Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mamba-Based Ensemble learning for White Blood Cell Classification

Lewis Clifton, Xin Tian, Duangdao Palasuwan, Phandee Watanaboonyongcharoen, Ponlapat Rojnuckarin, Nantheera Anantrasirichai

TL;DR

This paper addresses the challenge of efficient and accurate white blood cell classification in resource-constrained environments by introducing a Mamba-based ensemble framework. It combines five Mamba-derived architectures to form a robust classifier and uses data augmentation with weighted loss to mitigate class imbalance, validated on BloodMNIST and a new Chula-WBC-8 dataset. The results show that Mamba models can surpass traditional DI-60 and several DL baselines, with the ensemble achieving the highest accuracy on BloodMNIST and competitive performance on Chula-WBC-8, highlighting practical benefits for clinical workflows. The work provides a scalable, efficient alternative to Transformer-based approaches and suggests future directions toward automated leukemia detection and broader Mamba-enabled diagnostic tools.

Abstract

White blood cell (WBC) classification assists in assessing immune health and diagnosing various diseases, yet manual classification is labor-intensive and prone to inconsistencies. Recent advancements in deep learning have shown promise over traditional methods; however, challenges such as data imbalance and the computational demands of modern technologies, such as Transformer-based models which do not scale well with input size, limit their practical application. This paper introduces a novel framework that leverages Mamba models integrated with ensemble learning to improve WBC classification. Mamba models, known for their linear complexity, provide a scalable alternative to Transformer-based approaches, making them suitable for deployment in resource-constrained environments. Additionally, we introduce a new WBC dataset, Chula-WBC-8, for benchmarking. Our approach not only validates the effectiveness of Mamba models in this domain but also demonstrates their potential to significantly enhance classification efficiency without compromising accuracy. The source code can be found at https://github.com/LewisClifton/Mamba-WBC-Classification.

Mamba-Based Ensemble learning for White Blood Cell Classification

TL;DR

This paper addresses the challenge of efficient and accurate white blood cell classification in resource-constrained environments by introducing a Mamba-based ensemble framework. It combines five Mamba-derived architectures to form a robust classifier and uses data augmentation with weighted loss to mitigate class imbalance, validated on BloodMNIST and a new Chula-WBC-8 dataset. The results show that Mamba models can surpass traditional DI-60 and several DL baselines, with the ensemble achieving the highest accuracy on BloodMNIST and competitive performance on Chula-WBC-8, highlighting practical benefits for clinical workflows. The work provides a scalable, efficient alternative to Transformer-based approaches and suggests future directions toward automated leukemia detection and broader Mamba-enabled diagnostic tools.

Abstract

White blood cell (WBC) classification assists in assessing immune health and diagnosing various diseases, yet manual classification is labor-intensive and prone to inconsistencies. Recent advancements in deep learning have shown promise over traditional methods; however, challenges such as data imbalance and the computational demands of modern technologies, such as Transformer-based models which do not scale well with input size, limit their practical application. This paper introduces a novel framework that leverages Mamba models integrated with ensemble learning to improve WBC classification. Mamba models, known for their linear complexity, provide a scalable alternative to Transformer-based approaches, making them suitable for deployment in resource-constrained environments. Additionally, we introduce a new WBC dataset, Chula-WBC-8, for benchmarking. Our approach not only validates the effectiveness of Mamba models in this domain but also demonstrates their potential to significantly enhance classification efficiency without compromising accuracy. The source code can be found at https://github.com/LewisClifton/Mamba-WBC-Classification.

Paper Structure

This paper contains 22 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work and Background
  3. Existing white blood cell classification methods
  4. Mamba
  5. Datasets and Preparation
  6. BloodMNIST
  7. Chula-WBC-8
  8. Methodology
  9. Mamba architectures
  10. Vision Mamba (ViM) vim
  11. VMamba vmamba
  12. MambaVision mambavision
  13. LocalMamba localmamba
  14. MedMamba medmamba
  15. Ensemble learning
  16. ...and 7 more sections

Figures (4)

  • Figure 1: Example images from the Chula-WBC-8 dataset of its eight WBC types: segmented neutrophils (SNE), lymphocytes, monocytes, band neutrophils (BNE), eosinophils, myeloblasts, basophils, and metamyelocytes in peripheral blood smears.
  • Figure 2: Diagram of the proposed Mamba-based ensemble learning framework (n=5 in this paper).
  • Figure 3: Comparison of confusion matrices: Sysmex DI-60 (left) vs. Mamba-based ensemble learning (right).
  • Figure 4: Heatmap visualizations of key features (red indicates higher importance) across scenarios: (Left) Both correct, (Middle) Only LocalMamba correct, (Right) Both incorrect. Rows: LocalMamba (top), Swin Transformer (bottom).