Granted Patent B2

US 12,614,378 · App. 17/852,884 · Granted Apr 28, 2026

Inventors: Eric Robert (Quebec City, CA); George Shaikovski (London, GB); Christopher Kanan (Pittsford, NY)

Assignee: Paige.AI, Inc.

G06V10/774G06V10/469

Abstract

Claims (71)

1 . A computer-implemented method for processing one or more digital whole slide images (WSIs), the method comprising:

receiving, by a trained artificial intelligence (AI) system at an electronic storage of the AI system, one or more digital WSIs associated with a tissue specimen;

breaking, by the trained AI system, the one or more digital WSIs into a plurality of tile images;

determining, by the trained AI system, one or more foreground tiles of the plurality of tile images;

extracting, by the trained AI system, a plurality of tile feature vectors for the one or more foreground tiles;

clustering, by the trained AI system, the plurality of tile feature vectors into one or more clusters;

determining, by the trained AI system, cluster membership for the one or more clusters based on clustering information determined during training of the AI system;

combining, by the trained AI system, tile feature vectors in associated clusters sharing cluster membership to generate one or more vectors of features representing a slide signature; and

generating, using the trained AI system, an inference of one or more quality issues present in the one or more digital WSIs based on the one or more vectors of features representing the slide signature.

2 . The computer-implemented method of claim 1 , wherein determining the one or more foreground tiles comprises receiving a selection of the one or more foreground tiles.

3 . The computer-implemented method of claim 1 , further comprising:

removing one or more background tiles from the plurality of tile images prior to extracting the plurality of tile feature vectors from the one or more foreground tiles.

4 . The computer-implemented method of claim 1 , further comprising:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

predicting a type of the one or more quality issues after determining that the one or more quality issues are present.

5 . The computer-implemented method of claim 1 , further comprising:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

producing a visualization of a result of determining whether the one or more quality issues are present.

6 . A computer system for processing one or more digital whole slide images (WSIs), the computer system comprising:

at least one memory storing instructions; and

at least one processor configured to execute the instructions to perform operations comprising:

receiving, by a trained artificial intelligence (AI) system at an electronic storage of the AI system, one or more digital WSIs associated with a tissue specimen;

breaking, by the trained AI system, the one or more digital WSIs into a plurality of tile images;

determining, by the trained AI system, one or more foreground tiles of the plurality of tile images;

extracting, by the trained AI system, a plurality of tile feature vectors for the one or more foreground tiles;

clustering, by the trained AI system, the plurality of tile feature vectors into one or more clusters;

determining, by the trained AI system, cluster membership for the one or more clusters based on clustering information determined during training of the AI system;

combining, by the trained AI system, tile feature vectors in associated clusters sharing cluster membership to generate one or more vectors of features representing a slide signature; and

generating, using the trained AI system, an inference of one or more quality issues present in the one or more digital WSIs based on the one or more vectors of features representing the slide signature.

7 . The computer system of claim 6 , wherein determining the one or more foreground tiles comprises receiving a selection of the one or more foreground tiles prior to extracting the one or more vectors of features from the one or more foreground tiles.

8 . The computer system of claim 6 , wherein the operations further comprise:

removing one or more background tiles from the plurality of tile images prior to extracting the plurality of tile feature vectors from the one or more foreground tiles.

9 . The computer system of claim 6 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

predicting a type of the one or more quality issues after determining that the one or more quality issues are present.

10 . The computer system of claim 6 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

producing a visualization of a result of determining whether the one or more quality issues are present.

11 . A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform operations for processing one or more digital whole slide images (WSIs), the operations comprising:

receiving, by a trained artificial intelligence (AI) system at an electronic storage of the AI system, one or more digital WSIs associated with a tissue specimen;

breaking, by the trained AI system, the one or more digital WSIs into a plurality of tile images;

determining, by the trained AI system, one or more foreground tiles of the plurality of tile images;

extracting, by the trained AI system, a plurality of tile feature vectors for the one or more foreground tiles;

clustering, by the trained AI system, the plurality of tile feature vectors into one or more clusters;

determining, by the trained AI system, cluster membership for the one or more clusters based on clustering information determined during training of the AI system;

combining, by the trained AI system, tile feature vectors in associated clusters sharing cluster membership to generate one or more vectors of features representing a slide signature; and

generating, using the trained AI system, an inference of one or more quality issues present in the one or more digital WSIs based on the one or more vectors of features representing the slide signature.

12 . The non-transitory computer-readable medium of claim 11 , wherein determining the one or more foreground tiles comprises receiving a selection of the one or more foreground tiles prior to extracting the one or more vectors of features from the one or more foreground tiles.

13 . The non-transitory computer-readable medium of claim 11 , wherein the operations further comprise:

removing one or more background tiles from the plurality of tile images prior to extracting the plurality of tile feature vectors from the one or more foreground tiles.

14 . The non-transitory computer-readable medium of claim 11 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

predicting a type of the one or more quality issues after determining that the one or more quality issues are present.

15 . The non-transitory computer-readable medium of claim 11 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

producing a visualization of a result of determining whether the one or more quality issues are present.

16 . The computer-implemented method of claim 1 , further comprising:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

generating and outputting a report indicating a result of determining whether the one or more quality issues are present.

17 . The computer system of claim 6 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

generating and outputting a report indicating a result of determining whether the one or more quality issues are present.

18 . The non-transitory computer-readable medium of claim 11 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

generating and outputting a report indicating a result of determining whether the one or more quality issues are present.

19 . The computer-implemented method of claim 1 , further comprising:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

storing, by the trained AI system to the electronic storage, a result of determining whether the one or more quality issues are present.

20 . The computer system of claim 6 , wherein the operations further comprise:

determining, based on the inference of the trained AI system, whether one or more quality issues are present in the one or more digital WSIs; and

storing, by the trained AI system to the electronic storage, a result of determining whether the one or more quality issues are present.

Assignments (3)

Continuity (2)

References Cited (13)