Systems and methods for utilizing synthetic medical images generated using a neural network

Patent metadata

Field	Value
Publication number	US-12315149-B2
Application number	US-202117793201-A
Country	US
Kind code	B2
Filing date	Jan 19, 2021
Priority date	Jan 17, 2020
Publication date	May 27, 2025
Grant date	May 27, 2025

Abstract

A system for completing a medical image having at least one obscured region includes an input for receiving a first classification map generated using an acquired optical coherence tomography (OCT) image having at least one obscured region, the acquired OCT image acquired using an imaging system and a pre-processing module coupled to the input and configured to create an obscured region mask. The pre-processing module also generates a second classification map that has the at least one obscured region filled in. The system also includes a generative network coupled to the pre-processing module and configured to generate a synthetic OCT image based on the second classification map and a post-processing module coupled to the generative network. The post-processing module is configured to receive the synthetic OCT image and the acquired OCT image and to generate a completed image based on the synthetic OCT image and the acquired OCT image.

First claim

The invention claimed is:

1. A system for completing a medical image having at least one obscured region, the system comprising:
an input for receiving a first classification map generated using an acquired optical coherence tomography (OCT) image having at least one obscured region, the acquired OCT image acquired using an imaging system;
a pre-processing module coupled to the input and configured to create an obscured region mask and to generate a second classification map that has the at least one obscured region filled in;
a generative network coupled to the pre-processing module and configured to generate a synthetic OCT image based on the second classification map; and
a post-processing module coupled to the generative network and configured to receive the synthetic OCT image and the acquired OCT image and to generate a completed image based on the synthetic OCT image and the acquired OCT image.

2. The system according to claim 1 , further comprising a memory coupled to the post-processing module for storing the completed image.

3. The system according to claim 1 , further comprising a display coupled to the post-processing module and configured to display the completed image.

4. The system according to claim 1 , wherein the obscured region mask is created based on the classification map or the acquired OCT image.

5. The system according to claim 1 , wherein the completed image is generated by replacing obscured pixels in the acquired OCT image with corresponding pixels in the synthetic OCT image.

6. The system according to claim 1 , wherein the generative network is trained using a conditional generative adversarial network.

7. The system according to claim 1 , wherein the classification map is generated by identifying a wall area of a vessel in the acquired OCT image and classifying at least one type of tissue in the wall area using a convolution neural network.

8. The system according to claim 7 , wherein the at least one tissue type is one of calcium, lipid tissue, fibrous tissue, mixed tissue, non-pathological tissue or media, and no visible tissue.

9. The system according to claim 1 , wherein the at least one obscured region is filled by determining the expected, likely or nominal classifications of a plurality of pixels in the obscured region.

10. A method for completing a medical image having at least one obscured region, the method comprising:
receiving a first classification map generated using an acquired optical coherence tomography (OCT) image having at least one obscured region, the acquired OCT image acquired using an imaging system;
creating an obscured region mask;
generating a second classification map that has the at least one obscured region filled in;
generating a synthetic OCT image based on the second classification map using a generative network;
generating a completed image based on the synthetic OCT image and the acquired OCT image; and
displaying the completed image on a display or storing the completed image in a memory.

11. The method according to claim 10 , wherein the obscured region mask is created based on the classification map or the acquired OCT image.

12. The method according to claim 10 , wherein generating the completed image includes replacing obscured pixels in the acquired OCT image with corresponding pixels in the synthetic OCT image.

13. The method according to claim 10 , wherein the generative network is trained using a conditional generative adversarial network.

14. The method according to claim 10 , wherein the first classification map is generated by identifying a wall area of a vessel in the acquired OCT image and classifying at least one type of tissue in the wall area using a convolution neural network.

15. The method according to claim 14 , wherein the at least one tissue type is one of calcium, lipid tissue, fibrous tissue, mixed tissue, non-pathological tissue or media, and no visible tissue.

16. The method according to claim 10 , wherein filling the obscured region mask of the classification map includes determining the expected, likely or nominal classifications of a plurality of pixels in the obscured region.

Assignees

• Massachusetts Inst Technology

Inventors

• Olender Max Louis
• Edelman Elazer R

Classifications

• G06T12/10

  Image preprocessing, e.g. calibration, positioning of sources or scatter correction · CPC title

• G06T5/77

  Retouching; Inpainting; Scratch removal · CPC title

• G06T2207/20221

  Image fusion; Image merging · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2207/20081

  Training; Learning · CPC title