Assessing risk of breast cancer recurrence

Patent metadata

Field	Value
Publication number	US-12100146-B2
Application number	US-202217685115-A
Country	US
Kind code	B2
Filing date	Mar 2, 2022
Priority date	Jun 10, 2014
Publication date	Sep 24, 2024
Grant date	Sep 24, 2024

Abstract

The subject disclosure presents systems and computer-implemented methods for assessing a risk of cancer recurrence in a patient based on a holistic integration of large amounts of prognostic information for said patient into a single comparative prognostic dataset. A risk classification system may be trained using the large amounts of information from a cohort of training slides from several patients, along with survival data for said patients. For example, a machine-learning-based binary classifier in the risk classification system may be trained using a set of granular image features computed from a plurality of slides corresponding to several cancer patients whose survival information is known and input into the system. The trained classifier may be used to classify image features from one or more test patients into a low-risk or high-risk group.

First claim

The invention claimed is:

1. A method comprising:
accessing a first image depicting at least part of a training tissue sample stained with a first type of stain;
accessing a second image depicting at least part of the training tissue sample stained with a second type of stain;
identifying, for the first image, a first plurality of marker-specific features;
identifying, for the second image, a second plurality of marker-specific features, wherein each marker-specific feature of the first or second plurality of marker-specific features includes a value that identifies one or more characteristics of a biological feature of the training tissue sample;
applying a machine-learning model to one or more marker-specific features of the first plurality of marker-specific features and one or more marker-specific features of the second plurality of marker-specific features to determine a set of prognostic marker-specific features of the training tissue sample;
computing a plurality of inter-marker features of the first and second images, wherein each inter-marker feature of the plurality of inter-marker features is computed based on a first marker-specific feature of the first plurality of marker-specific features and a second marker-specific feature of the second plurality of marker-specific features;
applying the machine-learning model to one or more inter-marker features of the plurality of inter-marker features to determine a set of prognostic inter-marker features of the training tissue sample;
training a breast-cancer recurrence model using the set of prognostic marker-specific features and the set of prognostic inter-marker features, wherein the breast-cancer recurrence model is trained to process another image to predict a breast-cancer recurrence risk level for a particular subject; and
outputting the trained breast-cancer recurrence model.

2. The method of claim 1 , wherein the set of prognostic inter-marker features and/or the set of prognostic marker-specific features are determined further based on applying the machine-learning model to survival data, wherein the survival data include breast-cancer recurrence data of a subject from which the training tissue sample has been obtained.

3. The method of claim 1 , wherein the machine-learning model includes an L1-regularized logistic regression algorithm.

4. The method of claim 1 , wherein the first plurality of marker-specific features include a nucleus blob shape feature, a nucleus blob area feature, a nucleus blob compactness feature, a nucleus blob density feature, a normalized color feature, an absolute color feature, a center vote strength, an annular region based feature, or a blob topography feature.

5. The method of claim 1 , wherein the first type of stain includes a hematoxylin-and-eosin (H&E) stain.

6. The method of claim 1 , wherein the second type of stain includes a multiplex immunohistochemical (IHC) stain.

7. The method of claim 1 , wherein:
a first number of the first plurality of marker-specific features is less than 500; and
a second number of the set of prognostic marker-specific features is less than 50, wherein the second number is less than the first number.

8. The method of claim 1 , wherein a number of the plurality of inter-marker features is greater than 500.

9. The method of claim 8 , wherein the number of the plurality of inter-marker features is less than 2000.

10. A system comprising:
one or more processors; and
a memory coupled to the processor, the memory storing computer readable instructions that, when executed by the processor, cause the one or more processors to perform operations comprising:
accessing a first image depicting at least part of a training tissue sample stained with a first type of stain;
accessing a second image depicting at least part of the training tissue sample stained with a second type of stain;
identifying, for the first image, a first plurality of marker-specific features;
identifying, for the second image, a second plurality of marker-specific features, wherein each marker-specific feature of the first or second plurality of marker-specific features includes a value that identifies one or more characteristics of a biological feature of the training tissue sample;
applying a machine-learning model to one or more marker-specific features of the first plurality of marker-specific features and one or more marker-specific features of the second plurality of marker-specific features to determine a set of prognostic marker-specific features of the training tissue sample;
computing a plurality of inter-marker features of the first and second images, wherein each inter-marker feature of the plurality of inter-marker features is computed based on a first marker-specific feature of the first plurality of marker-specific features and a second marker-specific feature of the second plurality of marker-specific features;
applying the machine-learning model to one or more inter-marker features of the plurality of inter-marker features to determine a set of prognostic inter-marker features of the training tissue sample;
training a breast-cancer recurrence model using the set of prognostic marker-specific features and the set of prognostic inter-marker features, wherein the breast-cancer recurrence model is trained to process another image to predict a breast-cancer recurrence risk level for a particular subject; and
outputting the trained breast-cancer recurrence model.

11. The system of claim 10 , wherein the set of prognostic inter-marker features and/or the set of prognostic marker-specific features are determined further based on applying the machine-learning model to survival data, wherein the survival data include breast-cancer recurrence data of a subject from which the training tissue sample has been obtained.

12. The system of claim 10 , wherein the machine-learning model includes an L1-regularized logistic regression algorithm.

13. The system of claim 10 , wherein the first plurality of marker-specific features include a nucleus blob shape feature, a nucleus blob area feature, a nucleus blob compactness feature, a nucleus blob density feature, a normalized color feature, an absolute color feature, a center vote strength, an annular region based feature, or a blob topography feature.

14. The system of claim 10 , wherein the first type of stain includes a hematoxylin-and-eosin (H&E) stain.

15. The system of claim 10 , wherein the second type of stain includes a multiplex immunohistochemical (IHC) stain.

16. A computer program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform operations comprising:
accessing a first image depicting at least part of a training tissue sample stained with a first type of stain;
accessing a second image depicting at least part of the training tissue sample stained with a second type of stain;
identifying, for the first image, a first plurality of marker-specific features;
identifying, for the second image, a second plurality of marker-specific features, wherein each marker-specific feature of the first or second plurality of marker-specific features includes a value that identifies one or more characteristics of a biological feature of the training tissue sample;
applying a machine-learning model to one or more marker-specific features of the first plurality of marker-specific features and one or more marker-specific features of the second plurality of marker-specific features to determine a set of prognostic marker-specific features of

Assignees

• Ventana Med Syst Inc
• Univ Leland Stanford Junior

Inventors

• Barnes Michael
• Chukka Srinivas
• Knowles David

Classifications

• G06V10/806

  of extracted features · CPC title

• G06V10/771

  Feature selection, e.g. selecting representative features from a multi-dimensional feature space · CPC title

• G06F18/253

  of extracted features · CPC title

• G06F18/211

  Selection of the most significant subset of features · CPC title

• G06V20/698

  Matching; Classification · CPC title