LadderMIL: Multiple Instance Learning with Coarse-to-Fine Self-Distillation

TL;DR

LadderMIL introduces a novel framework for multiple instance learning in pathology that combines self-distillation for instance supervision and contextual encoding for inter-instance relationships, significantly improving performance on various clinical tasks.

Contribution

The paper proposes LadderMIL, a new MIL framework that employs a Coarse-to-Fine Self-Distillation paradigm and a Contextual Encoding Generator to enhance instance-level learning and inter-instance context understanding.

Findings

Achieved average improvements of 8.1% in AUC across benchmarks.

Demonstrated 11% average increase in F1-score.

Showed 2.4% improvement in C-index for prognosis prediction.

Abstract

Multiple Instance Learning (MIL) for whole slide image (WSI) analysis in computational pathology often neglects instance-level learning as supervision is typically provided only at the bag level, hindering the integrated consideration of instance and bag-level information during the analysis. In this work, we present LadderMIL, a framework designed to improve MIL through two perspectives: (1) employing instance-level supervision and (2) learning inter-instance contextual information at bag level. Firstly, we propose a novel Coarse-to-Fine Self-Distillation (CFSD) paradigm that probes and distils a network trained with bag-level information to adaptively obtain instance-level labels which could effectively provide the instance-level supervision for the same network in a self-improving way. Secondly, to capture inter-instance contextual information in WSI, we propose a Contextual Encoding Generator (CEG), which encodes the contextual appearance of instances within a bag. We also theoretically and empirically prove the instance-level learnability of CFSD. Our LadderMIL is evaluated on multiple clinically relevant benchmarking tasks including breast cancer receptor status classification, multi-class subtype classification, tumour classification, and prognosis prediction. Average improvements of 8.1%, 11% and 2.4% in AUC, F1-score, and C-index, respectively, are demonstrated across the five benchmarks, compared to the best baseline.