A MULTIPLE INSTANCE LEARNING FRAMEWORK

Summary from the TLDR AI newsletter

A new multiple instance learning framework for whole slide image classification introduces a dual-scale vision-language approach, using a prototype-guided patch decoder and a context-guided text decoder to enhance model performance on pathology tasks.