Shedding Light on the Black Box of a Neural Network Used to Detect Prostate Cancer in Whole Slide Images by Occlusion-Based Explainability
V. Krajnansky, M. Gallo, R. Nenutil, M. Nemecek, P. Holub, T. Brazdil
This paper presents a deep learning system for prostate cancer detection in whole slide images and uses occlusion-based explainability to identify key histomorphological features used for classification.
bioRxiv · Mar 31, 2022 · PreprintPathology