WBCBench 2026: A Challenge for Robust White Blood Cell Classification Under Class Imbalance

2026-04-12

AI Summary

This work addresses three major challenges in automated white blood cell classification: severe class imbalance, patient-level data isolation, and domain shift under real-world conditions. To this end, the authors establish a benchmark challenge featuring 13 fine-grained morphological white blood cell classes. The evaluation is structured in two phases to simulate development and deployment discrepancies: Phase I provides clean training data, while Phase II introduces degraded images with synthetic noise, blur, and illumination perturbations, with all splits strictly partitioned by patient. This is the first effort to jointly model these challenges in this task, adopting macro-averaged F1 score as the primary metric and providing an open-source evaluator with a standardized submission format. An international challenge based on this benchmark successfully spurred advances in robust medical image classification under extreme class imbalance and domain shift.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Medical and Biological Imaging
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

We present WBCBench 2026, an ISBI challenge and benchmark for automated WBC classification designed to stress-test algorithms under three key difficulties: (i) severe class imbalance across 13 morphologically fine-grained WBC classes, (ii) strict patient-level separation between training, validation and test sets, and (iii) synthetic scanner- and setting-induced domain shift via controlled noise, blur and illumination perturbations. All images are single-site microscopic blood smear acquisitions with standardised staining and expert hematopathologist annotations. This paper reviews the challenge and summarises the proposed solutions and final outcomes. The benchmark is organised into two phases. Phase 1 provides a pristine training set. Phase 2 introduces degraded images with split-specific severity distributions for train, validation and test, emulating a realistic shift between development and deployment conditions. We specify a standardised submission schema, open-source evaluator, and macro-averaged F1 score as the primary ranking metric.

Problem

• white blood cell classification
• class imbalance
• domain shift
• patient-level separation
• morphologically fine-grained

Innovation

• class imbalance
• domain shift
• patient-level separation
• white blood cell classification
• robustness benchmark