Ensemble of Small Classifiers For Imbalanced White Blood Cell Classification

Siddharth Srivastava; Adam Smith; Scott Brooks; Jack Bacon; Till Bretschneider

cs.CVMar 21, 2026

Ensemble of Small Classifiers For Imbalanced White Blood Cell Classification

Authors: Siddharth Srivastava, Adam Smith, Scott Brooks, Jack Bacon, Till Bretschneider

Organizations: Department of Computer Science, University of Warwick · Warwick Medical School, University of Warwick · Intelligent Imaging Innovations Ltd

Abstract

Automating white blood cell classification for diagnosis of leukaemia is a promising alternative to time-consuming and resource-intensive examination of cells by expert pathologists. However, designing robust algorithms for classification of rare cell types remains challenging due to variations in staining, scanning and inter-patient heterogeneity. We propose a lightweight ensemble approach for classification of cells during Haematopoiesis, with a focus on the biology of Granulopoiesis, Monocytopoiesis and Lymphopoiesis. Through dataset expansion to alleviate some class imbalance, we demonstrate that a simple ensemble of lightweight pretrained SwinV2-Tiny, DinoBloom-Small and ConvNeXT-V2-Tiny models achieves excellent performance on this challenging dataset. We train 3 instantiations of each architecture in a stratified 3-fold cross-validation framework; for an input image, we forward-pass through all 9 models and aggregate through logit averaging. We further reason on the weaknesses of our model in confusing similar-looking myelocytes in granulopoiesis and lymphocytes in lymphopoiesis. Code: https://gitlab.com/siddharthsrivastava/wbc-bench-2026.

Figures & tables

Figure 1: Example phase 1 (unperturbed) images for each class from the WBCBench competition dataset.

Figure 2: Number of samples per class in our expanded dataset.

Panel of Figure 3 — Figure 3: Model architecture. Left: our model consists of a $3$ -ensemble of $3$ different architectures: a ConvNeXT-V2 Tiny [ 20 ] , a SwinV2-Tiny [ 10 ] , and a DinoBloom-Small [ 8 ] model. In each, we replace the pretrained classification head with a simple MLP $h$ . Each instantiation of an architecture is fined-tuned independently on $2$ out of $3$ folds of data and validated on the last fold to mitigate overfitting, giving 3 models for each architecture and a total of $9$ models. The final model $\mathcal{M}$ passes an input image through all $9$ architectures and averages the logits to produce the $13$ -class logit vector. Right : At inference, we generate multiple augmented views of each image via random flips and rotations. These views are independently processed by $\mathcal{M}$ and their logits are averaged to obtain the final class prediction.

Figure 4: Confusion matrices of our ensemble trained on the expanded dataset. Left : confusion matrix of

1149

high-confidence misclassified examples,

841

of which come from WBCBench and the rest from Acevedo20; found using confident learning [ 14 ] . The dotted red box groups the classes in granulopoiesis and the green box captures lymphopoiesis; MO occurs in monocytopoiesis. Right confusion matrix generated using out-of-fold examples for each group of models.

		Training Dataset: WBCBench Train + Eval					Training Dataset: Expanded Dataset
		Eval. Set: WBCBench Train + Eval Out-of-Fold				WBCBench Test	Eval. Set: WBCBench Train + Eval Out-of-Fold				WBCBench Test
	Model	Macro F1	Balanced Accuracy	Macro Precision	Macro Sensitivity	Macro F1	Macro F1	Balanced Accuracy	Macro Precision	Macro Sensitivity	Macro F1
	$3\times$ SwinV2	0.7351	0.7759	0.7126	0.9932		0.7410	0.7523	0.7412	0.9936
	$3\times$ ConvNeXt	0.7523	0.7669	0.7488	0.9938		0.7662	0.7660	0.7741	0.9942
	$3\times$ DinoBloom	0.7276	0 .8095	0.6863	0.9925	-	0.7417	0.7739	0.7296	0.9936	-
No TTA	Ensemble	0 .7666	0.789	0 .7586	0 .9941	0 .6638	0 .7791	0 .7736	0 .7977	0.9945	0 .6737

Table 1: Quantitative comparison of SwinV2-Tiny, DinoBloom-Small, ConvNeXt-Tiny, and their ensemble trained on two datasets: WBCBench Train + Eval, and the Expanded Dataset. Results are reported both with and without Test Time Augmentation (TTA). Metrics are computed from out-of-fold predictions to provide an unbiased evaluation; Test Macro F1 (highlighted in pink) serves as the held-out benchmark. Higher values indicate better performance; Bold indicates best performance and underline indicates second-best. Our best model is the ensemble with TTA trained on the expanded dataset.

Explore similar work

Multi-Stage Fine-Tuning of Pathology Foundation Models with Head-Diverse Ensembling for White Blood Cell Classification
Mar 20, 2026Antony Gitau, Martin Paulson, Bjørn-Jostein Singstad +3Fine-Grained Image Classification Pathology Foundation Models
University of South-Eastern Norway · Vestfold Hospital Trust
Synergizing Deep Learning and Biological Heuristics for Extreme Long-Tail White Blood Cell Classification
Mar 17, 2026Duc T. Nguyen, Hoang-Long Nguyen, Huy-Hieu PhamMedical Image Classification Long-Tailed Classification
College of Engineering & Computer Science, VinUniversity, Hanoi, Vietnam · VinUni-Illinois Smart Health Center, VinUniversity, Hanoi, Vietnam · The Computer Vision and Medical AI Lab, VinUniversity, Hanoi, Vietnam
A Hierarchical Ensemble Inference Pipeline for Robust White Blood Cell Classification Under Domain Shifts
Apr 25, 2026Ruyi Dai, Tingkwong Ng, Hao ChenHierarchical Classification Ensemble Learning
Department of Computer Science and Engineering, The Hong Kong University of Science and Technology, Hong Kong, China · Department of Chemical and Biological Engineering, The Hong Kong University of Science and Technology, Hong Kong, China · Division of Life Science, Hong Kong University of Science and Technology, Hong Kong, China +2

Mar 20, 2026J/K move · Enter open · S save

Multi-Stage Fine-Tuning of Pathology Foundation Models with Head-Diverse Ensembling for White Blood Cell Classification

Antony Gitau, Martin Paulson, Bjørn-Jostein Singstad, Karl Thomas Hjelmervik+2

University of South-Eastern Norway · Vestfold Hospital Trust

Fine-Grained Image Classification Pathology Foundation Models Ensemble Learning

Open paper AskPDF arXiv

The classification of white blood cells (WBCs) from peripheral blood smears is critical for the diagnosis of leukemia. However, automated approaches still struggle due to challenges including class imbalance, domain shift, and morphological continuum confusion, where adjacent maturation stages exhibit subtle, overlapping features. We present a multi-stage fine-tuning methodology for 13-class WBC classification in the WBCBench 2026 Challenge (ISBI 2026). Our best-performing model is a fine-tuned DINOBloom-base, on which we train multiple classifier head families (linear, cosine, and multilayer perceptron (MLP)). The cosine head performed best on the mature granulocyte boundary (Band neutrophil (BNE) F1 = 0.470), the linear head on more immature granulocyte classes (Metamyelocyte (MMY) F1 = 0.585), and the MLP head on the most immature granulocyte (Promyelocyte (PMY) F1 = 0.733), revealing class-specific specialization. Based on this specialization, we construct a head-diverse ensemble, where the MLP head acts as the primary predictor, and its predictions within the four predefined confusion pairs are replaced only when two other head families agree. We further show that cases consistently misclassified by all models are substantially enriched for probable labeling errors or inherent morphological ambiguity.

Abstract

Explore similar work

Multi-Stage Fine-Tuning of Pathology Foundation Models with Head-Diverse Ensembling for White Blood Cell Classification

Synergizing Deep Learning and Biological Heuristics for Extreme Long-Tail White Blood Cell Classification

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