Weakly Supervised Instance-Level Gleason Pattern Estimation Using Primary and Secondary Labels

2026-07-26

AI Summary

This study addresses the challenge in prostate cancer histopathology where only slide-level Gleason primary and secondary pattern labels are available, lacking instance-level annotations and thereby hindering precise lesion identification. To overcome this limitation, the authors propose a multiple instance learning (MIL) framework that, for the first time, aligns instance predictions with the clinical definition of Gleason scoring by aggregating them into class counts. The method explicitly models both primary and secondary Gleason patterns along with their dominance relationship, enabling effective instance-level Gleason pattern estimation from weakly supervised slide-level labels. Evaluated on the SICAP-MIL dataset, the proposed approach outperforms existing MIL methods and achieves significantly improved accuracy at the instance level.

Abstract

In prostate cancer histopathology, the Gleason Score is determined by the most frequent (Primary) and second most frequent (Secondary) Gleason patterns within a whole-slide image. Although these slide-level labels are routinely available in clinical practice, instance-level Gleason annotations are rarely provided, making patch-level learning challenging. We propose a Multiple Instance Learning (MIL) framework that estimates instance-level Gleason patterns from slide-level Primary and Secondary labels. The proposed method formulates instance-level learning according to the clinical definition of the Gleason Score by aggregating instance predictions into class counts and explicitly modeling the Primary pattern, Secondary pattern, and their dominance. Experimental results demonstrate that the proposed formulation enables effective instance-level learning and outperforms existing MIL approaches on the SICAP-MIL dataset.

Problem

• Weakly Supervised Learning
• Instance-Level Annotation
• Gleason Pattern
• Multiple Instance Learning
• Prostate Cancer Histopathology

Innovation

• Multiple Instance Learning
• Gleason Score
• Weakly Supervised Learning
• Instance-level Prediction
• Histopathology