended to be exhaustive or to limit the invention to the
precise form disclosed. Modifications and variations of the
invention will be apparent to those skilled in the art in light of
the above teaching. For example, while extrema are described herein
as minimums, under an alternative metric, the invention can be
performed using maximums. The embodiments were chosen and described
to explain the principles of the invention and its practical
application to enable others skilled in the art to best use the
invention in various embodiments and with various modifications
suited to the particular use contemplated. The scope of the
invention is defined by the claims, which include known equivalents
and unforeseeable equivalents at the time of filing of this
application.
[0056] The following references are hereby incorporated herein by
reference.
U.S. Patent Documents
[0057] U.S. Pat. No. 5,189,369 Takane et al., NMR imaging method of
low flow rate fluid [0058] U.S. Pat. No. 5,321,501 Swanson, et al.,
Method and apparatus for optical imaging with means for controlling
the longitudinal range of the sample [0059] U.S. Pat. No. 5,779,641
Hatfield, et al., Method and apparatus for three-dimensional
ultrasound imaging by projecting filtered pixel data [0060] U.S.
Pat. No. 6,436,049 Kamiyama et al., Three-dimensional ultrasound
diagnosis based on contrast echo technique [0061] U.S. Pat. No.
6,501,272 Haacke, et al. Application-specific optimization of echo
time in MR pulse sequences for investigating materials with
susceptibilities different from that of the background in which
they are embedded [0062] U.S. Pat. No. 6,505,064 Liu et al.,
Diagnostic imaging systems and methods employing temporally
resolved intensity tracing [0063] U.S. Pat. No. 6,658,280 Haacke;
E. Mark, Susceptibility weighted imaging [0064] U.S. Pat. No.
7,301,644 by Knighton et al., Enhanced optical coherence tomography
for anatomical mapping [0065] U.S. Pat. No. 2008/0100612
Dastmalchi; Shahram Shawn et al., User interface for efficiently
displaying relevant OCT imaging data [0066] U.S. Pat. No.
2007/0291277 Everett; Matthew J.; et al., Spectral domain optical
coherence tomography system
Other Publications
[0066] [0067] Leitgeb, R. A., et al. (2003). "Performance of
Fourier domain vs. time domain optical coherence tomography."
Optics Express 11(8): 889-894. [0068] Hausler, G. and M. W. Lindner
(1998). ""Coherence Radar" and "Spectral Radar"--New Tools for
Dermatological Diagnosis." Journal of Biomedical Optics 3(1):
21-31. [0069] Napel S, Rubin G D, Jeffrey R B Jr., "STS-MIP: a new
reconstruction technique for CT of the chest", J. Comput Assist
Tomogr 1993;17:832-838. [0070] Narayana D L V Rao, Manpreet Singh
Gulati, Shashi Bala Paul, Girish Kumar Pande, Peush Sahni and
Tushar Kanti Chattopadhyay, "Three-dimensional helical computed
tomography cholangiography with minimum intensity projection in
gallbladder carcinoma patients with obstructive jaundice:
Comparison with magnetic resonance cholangiography and percutaneous
transhepatic cholangiography", Journal of Gastroenterology and
Hepatology, Volume 20, Issue 2, Pages 304-308, 19 Jan. 2005. [0071]
A. Salles, M. Nino-Murcia, R. Jeffrey, "CT of pancreas: minimum
intensity projections", Abdominal Imaging, Volume 33, Number 2,
March 2008 , pp. 207-213(7).
* * * * *