Opening The Ai Box: Can Deep Learning Predict Cancer Recurrence?

Researchers from the RIKEN Center for Advanced Intelligence Project (AIP) in Japan have developed an artificial intelligence (AI) framework capable of extracting interpretable features from annotation-free histopathology images from prostate cancer patients. This AI-generated features deliver high accuracy and outperformed the prediction of biochemical recurrence using conventional, Gleason Score-based methods. The AI framework is capable of acquiring interpretable features from whole-mount pathology images acquired at three different centres, including images from 842 patients at Nippon Medical School Hospital (NMSH), plus 95 patients from St. Marianna University hospital (SMH) and Aichi Medical University (AMU).