U.S. patent application number 15/474513 was filed with the patent office on 2017-10-05 for methods of obtaining 3d retinal blood vessel geometry from optical coherent tomography images and methods of analyzing same.
The applicant listed for this patent is Bio-Tree Systems, Inc.. Invention is credited to Raul A. Brauner, Kongbin Kang, Yanchun Wu.
Application Number | 20170287131 15/474513 |
Document ID | / |
Family ID | 59959502 |
Filed Date | 2017-10-05 |
United States Patent
Application |
20170287131 |
Kind Code |
A1 |
Kang; Kongbin ; et
al. |
October 5, 2017 |
Methods Of Obtaining 3D Retinal Blood Vessel Geometry From Optical
Coherent Tomography Images And Methods Of Analyzing Same
Abstract
Embodiments relate to extracting blood vessel geometry from one
or more optical coherent tomography (OCT) images for use in
analyzing biological structures for diagnostic and therapeutic
applications for diseases that can be detected by vascular changes
in the retina. An OCT image refers generally to one or more images
of any dimension obtained using any one or combination of OCT
techniques. Some embodiments include a method of identifying a
region of interest of a retina from a plurality of retinal blood
vessels in at least one optical coherence tomography (OCT) image of
at least a portion of the retina. Some embodiments include a method
of distinguishing between a plurality of retinal layers from vessel
morphology information of retinal blood vessels in at least one
optical coherence tomography (OCT) image of at least a portion of
the retina.
Inventors: |
Kang; Kongbin; (Providence,
RI) ; Wu; Yanchun; (Sharon, MA) ; Brauner;
Raul A.; (Framingham, MA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Bio-Tree Systems, Inc. |
Framingham |
MA |
US |
|
|
Family ID: |
59959502 |
Appl. No.: |
15/474513 |
Filed: |
March 30, 2017 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
62316490 |
Mar 31, 2016 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06T 7/0012 20130101;
G06T 2207/10101 20130101; G06T 2207/30041 20130101; G06T 2207/30101
20130101 |
International
Class: |
G06T 7/00 20060101
G06T007/00 |
Claims
1. A method of identifying a 3D region of interest of a retina from
a plurality of retinal blood vessels in a 3D optical coherence
tomography (OCT) image, comprising more than one OCT image, of at
least a portion of the retina, the method comprising: using at
least one computer programmed to perform: evaluating at least one
first morphological feature of a plurality of retinal blood vessels
detected in the 3D OCT image at each of a plurality of voxel
locations in the 3D OCT image; and identifying a first region of
interest based at least in part on evaluating the at least one
first morphological feature of the plurality of blood vessels.
2. The method of claim 1, wherein identifying the first region of
interest comprises constructing a likelihood function of voxels
belong to the first region of interest and a smoothness constraint
to force the first region of interest to be continuous, based on at
least one morphological feature.
3. The method of claim 1, wherein evaluating at least one first
morphological feature comprises determining a feature field
comprising the at least one morphological feature at each of a
plurality of voxel locations, and wherein identifying the first
region of interest comprises evaluating the feature field.
4. The method of claim 3, wherein the feature field comprises a
vessel diameter field, a vessel density field, a vessel branching
point density field, vessel branch length field, vessel curvature
field, vessel tortuosity field and/or a vessel branching angle
field.
5. The method of claim 3, where the feature field is evaluated
relative to known characteristics of each retinal layer.
6. The method of claim 1, wherein the at least one 3D OCT image
comprises an OCT angiography (OCTA) image.
7. A method of distinguishing between a plurality of retinal layers
from vessel morphology information of retinal blood vessels in a 3D
optical coherence tomography (OCT) image, comprising a plurality of
OCT images, of at least a portion of a retina, the method
comprising: using at least one computer programmed to perform:
evaluating at least one first morphological feature of at least one
retinal blood vessel at each of a plurality of voxel locations in
the 3D OCT image; and determining that at least one first voxel
location of the plurality of voxel locations belongs to a first
layer of the plurality of retinal layers and that at least one
second location of the plurality of voxel locations belongs to a
second layer of the plurality of retinal layers based on evaluating
the at least one first morphological feature.
8. The method of claim 7, wherein the at least one computer is
programmed to determine to which of the plurality of retinal layers
each of the plurality of voxel locations belongs.
9. The method of claim 7, wherein the at least one first
morphological feature comprises vessel directional vessel density
field at each of the plurality of voxel locations.
10. The method of claim 7, wherein evaluating the at least one
morphological feature comprises evaluating a vessel diameter field,
a vessel density field, a vessel branching point density field,
vessel branch length field, vessel curvature field, vessel
tortuosity field and/or a vessel branching angle field at each of
the plurality of voxel locations relative to known characteristics
of each retinal layer.
11. The method of claim 7, wherein evaluating the at least one
morphological feature comprises determining a feature field
comprising the at least one morphological feature at each of a
plurality of voxel locations, and wherein determining a retinal
layer comprises evaluating the feature field.
12. The method of claim 11, wherein the feature field comprises a
vessel diameter field, a vessel density field, a vessel branching
point density field, vessel branch length field, vessel curvature
field, vessel tortuosity field and/or a vessel branching angle
field.
13. The method of claim 11, wherein the feature field is evaluated
relative to known characteristics of each retinal layer.
14. The method of claim 7, wherein the 3D OCT image comprises an
OCT angiography (OCTA) image.
Description
[0001] This application claims priority from U.S. Provisional
Patent Application Ser. No. 62/316,490, filed Mar. 31, 2016, the
disclosure of which is incorporated herein by reference in its
entirety.
BACKGROUND
[0002] A wide range of imaging methods and devices are commonly
used to evaluate physiological conditions. Tools have been
developed to image body structures based on different physical
properties. For example, X-rays, CT scans, MRIs, PET scans, IR
analyses and other technologies have been developed to obtain
images of various body structures. These tools are routinely used
for diagnostic, therapeutic, and research applications. More
recently, Optical Coherent Tomography (OCT) has been used to
capture three-dimensional images from with biological tissue, and
specifically the eye. In fact, OCT is well suited to ophthalmic
applications. In these applications, OCT can be used to take a
cross-section of the retina.
[0003] However, the data obtained using OCT must be processed and
analyzed. It would be beneficial if there were a method and
apparatus that was capable of processing these OCT images. More
particularly, it would be advantageous if this method could
determine which layer of the retina each point in the OCT scan
belonged to. Further, it would be beneficial if the method and
apparatus could accurately reconstruct the vasculature from the OCT
scan.
SUMMARY
[0004] Some embodiments relate to extracting blood vessel geometry
from 3D images comprising more than one optical coherent tomography
(OCT) images for use in analyzing biological structures for
diagnostic and therapeutic applications for diseases that can be
detected by vascular changes in the retina. A 3D OCT image refers
generally to one or more images of any dimension obtained using any
one or combination of OCT techniques including, but not limited to,
OCT, contrasted OCT, OCT angiography (OCTA), swept source OCT, en
face OCT, etc. In particular, certain embodiments relate to
extracting geometry from one or more OCT images of blood vessels to
identify structural features useful for detecting, monitoring,
and/or treating diseases, and/or for evaluating and validating new
therapies.
[0005] Some embodiments include a method of identifying a 3D region
of interest of a retina from a plurality of retinal blood vessels
in a 3D optical coherence tomography (OCT) image, comprising a
plurality of OCT images, of at least a portion of the retina, the
method comprising using at least one computer programmed to perform
evaluating at least one first morphological feature of a plurality
of retinal blood vessels detected in the 3D OCT image, and
identifying a first region of interest based at least in part on
evaluating the at least one first morphological feature of the
plurality of blood vessels.
[0006] Some embodiments include a method of distinguishing between
a plurality of retinal layers from vessel morphology information of
retinal blood vessels in a 3D optical coherence tomography (OCT)
image, comprising a plurality of OCT images, of at least a portion
of the retina, the method comprising using at least one computer
programmed to perform evaluating at least one first morphological
feature of at least one retinal blood vessel at each of a plurality
of voxel locations in the 3D OCT image, and determining that at
least one first voxel location of the plurality of voxel locations
belongs to a first layer of the plurality of retinal layers and
that at least one second location of the plurality of voxel
locations belongs to a second layer of the plurality of retinal
layers based on evaluating the at least one first morphological
feature.
[0007] Some embodiments include a method of determining scale at
each of a plurality of voxels of a 3D coherence tomography (OCT)
image, comprising a plurality of OCT images, of at least one
retinal blood vessel using an orientation independent scale
detector having a size defined by a radius, the scale indicative of
a distance from the respective voxel to a wall of a retinal blood
vessel, the method comprising using at least one computer
programmed to perform applying the scale detector at a first voxel
of the plurality of voxels using a plurality of radii to obtain a
response for each of the plurality of radii, each response having
plurality of values, ordering the plurality of values in each
response to obtain a plurality of ordered responses, each of the
plurality of ordered responses corresponding to a respective
radius, selecting one of the plurality of radii as a first scale
candidate based, at least in part, on a result of evaluating each
of the plurality of ordered responses according to a first
function, selecting one of the plurality of radii as a second scale
candidate based, at least in part, on a result of evaluating each
of the plurality of ordered responses according to a second
function, selecting, as the scale at the first voxel, either the
first scale candidate or the second scale candidate based on a
first criteria.
[0008] Some embodiments include a method of determining scale at
each of a plurality of voxels of a 3D optical coherence tomography
(OCT) image, comprising a plurality of OCT images, of at least one
retinal blood vessel using an orientation independent scale
detector having a size defined by a radius, the scale indicative of
a distance from the respective voxel to a wall of a retinal blood
vessel, the method comprising at least one computer programmed to
perform applying the scale detector at a first voxel of the
plurality of selected voxels using a plurality of radii to obtain a
response for each of the plurality of radii, each response having
plurality of values, ordering the plurality of values in each
response to obtain a plurality of ordered responses, selecting one
of the plurality of radii as a first scale candidate based, at
least in part, on a result of evaluating an average of the lowest k
values in each respective ordered response.
BRIEF DESCRIPTION OF THE FIGURES
[0009] For a better understanding of the present disclosure,
reference is made to the accompanying drawings, which are
incorporated herein by reference and in which:
[0010] FIG. 1 illustrates a flow chart for the invivo retina vessel
morphology information extraction system;
[0011] FIG. 2 shows an illustration showing the retina layers
defined by vessels;
[0012] FIGS. 3A-B are an example of layer detection based on retina
vessel morphological differences between the deep and the
intermediate layers;
[0013] FIGS. 4A-B show the improvements of the centerline filter on
a vein. FIG. 4A shows the result from a classical centerline filter
and FIG. 4B shows the result from the improved centerline filter,
which results in a more consistent vein segmentation and
linking;
[0014] FIGS. 5A-B show examples of retina vessels segmented and
linked from 3D OCTA datasets. FIG. 5A shows non-proliferative
diabetic retinopathy and FIG. 5B shows normal retinopathy; and
[0015] FIGS. 6A-6B show examples of vasometrics on retinopathy.
[0016] According to some embodiments, methods and apparatus for
retina micro-vessel segmentation and morphology based region
retrieval from 3D OCT Angiograph images are described.
[0017] Methods and apparatus for extracting vessel geometry from
OCT images (e.g., a 3D OCT image of a retina or a portion of a
retina) are disclosed. These methods and apparatus relate to
obtaining vessel geometry, determining one or more structural
features from the vessel geometry, and/or analyzing the one or more
structural features for retinal diagnostic, prognostic, and/or
research applications. The methods described herein are useful for
obtaining a geometrical representation of a vascular tree or vessel
network of retinal blood vessels from more than one OCT images
(e.g., a 3D OCT image) that contains data relating to
three-dimensional location, orientation and/or size points in the
vascular tree. In some embodiments, a vascular tree may be
represented by a series of disks or poker chips (e.g., circular or
elliptical disks) that are linked together to form a
three-dimensional structure containing information relating to the
local size, shape, branching, and other structural features at any
point in the vascular tree (e.g., a retinal blood vessel
network).
[0018] It should be appreciated that the entire vascular tree of a
retina may be represented by a network of linked poker chips (e.g.,
circular or elliptical disks). However, in many embodiments, only a
subset or a portion of a vascular tree of the retina may be
represented or analyzed. In some embodiments, a portion of a
vascular tree can be represented by a single disc or poker chip
that contains information relating to the location of the center of
the vessel, vessel size (diameter), and/or orientation (e.g., the
direction of the centerline of the vessel). In some embodiments, a
portion of a vascular tree may be represented by a dataset that
describes one or more poker chips along with information relating
to the linkage between the poker chips within a region of interest
of the vascular tree.
[0019] The present disclosure includes techniques for detecting at
least one feature associated with a retinal blood vessel in a 3D
OCT image of the retina or a portion of the retina and obtaining
the geometry of the retinal blood vessels in the 3D OCT image. Some
exemplary techniques for detecting and extracting geometry of blood
vessels from one or more images (e.g., an OCT image) are described
below and in U.S. Pat. No. 8,761,466, titled "Methods of Obtaining
Geometry from Images," which is incorporated by reference in its
entirety. Some techniques described therein may be referred to
herein generally as AngioTrack.TM. and/or, in the context of the
retina, RetinaAngioTrack.TM. or RetinaAngioTree.TM.. However, it
should be appreciated that these techniques are merely exemplary,
as the embodiments are not limited to the techniques described
therein. AngioTrack.TM. techniques may be used to obtain the
geometry of a blood vessel network from one or more images (e.g., a
3D OCT image) for use in detecting, monitoring, and/or treating
diseases, and/or for evaluating and validating new therapies (e.g.,
in human subjects or in other animals, for example other
mammals).
[0020] Some embodiments include a method of linking geometry
obtained from one or more images (e.g., a 3D OCT image) to form a
linked blood vessel network (e.g., a retinal blood vessel network)
that can be analyzed and mined for, for example, diagnostic and/or
therapeutic purpose. Exemplary techniques for linking geometry
(e.g., poker chips obtained from one or more images forming a Poker
Chip representation) are described in further detail in the U.S.
Pat. No. 8,761,466 and in International Publication Number WO
2014/143974 A1, titled "Methods and System for Linking Geometry
Obtained from Images," which is incorporated herein by reference in
its entirety. Some techniques described in these references are
generally referred to as linking or AngioLinking.TM. and/or, in the
context of the retina, RetinaAngioLinking.TM. or
RetinaAngioTree.TM.. It should be appreciated that the techniques
described in these references are merely exemplary, as the
embodiments are not limited to the techniques described therein.
Techniques described herein can be used on humans or other
animals.
[0021] In certain embodiments, the geometry of a blood vessel
network (e.g., a retinal blood vessel network obtained from one or
more OCT images) may be mined for physiological, biological, and/or
medical purposes. For example, a linked or unlinked poker chip
representation of a retinal blood vessel network may be analyzed
for diagnostic, monitoring and/or therapeutic purposes.
Accordingly, aspects of the invention relate to obtaining vessel
geometry, determining one or more structural features from the
vessel geometry, and/or analyzing the one or more structural
features for medical diagnostic, prognostic, and/or research
applications. Techniques for analyzing blood vessel geometry (e.g.,
a retinal blood vessel network) are described below, the '466
patent, the '971 Publication, and U.S. Publication No.
2015/0302584, titled "Vascular Analysis Methods and Apparatus,"
which is incorporated by reference in its entirety. Some techniques
described therein may be referred to generally as AngioProbe.TM.
or, in the context of the retina, RetinaAngioProbe.TM..
[0022] Some embodiments relate to automatically segmenting vessel
structure out of 3D OCTA dataset and mining information for the
earlier detection, intervention and monitoring the eye diseases,
like diabetic retinopathy, etc. The system to achieve this result
may include 4 subsystems: 1) vessel segmentation, 2) linking, 3)
vessel morphology based ROI extraction and 4) vasometrics
retrieval.
[0023] As shown in FIG. 1, the blocks represent exemplary
subsystems. The blocks can be implemented in a number of different
ways. For example, the blocks may be implemented using hardware,
software or a combination thereof. When implemented in software,
the software code can be executed on any suitable processor or
collection of processors, whether provided in a single computer or
distributed among multiple computers. It should be appreciated that
any component or collection of components that perform the
functions described herein can be generically considered as one or
more controllers that perform the disclosed functions. The
controllers can be implemented in numerous ways, such as with
dedicated hardware, or with general purpose hardware, such as
personal computers, that is programmed using microcode or software
to perform the functions recited herein.
[0024] It should be appreciated that the various methods outlined
herein may be coded as software that is executable on one or more
processors that employ any one of a variety of operating systems or
platforms. Additionally, such software may be written using any of
number of suitable programming languages and/or conventional
programming or scripting tools, and also may be compiled as
executable machine language code. It should be appreciated that one
embodiment is directed to a non-transitory computer-readable medium
encoded with one or more programs that, when executed, on one or
more computers or other processors, perform the methods that
implement the various embodiments disclosed herein.
[0025] Returning to FIG. 1, the data acquisition module will
generate a set of 3D images in the region of interest. The 3D
invivo images may either be from a OCT Angiography or contrasted
OCT scan. In the former case, blood vessels are contrasted by blood
flow speed or the decorrelation. In the latter case, patient is
injected contrast agent to increase the signal noise ratio between
vessels and tissues, then a retina region of interest, i.e. macula
region, is scanned by a high resolution OCT scanner.
[0026] This set of images is fed into a vessel segmentation block
to generate a set of unlinked poker chips which is used for data
mining. The unlinked poker chips output from this vessel
segmentation block contains information relating to the location of
the center of the vessel, vessel size (diameter), and/or
orientation (e.g., the direction of the centerline of the vessel).
Then those poker chips are fed into the Angiolinkage block to
recover all the morphological branching structure of a vessel
network. Based on the recovered vessel morphology, the region of
interests can be extracted by the ROI detector block. Those ROI can
be a disease region for the purpose of disease monitoring, a retina
layer for diabetes diagnosis, and etc. The final result of
Vasometrics is output by a vasometric block and saved to storage or
output, such as to a graph. Each of these blocks is described in
more detail below.
[0027] Some embodiments exhibit one or more of the following
advantages, though exhibiting such advantages is not a requirement:
[0028] 1. An 3D Angiotrack (not 1D or 2D feather) based system to
reliably extract vessels segments (poker chips) from OCTA invivo
scan. [0029] 2. An Angiolinking based system to extract branching
structures and vessel morphology structures. [0030] 3. Use vessel
morphologic information to separate Region of Interests [0031] 4.
An set of branching-morphology based vasometrics may be used for
the purpose of earlier detection, intervention and monitoring the
eye disease, i.e. diabetic retinopathy
[0032] The exemplary systems can also be applied to CT scan of the
exvivo retina data.
[0033] Region of Interest (ROI) Extraction Using Vessel
Morphology
[0034] The block labelled "Vessel Morphology based ROI detection"
in FIG. 1 is now described in more detail. Many retina functional
regions, e.g. layers and disease regions, can be separated based on
the vessel morphology structures. FIG. 2 shows the layers of
vessels and tissues. As shown in this illustration, the vessels lay
flat on the layers. There are vessel morphology differences between
the various layers. For example, the deep layer has vessels with
dense branching structure and relative uniform diameters; while the
intermediate layer has vessels of sparse branching points and large
range of vessel diameters. Using such vessel morphology
differences, retina layers are separable and differentiated.
[0035] According to some embodiments, the procedure to extract ROI
based on vessel morphology is as follows: [0036] compute a
morphology based feature field D(x, y, z) which is a function of
vessel diameter, vessel density, vessel branching point density,
vessel direction and vessel diameter. [0037] construct a likelihood
or energy function of the D(x, y, z) belonging to ROI [0038]
construct a penalty function to control how much influence neighbor
positions can affect each other (smoothness term) [0039] Extract
ROI based on the smoothness requirement of the ROI and data
agreement with model
[0040] Provided below is an example of separating Retina layers. As
described above, each layer of the retina may have specific
characteristics that are unique to that layer. Consequently, by
comparing the vasculature to predetermined characteristics
associated with various layers, it is possible to determine which
layer each point in a scan belongs to.
[0041] For example, according to some embodiments, a 3D scalar
density field, D={.phi.(x, y, z)}, of vessels information at every
point (x, y, z) is computed from linked or unlinked vessel poker
chips. The goal or target is to label each point l(x, y, z) for the
layer it belongs to. Suppose that L is the set of labels of the
layer number on each point, L={l(x, y, z)} for any (x, y, z).
[0042] According to some embodiments, retinal layers are separated
or labeled from a directional vessel density field D(x, y, z). In
other embodiments, other parameters may be used to separated or
label the layers. In this example, D is the density of vessels
projected to the trans-axis plane. The labelling can be written
as:
L * = max L P ( L D ) .varies. max L P ( L , D ) Consider P ( L , D
) = x , y , z P ( D ( x , y , z ) L ( x , y , z ) ) x ' , y ' , z '
.noteq. x , y , z P ( L ( x ' , y ' , z ' ) , L ( x , y , z ) ) ( 1
) ##EQU00001##
[0043] A piece wise model can be used, then the joint probability
of neighbor point i and j can be denoted as .PSI.(i,j). If the
posterior likelihood is denoted as .phi.(i), then the joint
probability can be rewritten as:
P = i .PHI. ( i ) j .di-elect cons. N ( i ) \ i .PSI. ( i , j )
##EQU00002##
[0044] where N(i)\j is the neighborhood of i except j. In one
example,
.PHI. ( i ) = 1 2 .pi. .sigma. l ( i ) e - ( D ( i ) - u l ( i ) )
2 2 .sigma. l ( i ) 2 ##EQU00003## and ##EQU00003.2## .PSI. ( i , j
) = 1 2 .pi. .alpha. e - ( l i - l j ) 2 2 .alpha.
##EQU00003.3##
[0045] This equation is solvable by Belief Propagation. The message
that node i sends to node j may be denoted as m.sub.ij, and
m.sub.ii is the message that the observation D.sub.i sends to node
i. We also denote b.sub.i as the belief of node i. Then the update
rules are as follows:
m ij ( l j ) = 1 Z l i .PSI. ( i , j ) m ii ( l i ) k .di-elect
cons. N ( i ) \ i m ki ( l i ) ##EQU00004## b i ( l i ) = 1 Z m ii
( l i ) k .di-elect cons. N ( i ) \ i m ki ( l i )
##EQU00004.2##
[0046] The label L* is obtained by initially assigning a uniform
message m.sub.ij(l). On the iteration t, the message m.sub.ij.sup.i
and belief b.sub.i.sup.t are updated according to the above rule.
The label L*.sup.t is computed by
L ( i ) * t = arg min I i b i ( l i ) ##EQU00005##
the model parameter at iteration t may be computed using:
.mu. l t = l t ( i ) = l D (i) N - 1 ##EQU00006## and
##EQU00006.2## .sigma. l t = l t ( i ) = l ( D ( i ) - .mu. l t ) 2
N - 1 ##EQU00006.3##
[0047] Using this algorithm, the specific layer of the retina may
be determined. FIG. 3 shows an example of layer detection based on
retina morphological differences. As stated above, the intermediate
layer has vessels of sparse branching points and large range of
vessel diameters. The algorithm described above is able to identify
the right figure as the intermediate layer.
Vessel Segmentation and Linking
[0048] Vessel segmentation of retinal blood vessels from OCT,
contrasted OCT and/or OCTA images (collectively referred to as OCT)
may be performed using various techniques. According to some
embodiments, these techniques may be supplemented with further
techniques described herein to produce a linked poker chips for
retina blood vessels from one or more OCT images of the retina or a
portion thereof.
[0049] Normally, the intensity of 3D image outside of a vessel is
significant lower than the intensity inside the vessel. This rapid
intensity decay enables the detection of scale. Mathematically, a
ratio of a given location X is calculated for measuring the
intensity decay by:
( X , r ) = f - ( { I ( X ' ) : X ' - X = r + 1 } ) min r { f + ( {
I ( X ' ) : X ' - X = 1 , , r } ) } ##EQU00007##
[0050] where f.sub.- and f.sub.+ are a rank functions,
respectively. Given noise models, there are a lot of ways to choose
those rank functions. In order to cope with reconstruction effects,
f.sub.- may be chosen as the median value of last 8 lowest
intensities and f.sub.+ may be chosen as the median value of top 8
highest intensities. With the ratio response, the scale
.sigma..sub.r(X) may obtained by finding the minimum radius r so
that (X, r) reach the threshold a and the next larger radius has
smaller ratio:
.sigma. ( X ) = min r { ( X , r ) < 1 .alpha. ( X , r + 0.5 )
.ltoreq. ( X , r ) } ##EQU00008##
[0051] This formula holds for the large vessels in the OCTA 3D
images. However, due to the weak signal of the velocity or phase
information presented in the small and medium size vessels, the
angiotrack scale detection may behave incorrectly in the small and
middle size vessels. A more robust scale estimator for small and
middle size vessels has to be constructed. Some embodiments include
a rank based scale filter which is not sensitive to the weak
signals of the noise and outliers. Given a point X inside a vessel,
the rank based scale filter may be defined as:
1 ( X , r ) = { X ' : r < X ' - X .ltoreq. r + 1 I ( X ' ) <
T B } { X ' : X ' - X .ltoreq. r } ##EQU00009##
[0052] where T.sub.B is an intensity threshold and the normal of
the set, .parallel. .parallel., is defined as the number of points
of the set. Then the scale .sigma..sup.1(X) is obtained by finding
the minimum radius r so that .sub.1(X, r) reach the threshold
T.sub.R1 and the next larger radius has smaller ratio:
.sigma. 1 ( X ) = min r { 1 ( X , r ) < T R 1 1 ( X , r + 0.5 )
.ltoreq. 1 ( X , r ) } ##EQU00010##
[0053] Here the value of T.sub.R1 normally is chosen to be 0.05. To
facilitate scale detection on both small vessels and large vessels,
the true scale for a given point X regardless of the vessel size is
determined by the minimum of both algorithms.
.sigma.(X)=min{.sigma..sup.1(X),.sigma.(X)}
[0054] In summary, scale detection uses rank filters to deal with
the problem of variable signal: [0055] construct a set of shells
with different radius and fixed shell thickness. Those shells are
used as probes to find out the size of a vessel [0056] sort the
intensity information inside shell and assign them ranks/orders
[0057] use rank information to decide which shells touches the
outside of vessels [0058] use the radius of this shell as the
scale
Anisotropic Images
[0059] It is well known that the OCT, including OCTA, images have
lower resolution on the axis direction than the trans-axis
directions. This causes the vessel cross-section to be an ellipse,
instead of a circle. Furthermore, veins normally have a more flat
cross-section than arteries. An improved centerline filter is used
to reduce the effects.
f s ( r , z ) = { 1 r .ltoreq. s and z .ltoreq. 2 s 0 r > sor z
.ltoreq. 2 s ( 2 ) ##EQU00011##
[0060] Compared with the normal matching filter, this filter
removes the negative kernel in order to pin down the center better.
FIG. 4 shows the comparison results before and after modification
of the filter. We can see two most front vessels have no branching
points which means more poker chips are recovered before
linking.
Vasometrics to Detect Retinopathy and Monitor Disease
Progression
[0061] Vessel morphological pattern (vessel metrics) may be used to
detect retinopathy and/or monitoring the disease progression. The
candidate patterns include, but not limited to, the following
elements and the mathematic function of each of these variables:
[0062] The vessel pokerchip density. Number of vessel pokerchips
per unit volume [0063] The vessel branch density. Number of
branches per unit volume [0064] The vessel branching point density.
Number of branching points per unit volume [0065] The vessel
branching length distribution. The relative histogram (emperical
distribution) of vessel branching length. [0066] The vessel
curvature distribution. The relative histogram of vessel's
curvatures [0067] The vessel tortuosity distribution. The relative
histogram of vessel's tortuosity
[0068] Those features may used to monitor the progression of
disease progression and detect retinopathy. FIG. 5 shows an example
of diabetic retinopathy progression using the present system. From
the segmented vessel network, the vessel morphologic difference
between a non-proliferative retinopathy (FIG. 5a) and a healthy
retinopathy (FIG. 5b) can be clearly seen. The non-proliferative
retinopathy example shows the eliminations of the large vessels and
the increase in the number of small vessels, which can be captured
by the above variables. A quantitative changes of this example was
shown in FIG. 6 using the present system.
[0069] Blood vessel geometry obtained from one or more OCT images
may be analyzed according to any of the techniques described herein
to perform diagnosis or guide therapy of retinal disorders,
including but not limited to macular degeneration, diabetic
retinopathy, mild nonproliferative retinopathy, moderate
nonproliferative retinopathy, severe nonproliferative retinopathy,
proliferative diabetic retinopathy and diabetic macular edema
(DME). Blood vessel geometry obtained from one or more OCT images
may also be analyzed according to any of the techniques described
above to perform diagnosis or guide therapy of neurodegenerative
disorders, including but not limited to Parkinson's Disease,
Alzheimer's disease, amyotrophic lateral sclerosis,
olivopontocerebellar atrophy (OPCA), Ataxia telangiectasia, Batten
disease, Friedreich's ataxia, Spinal muscular atrophy, Huntington's
disease, Lewy body disease, or Hansen's disease. Further conditions
that can be detected or predicted via vascular changes in the
retina include stroke, cardiovascular disease, heart attack,
multiple sclerosis, drug toxicity, eye diseases (including but not
limited to Glaucoma), and cancers (including but not limited to
retinoblastoma). In some embodiments, diseases are evaluated in a
human subject.
[0070] In other words, RetinaAngioProbem includes methods for
analyzing structures or features of retinal blood vessels and
evaluating their association with disease, responsiveness to
therapeutic treatments, and/or other conditions. This methods
provide quantitative and analytical methods for evaluating and/or
comparing in vivo retinal blood vessel geometry obtained from
multiple OCT images. RetinaAngioProbem techniques can be useful in
assisting and/or automating the analysis of vascular patterns and
their association with disease diagnosis, prognosis, response to
therapy, etc., or any combination thereof, using for example, blood
vessel structural features (e.g., a distribution of vessel
parameters such as structural features or morphological parameters
within a region of interest may be generated and evaluated). In
some embodiments, vessel parameters may relate to the size, shape,
or number of vessels with a region of interest. A distribution may
be generated based on quantitative measurements related to one or
more parameters. In some embodiments, a distribution of blood
vessels may be a population distribution of blood vessels as a
function of quantitative measures of one or more parameters. For
example, a distribution may represent the number of blood vessels
(or the percentage of the blood vessel population) as a function of
their diameter, branching frequency, distance between branches,
degree of tortuousity, curvature, or any other quantitative
structural feature or morphological parameter, e.g., as described
herein, or any combination of two or more thereof.
[0071] As stated above, the above disclosure described a plurality
of methods which may be used to process OCT scans, and more
particularly OCT scans of the retina. In addition, each of these
methods may be executed by a controller, executing instructions
disposed on a non-transitory computer-readable storage medium.
Thus, the above disclosure also describes a system for performing
these methods, as well as a non-transitory computer-readable
storage medium which contains the instructions to perform these
methods.
[0072] The present disclosure is not to be limited in scope by the
specific embodiments described herein. Indeed, other various
embodiments of and modifications to the present disclosure, in
addition to those described herein, will be apparent to those of
ordinary skill in the art from the foregoing description and
accompanying drawings. Thus, such other embodiments and
modifications are intended to fall within the scope of the present
disclosure. Furthermore, although the present disclosure has been
described herein in the context of a particular implementation in a
particular environment for a particular purpose, those of ordinary
skill in the art will recognize that its usefulness is not limited
thereto and that the present disclosure may be beneficially
implemented in any number of environments for any number of
purposes. Accordingly, the claims set forth below should be
construed in view of the full breadth and spirit of the present
disclosure as described herein.
* * * * *