Docket #: S19-376

Clinical Evaluation of Prostate Cancer using Machine Learning-Based Pathology Report Generation

A machine learning-based framework for summarizing prostate cancer and related findings through a pathology report generator. This report generator reduces time-consuming annotations and tumor volume estimation during clinical routines by extracting and summarizing relevant information into a pathology report to aid pathologist workflow. The framework can also determine subsets with increased risk for genomic alterations, high-risk of biochemical recurrence, and cancer-specific survival from a set of histology images.

Stage of Research

Prototype

Demonstrated for clinical routine, document archiving, and research

Related Technology:

Stanford Docket S19-308 -"Plexus Convolutional Neural Network for Histologic Imaging Analysis with Smaller Training Datasets and Parameters"

Applications

Pathology report generator for clinical routine
Treatment success rate analysis and quality control

Advantages

Prepared pathology reports for clinical use

Tumor volume calculation & lesion-based summarization using 1/18-1/54 of data volume
Automated tumor lesion annotations

Reports are portable and easily shared
Extracted features can be used for prognostics development of molecular biological profiles

Publications

Eminaga et al. arXiv preprint: "Plexus Convolutional Neural Network (PlexusNet): A novel neural network architecture for histologic image analysis"