# US 2018/0020909 A1 · Automated quantification of nonperfusion in the retina using optical coherence tomography angiography

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2018/0020909 A1
- Document ID: US20180020909A1
- Status: Published application
- Kind: A1
- Application number: 15/655,040
- Filed: 2017-07-20
- Published: 2018-01-25
- Earliest claimed: 2016-07-20
- Assignees: Oregon Health & Science University
- Applicants: Oregon Health & Science University
- Inventors: Yali Jia; David Huang; Miao Zhang
- Web page: https://patentgrep.com/patent/US20180020909A1

## CPC

- A61B 3/0025: {characterised by electronic signal processing, e.g. eye models}
- A61B 3/102: {for optical coherence tomography [OCT]}
- A61B 3/1233: {for measuring blood flow, e.g. at the retina}
- A61B 3/14: Arrangements specially adapted for eye photography
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)

## Abstract

Disclosed are methods and systems for measuring areas of nonperfusion in the retina using OCT imaging. The disclosed methods and systems allow for the automated segmentation and quantification of avascular areas of the retina utilizing information obtained from both structural OCT and OCT angiography (OCTA) data. The disclosed methods include filtering approaches which enhance vessel structure while suppressing noise, dynamic thresholding approaches to mitigate the detrimental effects of within-scan variability and low scan quality, and distance transform-based approaches to improve detection of ischemic regions. When combined with methods such as projection-resolved OCTA, the sensitivity to detect nonperfusion within different plexuses of the inner retina is demonstrated. In the clinical setting of diabetic retinopathy, the disclosed methods and systems show high sensitivity and specificity to detect the mild non-proliferative form of the disease with high reproducibility.

## Claims

### Claim 1

A computer-based method of measuring nonperfusion in the retina, the method comprising:
obtaining an optical coherence tomography (OCT) angiogram;
thresholding the OCT angiogram, thereby generating a binary vessel mask;
generating a vessel distance map from the binary vessel mask; and
generating a nonperfusion map from the vessel distance map.

### Claim 2 (depends on 1)

The method of claim 1, wherein the OCT angiogram is a projection resolved OCT angiography (PR-OCTA) angiogram.

### Claim 3 (depends on 1)

The method of claim 1, further comprising, prior to the thresholding, applying a vesselness filter to the OCT angiogram to enhance the vessel structure in the OCT angiogram.

### Claim 4 (depends on 3)

The method of claim 3, wherein the vesselness filter is a two-scale vesselness filter.

### Claim 5 (depends on 1)

The method of claim 1, wherein thresholding the OCT angiogram comprises:
obtaining a structural OCT;
generating an en face reflectance image from the structural OCT;
filtering the en face reflectance image, thereby generating a reflectance-adjusted threshold image including respective reflectance-adjusted thresholds; and
thresholding the OCT angiogram using the respective reflectance-adjusted thresholds, thereby generating the binary vessel mask.

### Claim 6 (depends on 5)

The method of claim 5, wherein the structural OCT comprises the ganglion cell layer (GCL) and inner plexiform layer (IPL).

### Claim 7 (depends on 6)

The method of claim 6, wherein the thresholding the OCT angiogram using the reflectance-adjusted thresholds is performed according to:
B
xy
=
{
1
,
if


D
xy
>
T
xy
0
,
if


D
xy
<
T
xy
;
wherein Bxy is the binary vessel mask, Dxy is the OCT angiogram, and Txy is the reflectance-adjusted threshold image.

### Claim 8 (depends on 5)

The method of claim 5, wherein the filtering the en face reflectance image comprises:
calculating a logarithm of the en face reflectance image, thereby generating a log-transformed en face reflectance image;
applying a smoothing filter to the log-transformed en face reflectance image, thereby generating a smoothed log-transformed en face reflectance image;
scaling and offsetting the smoothed log-transformed en face reflectance image, thereby generating a scaled reflectance-adjusted threshold image; and
setting a maximum threshold in a region of the scaled reflectance-adjusted threshold image, thereby generating the reflectance-adjusted threshold image.

### Claim 9 (depends on 8)

The method of claim 8, wherein the scaling and offsetting is performed according to:
Txy=a×G(S)xy+b
wherein Txy is the reflectance-adjusted threshold image, G is a Gaussian operator, S corresponds to the log-transformed en face reflectance image, a is a scaling parameter, and b is an offset parameter.

### Claim 10 (depends on 8)

The method of claim 8, wherein the smoothing filter is a Gaussian filter.

### Claim 11 (depends on 10)

The method of claim 10, wherein the Gaussian filter is sized 15×15 pixels with a standard deviation of 8 pixels.

### Claim 12 (depends on 8)

The method of claim 8, wherein the region of the scaled reflectance-adjusted threshold image is a foveal avascular zone (FAZ) region.

### Claim 13 (depends on 12)

The method of claim 12, wherein the FAZ region comprises a 6 mm-diameter circular area centered on the FAZ.

### Claim 14 (depends on 1)

The method of claim 1, wherein generating the vessel distance map from the binary vessel mask comprises:
denoising the binary vessel mask at a first level, thereby generating a first denoised binary mask image;
denoising the binary vessel mask at a second level, thereby generating a second denoised binary mask image;
applying a distance transform to the first denoised binary mask image, thereby generating a level-1 vessel distance map;
applying a distance transform to the second denoised binary mask image, thereby generating a level-2 vessel distance map; and
averaging the level-1 vessel distance map and level-2 vessel distance map, thereby generating the vessel distance map.

### Claim 15 (depends on 14)

The method of claim 14, wherein the distance transform is a Euclidean distance transform calculated according to:
DTxy=minBx′y′=1√{square root over ((x−x′)2+(y−y′)2)}
wherein DTxy is the distance transform, (x, y) is a given pixel of the binary vessel mask, (x′, y′) corresponds to other pixels of the binary vessel mask, and Bx′y′=1 corresponds to pixels of the binary vessel mask that have a value to indicate a vessel.

### Claim 16 (depends on 1)

The method of claim 1, wherein generating the nonperfusion map from the vessel distance map comprises:
thresholding the vessel distance map, thereby generating a first thresholded image;
eroding the first thresholded image, thereby generating a second thresholded image;
filtering the second thresholded image, thereby generating a third thresholded image; and
dilating the third thresholded image, thereby generating the nonperfusion map.

### Claim 17 (depends on 16)

The method of claim 16, wherein the thresholding the vessel distance map is performed using a value of DT>4 when the vessel distance map is calculated according to:
DTxy=minBx′y′=1√{square root over ((x−x′)2+(y−y′)2)}
wherein DTxy is the distance transform, (x, y) is a given pixel of the binary vessel mask, (x′, y′) corresponds to other pixels of the binary vessel mask, and Bx′y′=1 corresponds to pixels of the binary vessel mask that have a value to indicate a vessel.

### Claim 18 (depends on 16)

The method of claim 16, wherein the eroding the first thresholded image is performed using a 5-pixel wide square kernel.

### Claim 19 (depends on 16)

The method of claim 16, wherein the filtering the second thresholded image comprises eliminating areas of the second thresholded image smaller than 8 pixels.

### Claim 20 (depends on 19)

The method of claim 19, wherein the filtering the second thresholded image further comprises eliminating areas of the second thresholded image whose minor axis length is smaller than 2 pixels.

### Claim 21 (depends on 16)

The method of claim 16, wherein the dilating the third thresholded image is performed using a 7-pixel wide square kernel.

### Claim 22

A system for quantifying nonperfusion in the retina, comprising:
an OCT system to acquire an OCT angiogram of a sample;
a logic subsystem; and
a data holding subsystem comprising machine-readable instructions stored thereon that are executable by the logic subsystem to:
threshold the OCT angiogram, thereby generating a binary vessel mask;
generate a vessel distance map from the binary vessel mask; and
generate a nonperfusion map from the vessel distance map.

### Claim 23 (depends on 22)

The system of claim 22, wherein the OCT angiogram is a projection resolved OCT angiography (PR-OCTA) angiogram.

### Claim 24 (depends on 22)

The system of claim 22, wherein the instructions are further executable by the logic subsystem to, prior to performing the thresholding, apply a vesselness filter to the OCT angiogram to enhance the vessel structure in the OCT angiogram.

### Claim 25 (depends on 22)

The system of claim 22, wherein, to threshold the OCT angiogram, the logic subsystem is to:
obtain a structural OCT of the sample;
generate an en face reflectance image from the structural OCT;
filter the en face reflectance image, thereby generating a reflectance-adjusted threshold image including respective reflectance-adjusted thresholds; and
threshold the OCT angiogram using the respective reflectance-adjusted thresholds, thereby generating the binary vessel mask.

### Claim 26 (depends on 25)

The system of claim 25, wherein the structural OCT comprises the ganglion cell layer (GCL) and inner plexiform layer (IPL).

### Claim 27 (depends on 6)

The system of claim 6, wherein, the logic subsystem is to threshold the OCT angiogram using the reflectance-adjusted thresholds according to:
B
xy
=
{
1
,
if


D
xy
>
T
xy
0
,
if


D
xy
<
T
xy
;
wherein Bxy is the binary vessel mask, Dxy is the OCT angiogram, and Txy is the reflectance-adjusted threshold image.

### Claim 28 (depends on 25)

The system of claim 25, wherein, to filter the en face reflectance image, the logic subsystem is to:
calculate a logarithm of the en face reflectance image, thereby generating a log-transformed en face reflectance image;
apply a smoothing filter to the log-transformed en face reflectance image, thereby generating a smoothed log-transformed en face reflectance image;
scale and offset the smoothed log-transformed en face reflectance image, thereby generating a scaled reflectance-adjusted threshold image; and
set a maximum threshold in a region of the scaled reflectance-adjusted threshold image, thereby generating the reflectance-adjusted threshold image.

### Claim 29 (depends on 28)

The system of claim 28, wherein the logic subsystem is to scale and offset the smoothed log-transformed en face reflectance image according to:
Txy=a×G(S)xy+b
wherein Txy is the reflectance-adjusted threshold image, G is a Gaussian operator, S corresponds to the log-transformed en face reflectance image, a is a scaling parameter, and b is an offset parameter.

### Claim 30 (depends on 28)

The system of claim 28, wherein the smoothing filter is a Gaussian filter sized 15×15 pixels with a standard deviation of 8 pixels.

### Claim 31 (depends on 28)

The system of claim 28, wherein the region of the scaled reflectance-adjusted threshold image is a foveal avascular zone (FAZ) region.

### Claim 32 (depends on 31)

The system of claim 31, wherein the FAZ region comprises a 6 mm-diameter circular area centered on the FAZ.

### Claim 33 (depends on 22)

The system of claim 22, wherein, to generate the vessel distance map from the binary vessel mask, the logic subsystem is to:
denoise the binary vessel mask at a first level, thereby generating a first denoised binary mask image;
denoise the binary vessel mask at a second level, thereby generating a second denoised binary mask image;
apply a distance transform to the first denoised binary mask image, thereby generating a level-1 vessel distance map;
apply a distance transform to the second denoised binary mask image, thereby generating a level-2 vessel distance map; and
average the level-1 vessel distance map and level-2 vessel distance map, thereby generating the vessel distance map.

### Claim 34 (depends on 33)

The system of claim 33, wherein the distance transform is a Euclidean distance transform calculated according to:
DTxy=minBx′y′=1√{square root over ((x−x′)2+(y−y′)2)}
wherein DTxy is the distance transform, (x, y) is a given pixel of the binary vessel mask, (x′, y′) corresponds to other pixels of the binary vessel mask, and Bx′y′=1 corresponds to pixels of the binary vessel mask that have a value to indicate a vessel.

### Claim 35 (depends on 22)

The system of claim 22, wherein, to generate the nonperfusion map from the vessel distance map, the logic subsystem is to:
threshold the vessel distance map, thereby generating a first thresholded image;
erode the first thresholded image, thereby generating a second thresholded image;
filter the second thresholded image, thereby generating a third thresholded image; and
dilate the third thresholded image, thereby generating the nonperfusion map.

### Claim 36 (depends on 35)

The system of claim 35, wherein the logic subsystem is to threshold the vessel distance map using distance transforms of the vessel distance map having a value greater than 4, and wherein the vessel distance map is calculated according to:
DTxy=minBx′y′=1√{square root over ((x−x′)2+(y−y′)2)}
wherein DTxy is the respective distance transform of the vessel distance map, (x, y) is a given pixel of the binary vessel mask, (x′, y′) corresponds to other pixels of the binary vessel mask, and Bx′y′=1 corresponds to pixels of the binary vessel mask that have a value to indicate a vessel.

### Claim 37 (depends on 36)

The system of claim 36, wherein the first thresholded image is eroded using a 5-pixel wide square kernel;
wherein the second thresholded image is filtered by eliminating areas of the second thresholded image smaller than 8 pixels and eliminating areas of the second thresholded image whose minor axis length is smaller than 2 pixels; and
wherein the third thresholded image is dilated using a 7-pixel wide square kernel.