Cross-Patient Pseudo Bags Generation and Curriculum Contrastive Learning for Imbalanced Multiclassification of Whole Slide Image

TL;DR

This paper introduces a novel framework for imbalanced multiclass classification of whole slide images using pseudo-bag generation and curriculum contrastive learning, significantly improving diagnostic accuracy.

Contribution

It proposes a new method that generates sub-bags and pseudo-bags, along with an affinity-based sample selection and curriculum contrastive learning, to better handle class imbalance in WSI analysis.

Findings

Achieved an average 4.39-point F1 score improvement over previous methods.

Demonstrated effectiveness on tumor classification and lymph node metastasis datasets.

Enhanced stability and representation learning in multi-instance bag analysis.

Abstract

Pathology computing has dramatically improved pathologists' workflow and diagnostic decision-making processes. Although computer-aided diagnostic systems have shown considerable value in whole slide image (WSI) analysis, the problem of multi-classification under sample imbalance remains an intractable challenge. To address this, we propose learning fine-grained information by generating sub-bags with feature distributions similar to the original WSIs. Additionally, we utilize a pseudo-bag generation algorithm to further leverage the abundant and redundant information in WSIs, allowing efficient training in unbalanced-sample multi-classification tasks. Furthermore, we introduce an affinity-based sample selection and curriculum contrastive learning strategy to enhance the stability of model representation learning. Unlike previous approaches, our framework transitions from learning bag-level representations to understanding and exploiting the feature distribution of multi-instance bags. Our method demonstrates significant performance improvements on three datasets, including tumor classification and lymph node metastasis. On average, it achieves a 4.39-point improvement in F1 score compared to the second-best method across the three tasks, underscoring its superior performance.