IP Library Granted Patent US 12675850
Granted Patent B2
US 12675850 · App. 17/618,190 · Granted Jul 7, 2026

System and method for removing noise and/or artifacts from an OCT image using a generative adversarial network

Inventors: John Galeotti (Pittsburgh, PA); Tejas Sudharshan Mathai (Seattle, WA); Jiahong Ouyang (Stanford, CA)
Assignee: Carnegie Mellon University
G06T5/70G06N3/045G06T7/0012G06T7/194G06V10/7747G06T2207/10101G06T2207/20081G06T2207/20084G06T2207/30041
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12675850
App. No.
17/618,190
Granted
Jul 7, 2026
Kind
B2
Abstract

Systems, methods, and computer program products are provided for removing noise and/or artifacts from an image. The method includes training a generative adversarial network (GAN) based on a plurality of images, the plurality of images comprising at least one undesired element comprising at least one of the following: noise, speckle patterns, artifacts, or any combination thereof, and generating a modified image based on processing an image of an eye or other object with the GAN to remove the at least one undesired element from the image that is above an outer surface of the eye or other object.

Claims (26)

1 . A method for removing noise and/or artifacts from an image, comprising:

training a generative adversarial network (GAN) based on a plurality of medical images, the plurality of medical images comprising at least one undesired element comprising at least one of the following: noise, speckle patterns, artifacts, or any combination thereof;

generating a pre-segmented medical image by inputting a medical image into the GAN to pre-segment an outer layer of an object in the medical image, the pre-segmented medical image removing or labeling at least one undesired element prior to the shallowest tissue interface of the object; and

generating a segmented medical image based on processing the pre-segmented medical image output by the GAN and the medical image with a segmentation network, the segmentation network configured to segment at least one object in the medical image based on the pre-segmented medical image, wherein the segmentation network is a different network than the GAN.

2 . The method of claim 1 , wherein generating the segmented medical image comprises identifying a plurality of background pixels corresponding to the at least one undesired element.

3 . The method of claim 1 , wherein the GAN is configured to assign different weights to foreground pixels and background pixels.

4 . The method of claim 3 , wherein the background pixels are weighed more than the foreground pixels.

5 . The method of claim 1 , wherein the plurality of medical images comprises a plurality of Optical Coherence Tomography (OCT) images.

6 . The method of claim 5 , wherein the plurality of OCT images is from a plurality of different OCT imaging systems.

7 . The method of claim 1 , wherein the at least one undesired element is beneath an outer surface of an eye, the outer surface comprising the shallowest tissue interface.

8 . The method of claim 1 , further comprising processing the pre-segmented medical image with at least one segmentation algorithm.

9 . A system for removing noise and/or artifacts from an image, comprising a computing device programmed or configured to:

train a generative adversarial network (GAN) based on a plurality of medical images, the plurality of medical images comprising at least one undesired element comprising at least one of the following: noise, speckle patterns, artifacts, or any combination thereof;

generate a pre-segmented medical image by inputting a medical image into the GAN to pre-segment an outer layer of an object in the medical image, the pre-segmented medical image removing or labeling at least one undesired element prior to the shallowest tissue interface of the object; and

generate a segmented medical image based on processing the pre-segmented medical image and the medical image with a segmentation network, the segmentation network configured to segment at least one object in the medical image, wherein the segmentation network is a different network than the GAN.

10 . The system of claim 9 , wherein generating the segmented medical image comprises identifying a plurality of background pixels corresponding to the at least one undesired element.

11 . The system of claim 9 , wherein the GAN is configured to assign different weights to foreground pixels and background pixels.

12 . The system of claim 11 , wherein the background pixels are weighed more than the foreground pixels.

13 . The system of claim 9 , wherein the plurality of medical images comprises a plurality of Optical Coherence Tomography (OCT) images.

14 . The system of claim 13 , wherein the plurality of OCT images is from a plurality of different OCT imaging systems.

15 . The system of claim 9 , wherein the at least one undesired element is beneath an outer surface of an eye, the outer surface comprising the shallowest tissue interface.

16 . The system of claim 9 , wherein the computing device is further programmed or configured to process the pre-segmented medical image with at least one segmentation algorithm.

17 . A computer program product for removing noise and/or artifacts from an image, comprising at least one non-transitory computer-readable medium comprising program instructions that, when executed by a computing device, cause the computing device to:

train a generative adversarial network (GAN) based on a plurality of medical images, the plurality of medical images comprising at least one undesired element comprising at least one of the following: noise, speckle patterns, artifacts, or any combination thereof;

generate a pre-segmented medical image by inputting a medical image into the GAN to pre-segment an outer layer of an object in the medical image, the pre-segmented medical image removing or labeling at least one undesired element prior to the shallowest tissue interface of the object; and

generate a segmented medical image based on processing the pre-segmented medical image and the medical image with a segmentation network, the segmentation network configured to segment at least one object in the medical image, wherein the segmentation network is a different network than the GAN.