Robust White Blood Cell Classification with Stain-Normalized Decoupled Learning and Ensembling

2026-03-02

AI Summary

This work addresses the challenges of appearance variation due to staining and scanning inconsistencies and severe class imbalance in real-world white blood cell classification. The authors propose a training framework that integrates stain normalization with decoupled learning: first, instance-balanced sampling is employed to learn transferable representations, followed by classifier rebalancing using a hybrid loss function that combines class-aware sampling with effective-number weighting and focal modulation. During inference, robustness is enhanced through multi-backbone ensembling and test-time augmentation. The method achieved first place in the ISBI 2026 WBCBench Challenge, significantly improving recognition accuracy for rare yet clinically critical white blood cell types and enhancing overall model generalization.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Medical and Biological Imaging
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

White blood cell (WBC) classification is fundamental for hematology applications such as infection assessment, leukemia screening, and treatment monitoring. However, real-world WBC datasets present substantial appearance variations caused by staining and scanning conditions, as well as severe class imbalance in which common cell types dominate while rare but clinically important categories are underrepresented. To address these challenges, we propose a stain-normalized, decoupled training framework that first learns transferable representations using instance-balanced sampling, and then rebalances the classifier with class-aware sampling and a hybrid loss combining effective-number weighting and focal modulation. In inference stage, we further enhance robustness by ensembling various trained backbones with test-time augmentation. Our approach achieved the top rank on the leaderboard of the WBCBench 2026: Robust White Blood Cell Classification Challenge at ISBI 2026.

Problem

• white blood cell classification
• stain variation
• class imbalance
• appearance variation
• rare cell types

Innovation

• stain normalization
• decoupled learning
• class imbalance
• ensemble learning
• test-time augmentation