Artificial intelligence prediction of prostate cancer outcomes

Inventors

MERMEL, Craig • Liu, Yun • Manoj, Naren • Symonds, Matthew • Stumpe, Martin • Peng, Lily • Nagpal, Kunal • Wulczyn, Ellery • Foote, Davis • Steiner, David F. • Chen, Po-Hsuan Cameron

Assignees

Verily Life Sciences LLC

Abstract

One example method for AI prediction of prostate cancer outcomes involves receiving an image of prostate tissue; assigning Gleason pattern values to one or more regions within the image using an artificial intelligence Gleason grading model, the model trained to identify Gleason patterns on a patch-by-patch basis in a prostate tissue image; determining relative areal proportions of the Gleason patterns within the image; assigning at least one of a risk score or risk group value to the image based on the determined relative areal proportions; and outputting at least one of the risk score or the risk group value.

Core Innovation

The invention relates to an artificial intelligence Gleason grading model that assesses a prognosis of a prostate cancer patient from an image of prostate tissue. The model assigns Gleason pattern values to one or more regions within the image in a patch-by-patch manner, and the Gleason pattern values are used to determine relative areal proportions of Gleason patterns within the image.

Based on the determined relative areal proportions, a continuous risk score is generated for the entire image. The continuous risk score is assigned as at least one of the output values for prognosis of a prostate cancer patient, and a risk group value is assigned based on the continuous risk score. At least one of the continuous risk score or the risk group value is output.

The system generates the continuous image-level risk score using a Cox proportional hazards regression model trained with survival data, and the risk score is discretized into AI risk groups. The disclosed approach aligns the AI risk group distributions with prognostic Gleason Grade Group distributions and reports prognostic performance using Harrell’s C-index and Kaplan–Meier stratification.

Claims Coverage

The independent claims are directed to a computer-implemented method, a system, and a non-transitory computer-readable medium that implement the same core pipeline. Across the independent claims, the inventive features include patch-by-patch AI Gleason pattern value assignment, computation of relative areal proportions, generation of a continuous image-level risk score, and assignment/output of a risk group value based on the continuous risk score.

Across the independent claims, the document covers the disclosed prognosis assessment pipeline that converts patch-by-patch AI Gleason pattern assignments into relative areal proportions, then into a continuous risk score for the entire image, followed by risk group assignment and output.

Stated Advantages

Documented Applications