Systems and methods for training a model to predict survival time for a patient
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-10650929-B1
Publication Date
2020-05-12
Expiration Date
Abstract
In some aspects, the described systems and methods provide for a method for training a model to predict survival time for a patient. The method includes accessing annotated pathology images associated with a first group of patients in a clinical trial. Each of the annotated pathology images is associated with survival data for a respective patient. Each of the annotated pathology images includes an annotation describing a tissue characteristic category for a portion of the image. Values for one or more features are extracted from each of the annotated pathology images. A model is trained based on the survival data and the extracted values for the features. The trained model is stored on a storage device.
Core Innovation
The invention provides a method for training a model to predict survival time for a patient using annotated pathology images. The method accesses annotated pathology images associated with patients in a randomized controlled clinical trial, where each image includes at least one annotation describing a tissue characteristic category for a portion of the image and is associated with survival data for a respective patient. The method extracts values for a plurality of features from the annotated pathology images and trains a model based on the survival data and the extracted feature values.
After training, the method applies the trained model to process the feature values to predict survival data for patients in the experimental treatment group and separately for patients in the control treatment group of the randomized controlled clinical trial. The method then determines a first prognostic performance of the trained model for the experimental treatment group using the predicted survival data and respective survival data, and determines a second prognostic performance for the control treatment group using predicted survival data and respective survival data.
To quantify clinically meaningful prognostic power, the method determines a specificity of the prognostic power by comparing the first prognostic performance with the second prognostic performance. The specificity includes a likelihood that the model will correctly identify a subset of patients that benefit from experimental treatment.
Claims Coverage
The independent claims cover a training and evaluation framework for survival prediction from annotated pathology images in a randomized controlled clinical trial, with separate prognostic performance for experimental and control treatment groups and a specificity measure tied to correct identification of a subset expected to benefit. Across the independent claims, the inventive features are the survival-prediction training, group-wise prognostic evaluation, and specificity determination based on comparative performance between experimental and control groups.
Survival-time model training from annotated pathology images in a randomized controlled clinical trial
accessing a first plurality of annotated pathology images associated with a first group of patients in a randomized controlled clinical trial, wherein each of the first plurality of annotated pathology images is associated with survival data for a respective patient, wherein each of the first plurality of annotated pathology images includes at least one annotation describing a tissue characteristic category for a portion of the image; extracting a first plurality of values for a plurality of features from each of the first plurality of annotated pathology images; training a model based on the survival data and the first plurality of values for the plurality of features
Group-wise survival prediction for experimental and control treatment groups
processing, using the trained model, the first plurality of values for the plurality of features, to predict survival data for patients in the first group of patients, wherein the first group of patients belongs to an experimental treatment group of the randomized controlled clinical trial; processing, using the trained model, a second plurality of values for the plurality of features extracted from a second plurality of annotated pathology images associated with a second group of patients, to predict survival data for patients in the second group of patients, wherein the second group of patients belongs to a control treatment group of the randomized controlled clinical trial
Separate prognostic performance determination for experimental and control groups and specificity based on comparative performance
determining a first prognostic performance of the trained model for the experimental treatment group based on the predicted survival data for the patients in the first group of patients and respective survival data; determining a second prognostic performance of the trained model for the control treatment group based on predicted survival data for the patients in the second group of patients and respective survival data; and determining a specificity of a prognostic power of the trained model by comparing the first prognostic performance of the trained model for the experimental treatment group and the second prognostic performance of the trained model for the control treatment group, wherein the specificity of the prognostic power of the trained model includes a likelihood that the model will correctly identify of a subset of patients that benefit from experimental treatment
The independent claims jointly define: (i) training a model for survival-time prediction from annotated pathology images with tissue characteristic category annotations and extracted feature values, (ii) predicting survival separately for experimental-treatment and control-treatment groups in a randomized controlled clinical trial, and (iii) determining prognostic performance for each group and computing a specificity of prognostic power via comparison, including a likelihood of correctly identifying a subset of patients expected to benefit from experimental treatment.
Stated Advantages
Documented Applications
No documented applications found
Interested in licensing this patent?