A Hierarchical Ensemble Inference Pipeline for Robust White Blood Cell Classification Under Domain Shifts

2026-04-25

AI Summary

This work addresses domain shift in white blood cell (WBC) classification caused by variations in staining protocols, scanning devices, and laboratory conditions. To enhance robustness in real-world clinical settings—particularly for identifying rare but critical subtypes such as blasts—the authors propose a novel paradigm integrating a feature memory mechanism with hierarchical k-nearest neighbor (kNN) ensembles. The approach employs a three-level hierarchical inference pipeline, leveraging a LoRA-finetuned DinoBloom backbone and a dynamic feature bank for storage and retrieval, with kNN-based decisions introduced at each level to reduce reliance on any single prediction path. Evaluated on the WBCBench dataset, the method achieves a macro F1 score placing it among the top ten submissions in the test phase, demonstrating its effectiveness and competitiveness in cross-domain WBC classification.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Segmentation

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Automated white blood cell (WBC) classification is essential for scalable leukaemia screening. However, real-world deployment is challenged by domain shifts caused by staining protocols, scanner characteristics, and inter-laboratory variability, which often degrade model performance. The White Blood Cell Classification Challenge (WBCBench) at ISBI 2026 aims to advance robust WBC recognition, with a focus on accurately identifying blast cells and other clinically critical rare subtypes. We propose a memory-augmented, hierarchical ensemble pipeline for WBC classification under domain shifts, leveraging a feature bank and a DinoBloom backbone fine-tuned with LoRA. Our three-stage inference hierarchy combines k-nearest neighbors (kNN) retrieval at each level, reducing over-reliance on any single decision. Evaluated on the WBCBench dataset, our method ranks within the top ten by macro F1-score in the final testing phase.

Problem

• white blood cell classification
• domain shifts
• blast cells
• rare subtypes
• robust recognition

Innovation

• hierarchical ensemble
• memory-augmented inference
• domain shift robustness
• LoRA fine-tuning
• kNN retrieval