Assessing risk of breast cancer recurrence

Patent metadata

Field	Value
Publication number	US-10489904-B2
Application number	US-201615374998-A
Country	US
Kind code	B2
Filing date	Dec 9, 2016
Priority date	Jun 10, 2014
Publication date	Nov 26, 2019
Grant date	Nov 26, 2019

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. An image processing method for analyzing at least first and second images obtained from an early stage breast cancer biopsy tissue sample, the tissue sample being ER-positive and HER2-negative, the tissue sample being marked by multiple stains for identification of biological features, the method comprising:
acquiring a predetermined plurality of prognostic marker-specific biological features by performing an image analysis on the at least first and second images individually, wherein each of the acquired prognostic marker-specific biological features has at least one numerical value assigned thereto that is descriptive of the biological feature, and at least one of the acquired marker-specific biological features includes 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 nearest neighbor feature, a center vote strength, an annular region based feature, or a blob topography feature;
computing a predefined plurality of prognostic inter-marker features by calculating combinations of numerical values of predefined pairs of the marker-specific biological features acquired from different ones of the at least first and second images; and
entering (i) the plurality of acquired prognostic marker-specific biological features and (ii) the plurality of prognostic inter-marker features into a trained predictor, wherein the trained predictor correlates (i) and (ii) with breast cancer recurrence data and cross-validates performance of the trained predictor;
wherein the predictor outputs a first signal or a second signal in response to the entry of the plurality of prognostic marker-specific biological features and prognostic inter-marker features, the first signal being indicative of a requirement for adjuvant chemotherapy and the second signal being indicative of the adjuvant chemotherapy being unnecessary.

2. The method of claim 1 , the calculation of at least some of the combinations being performed by multiplying the numerical value of the acquired biological feature of one of the at least first and second image by the numerical value of the acquired biological feature of the other of the at least first and second image.

3. The method of claim 2 , wherein the predetermined plurality of prognostic marker-specific biological features and the predefined plurality of prognostic inter-marker features are determined by L1-regularized logistic regression, and the trained predictor is obtained by L1-regularized logistic regression.

4. The method of claim 1 , the calculation of at least some of the combinations being performed by calculating a correlation coefficient indicative of a spatial correlation of the marker-specific biological features in each pair of the predefined pairs.

5. The method of claim 1 , further comprising acquiring a multi-channel image from the tissue sample being stained by multiple stains and obtaining the at least first and second images by unmixing the multi-channel image.

6. The method of claim 1 , wherein the at least first and second images are obtained by slicing the tissue sample into at least first and second tissue slices and individually staining the first and second tissue slices for acquisition of the respective at least first and second images.

7. The method of claim 6 , the at least first and second slices being adjacent slices and having a thickness of between 2 μm to 10 μm.

8. The method of claim 6 wherein the at least first and second slices are adjacent slices having a thickness of 6 μm.

Assignees

• Ventana Med Syst Inc
• Univ Leland Stanford Junior

Inventors

• Barnes Michael
• Chukka Srinivas
• Knowles David

Classifications

• G06T2207/10064

  Fluorescence image · CPC title

• G06T2207/30072

  Microarray; Biochip, DNA array; Well plate · CPC title

• G06T2207/10056

  Microscopic image · CPC title

• G06T2207/30096

  Tumor; Lesion · CPC title

• G06T2207/10024

  Color image · CPC title