Systems and methods for training a statistical model to predict tissue characteristics for a pathology image
Inventors
Beck, Andrew H. • Khosla, Aditya
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11756198-B1
Publication Date
2023-09-12
Expiration Date
Abstract
In some aspects, the described systems and methods provide for a method for training a statistical model to predict tissue characteristics for a pathology image. The method includes accessing annotated pathology images. Each of the images includes an annotation describing a tissue characteristic category for a portion of the image. A set of training patches and a corresponding set of annotations are defined using an annotated pathology image. Each of the training patches in the set includes values obtained from a respective subset of pixels in the annotated pathology image and is associated with a corresponding patch annotation determined based on an annotation associated with the respective subset of pixels. The statistical model is trained based on the set of training patches and the corresponding set of patch annotations. The trained statistical model is stored on at least one storage device.
Core Innovation
The invention predicts an entity of interest for a pathology image by receiving an annotated pathology image including one or more annotations for the pathology image, extracting values for one or more features from the annotated pathology image, and retrieving a machine learning model from at least one storage device. The extracted feature values are processed using the machine learning model to predict an entity of interest selected from survival time, drug response, patient level phenotype/molecular characteristics, and patient clinical outcomes, and the predicted entity of interest is stored on the at least one storage device.
The predictive modeling is grounded in feature values extracted from annotated pathology images, and the feature values are derived from clinical trial-related contexts and additional data sources. The machine learning model is trained using extracted feature values from annotated training pathology images collected from patients in a randomized controlled clinical trial, with optional clinical metadata values, and is further described with genomic data, transcriptomic data, and/or protein expression data, with output constrained to survival time.
The feature values include morphologic and quantitative histology metrics and spatial distance features. Feature examples include areas of epithelium, stroma, necrosis, cancer cells, macrophages, and lymphocytes; number of mitotic figures; average nuclear grade and standard deviation of nuclear grade; and spatial distances between defined cell types and structures. The document also describes using an annotated pathology image that includes a heat map and extracting feature values by applying one or more functions to the heat map.
Claims Coverage
The independent claims cover three forms of the same inventive concept: a method, a system, and a non-transitory computer-readable medium for predicting an entity of interest from an annotated pathology image using extracted feature values and a retrieved trained machine learning model, with storage of the predicted entity. The inventive features include survival time, drug response, patient level phenotype/molecular characteristics, and patient clinical outcomes, together with training-context and feature-extraction limitations.
Predicting an entity of interest from an annotated pathology image using a retrieved machine learning model
Receiving an annotated pathology image including one or more annotations for the pathology image; extracting values for one or more features from the annotated pathology image; retrieving a machine learning model from at least one storage device; processing, using the machine learning model, the values for the one or more features extracted from the annotated pathology image to predict an entity of interest selected from survival time, drug response, patient level phenotype/molecular characteristics, and patient clinical outcomes; and storing the predicted entity of interest on the at least one storage device.
Predicting an entity of interest for a pathology image using a system architecture with a retrieved trained machine learning model
A system comprising at least one computer hardware processor and at least one non-transitory computer-readable storage medium storing processor-executable instructions that cause the processor to perform receiving an annotated pathology image including one or more annotations for the pathology image; extracting values for one or more features from the annotated pathology image; retrieving a trained machine learning model from at least one storage device; processing, using the trained machine learning model, the values for the one or more features extracted from the annotated pathology image to predict an entity of interest selected from survival time, drug response, patient level phenotype/molecular characteristics, and patient clinical outcomes; and storing the predicted entity of interest on the at least one storage device.
Using a non-transitory computer-readable medium to execute predicting an entity of interest from annotated pathology image features
A non-transitory computer-readable medium containing instructions that, when executed, cause at least one computer hardware processor to perform receiving an annotated pathology image including one or more annotations for the pathology image; extracting values for one or more features from the annotated pathology image; retrieving a trained machine learning model from at least one storage device; processing, using the trained machine learning model, the values for the one or more features extracted from the annotated pathology image to predict an entity of interest selected from survival time, drug response, patient level phenotype/molecular characteristics, and patient clinical outcomes; and storing the predicted entity of interest on the at least one storage device.
The inventive coverage is directed to predicting survival time, drug response, patient level phenotype/molecular characteristics, or patient clinical outcomes from annotated pathology images by extracting feature values and applying a retrieved trained machine learning model, with the predicted outcome stored. Dependent claims further narrow training and outputs, including randomized controlled clinical trial training context, optional clinical metadata, genomic/transcriptomic/protein expression inputs for survival time, and morphologic, spatial, and heat-map-based feature extraction.
Stated Advantages
Documented Applications
Predicting survival time, drug response, patient level phenotype/molecular characteristics, and patient clinical outcomes from pathology images.
Survival analysis in clinical trial experimental versus control groups as described in the document’s discussion of subset survival analyses.
Using feature values extracted from fully annotated images to train downstream prognostic models for predicting survival, drug response, and patient outcomes as described in the document.
Interested in licensing this patent?