System and method for the segmentation of optical coherence tomography slices

Patent metadata

Field	Value
Publication number	US-10062162-B2
Application number	US-201615296790-A
Country	US
Kind code	B2
Filing date	Oct 18, 2016
Priority date	Oct 19, 2015
Publication date	Aug 28, 2018
Grant date	Aug 28, 2018

Abstract

The present disclosure describes a system and method to segment optical coherence tomography (OCT) images. The present system uses a hybrid method that employs both Bayesian level sets (BLS) and graph-based segmentation algorithms. The system first identifies retinal tissue within an OCT image using the BLS algorithms. The identified retinal tissue is then further segmented using the graph-based segmentation algorithms.

First claim

What is claimed:

1. A system to segment an optical coherence tomography (OCT) image, the system comprising one or more processors and a memory, wherein execution of instructions stored in the memory cause the one or more processors to:
retrieve, from the memory, an optical coherence tomography (OCT) image;
identify a first boundary representing an inner limiting membrane in the OCT image with a Bayesian level set based algorithm;
identify a second boundary representing a light penetrating layer in the OCT image with the Bayesian level set based algorithm;
classify a first portion of pixels of the OCT image between the first boundary and the second boundary into a retina class and a second portion of the pixels of the OCT image into a non-retina class;
calculate a weight for each of the first portion of pixels of the OCT image;
select the first portion of pixels of the OCT image as a search region;
identify a boundary of a retina tissue layer through the search region by determining a shortest path through the search region responsive to the weight of each of the pixels in the search region; and
generate a segmented OCT image based at least on the boundary of the retina tissue layer through the search region, the first boundary representing the inner limiting membrane, and the second boundary representing the light penetrating layer.

2. The system of claim 1 , wherein the shortest path through the selected portion of the retina class pixels identifies a boundary one of a subretinal hyper reflective material (SHRM) layer, a retina pigment epithelium (RPE) layer, a nerve fiber layer (NFL), or an inner-outer retina (IR/OR) interface.

3. The system of claim 1 , wherein execution of the instructions stored in the memory cause the one or more processors to classify each of the pixels with the Bayesian level set based algorithm.

4. The system of claim 1 , wherein execution of the instructions stored in the memory cause the one or more processor to iteratively:
calculate the weight for each of the first portion of pixels of the OCT image;
select a sub-portion of the retina class pixels as the search region; and
determine the shortest path through the search region responsive to the weight of each of the pixels in the search region.

5. The system of claim 4 , wherein during each iteration the selected sub-portion of the retina class pixels is a different portion of the retina class pixels.

6. The system of claim 4 , wherein the shortest path of each iteration identifies a different boundary of one of a subretinal hyper reflective material (SHRM) layer, a RPE layer, a NFL, or an IR/OR interface.

7. The system of claim 6 , wherein a first iteration identifies a boundary of the SHRM layer, a second iteration identifies a boundary of the RPE layer, a third iteration identifies a boundary of the NFL, and a fourth iteration identifies a boundary of the IR/OR interface.

8. The system of claim 1 , wherein the wherein execution of the instructions stored in the memory cause the one or more processor to preprocess the OCT image.

9. The system of claim 8 , wherein preprocessing comprises at least one of down-sampling, de-noising, or flattening the OCT image.

10. The system of claim 1 , wherein the shortest path is determined using Dijkstra's algorithm.

11. A method of segmenting an optical coherence tomography (OCT) image, the method comprising:
retrieving, by a segmentation agent executed by one or more processors coupled to a memory, an OCT image;
identifying, by the segmentation agent, a first boundary representing an inner limiting membrane in the OCT image with a Bayesian level set based algorithm;
identifying, by the segmentation agent, a second boundary representing a light penetrating layer in the OCT image with the Bayesian level set based algorithm;
classifying, by the segmentation agent, a first portion of pixels of the OCT image between the first boundary and the second boundary into a retina class and a second portion of the pixels of the OCT image into a non-retina class;
calculating, by the segmentation agent, a weight for each of the first portion of pixels of the segmented OCT image;
selecting, by the segmentation agent, the first portion of pixels of the OCT image as a search region;
identifying, by the segmentation agent, a boundary of a retina tissue layer through the search region by determining a shortest path through the search region responsive to the weight of each of the pixels in the search region; and
generating, by the segmentation agent, a segmented OCT image based at least on the boundary of the retina tissue layer through the search region, the first boundary representing the inner limiting membrane, and the second boundary representing the light penetrating layer.

12. The method of claim 11 , wherein the shortest path through the selected portion of the retina class pixels identifies a boundary of one of a subretinal hyper reflective material (SHRM) layer, a retina pigment epithelium (RPE) layer, a nerve fiber layer (NFL), or an inner-outer retina (IR/OR) interface.

13. The method of claim 11 , further comprising classifying each of the pixels with the Bayesian level set based algorithm.

14. The method of claim 11 , further comprising iteratively:
calculating, by the segmentation agent, the weight for each of the first portion of pixels of the OCT image;
selecting, by the segmentation agent, a sub-portion of the retina class pixels as the search region; and
determining, by the segmentation agent, the shortest path through the search region responsive to the weight of each of the pixels in the search region.

15. The method of claim 14 , wherein during each iteration the selected sub-portion of the retina class pixels is a different portion of the retina class pixels.

16. The method of claim 14 , wherein the shortest path of each iteration identifies a different boundary of one of a subretinal hyper reflective material (SHRM) layer, a RPE layer, a NFL, or an IR/OR interface.

17. The method of claim 16 , wherein a first iteration identifies a boundary of the SHRM layer, a second iteration identifies a boundary of the RPE layer, a third iteration identifies a boundary of the NFL, and a fourth iteration identifies a boundary of the IR/OR interface.

18. The method of claim 11 , further comprising preprocessing, by the segmentation agent, the OCT image.

19. The method of claim 18 , wherein preprocessing comprises at least one of down-sampling, de-noising, or flattening the OCT image.

20. The method of claim 11 , wherein the shortest path is determined using Dijkstra's algorithm.

Assignees

• Charles Stark Draper Laboratory Inc

Inventors

• Lowry Nathan
• Lommel Peter
• Mangoubi Rami
• Wood Richard
• Hamilton Lei
• Irvine John
• Duncan Stephen

Classifications

• G06T7/0081

  Physics · mapped topic

• G06T2207/30041

  Eye; Retina; Ophthalmic · CPC title

• G06T2207/10101

  Optical tomography; Optical coherence tomography [OCT] · CPC title

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G06T7/12 Primary

  Edge-based segmentation · CPC title