Vision Transformer–based Automated Classification of Peripheral Blood Cells for Hematological Cancer Screening

View at publisher Summarise with VIRAT
Authors: Eshika Jain, Vinay Kukreja
Cited by: 0
References: 17

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Early and accurate identification of abnormal blood cells plays a critical role in screening hematological disorders, including leukemia. Traditional manual microscopic examination takes time and is operator-dependent and susceptible to operator variation in diagnosis. To overcome these issues, this paper presents a Vision Transformer (ViT) based framework for automatic blood cell classification from high-resolution peripheral smear images. The dataset is composed of five clinically relevant cell types, and a custom preprocessing pipeline was designed for improved morphological feature representation (stain normalization, region of interest extraction, patch tokenization, and ImageNet normalization). The model was optimized on samples that have been preprocessed, and achieved a test accuracy of 95.7%, with consistently high precision, recall, and F1-scores for each class. Confusion matrix evaluation and qualitative analysis are further used to validate the robustness of the model with very little misclassification between morphologically similar cell types. The results show that ViT-based architectures can be used as reliable decision support systems for hematology and hold promising potential for being deployed on a large scale in clinical diagnostics.

Fields of Study & Subjects

Digital Imaging for Blood Diseases Cell Image Analysis Techniques AI in cancer detection

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1871–1876
Publication date
21 January 2026
Language
EN