Comparative Analysis of Convolutional Neural Networks and Vision Transformers for White Blood Cell Classification in Medical Imaging

Mohamad Abou Ali, Fadi Dornaika, Ignacio Arganda-Carreras, Abdallah Kassem
October 2025
Cite DOI URL

Abstract

Accurate white blood cell (WBC) classification in peripheral blood smears is essential for diagnosing and monitoring hematological disorders. Traditionally, Convolutional Neural Networks (CNNs) have been employed, but they often demand large datasets and substantial computational resources. Recently, Vision Transformers (ViTs) have shown promise by using selfattention mechanisms to capture global image dependencies. This study provides a comparative analysis of CNNs and ViTs for WBC classification, assessing their performance using a standardized dataset of peripheral blood smear images. We evaluate accuracy, computational efficiency, and robustness to varying image qualities. Our results indicate that while CNNs perform well in feature extraction, ViTs excel in managing complex patterns and achieving higher classification accuracy with fewer samples. Additionally, we explore the effects of data augmentation and hybrid models that combine CNNs with ViTs. These approaches enhance model generalization and performance, making them promising for clinical applications requiring diverse data and high accuracy. This research advances the understanding of deep learning in medical imaging, highlighting ViTs as a viable alternative to CNNs for WBC classification. Future research will focus on optimizing these models for real-time clinical use and exploring their application in other diagnostic fields.

Type
Conference paper
Publication
2025 Eighth International Conference on Advances in Biomedical Engineering (ICABME)
White Blood Cells Deep Learning Computer Vision Accuracy Transformers Data Models Robustness Real-Time Systems Convolutional Neural Networks Artificial Intelligence Medical Image Analysis Blood Cell Classification Artificial Intelligence (AI) Deep Learning (DL) Vision Transformers (ViTs) Convolutional Neural Networks (CNNs) Automation
Mohamad Abou Ali
Mohamad Abou Ali
Postdoctoral Researcher

I develop generalizable deep learning methods for biomedical imaging, advancing diagnostic accuracy, data augmentation, and intelligent healthcare systems.

Fadi Dornaika
Fadi Dornaika
Ikerbasque Research Professor

Ikerbasque Research Professor with expertise in computer vision, machine learning, and pattern recognition.

Ignacio Arganda-Carreras
Ignacio Arganda-Carreras
Ikerbasque Research Associate Professor

My research interests include image processing, computer vision, and deep learning for biomedical applications.