Health Care Informatics Smart Model for Breast Cancer Classification and Detection
Document Type
Conference Proceeding
Source of Publication
2026 11th International Conference on Machine Learning Technologies Icmlt 2026
Publication Date
5-1-2026
Abstract
Breast cancer detection from mammographic imaging continues to be a complex problem, particularly due to the presence of dense and structurally heterogeneous breast tissues that obscure lesion boundaries and reduce diagnostic sensitivity. In this study, we propose an improved transformer-based diagnostic framework using a hierarchical Vision Transformer (ViT) architecture, Adaptive Sharpness-Aware Minimization (ASAM), and Bayesian optimization (BO) to boost predictive performance and model robustness. The proposed model introduces multi-scale patch embedding and hybrid convolution-transformer blocks to effectively capture local and global contextual information from mammographic images, which are not typically captured by conventional ViT implementations. Experimental results show that the proposed framework can obtain 96.2% classification accuracy and 0.97 Area Under the Curve (AUC) and is significantly better than the traditional convolutional neural networks (CNN) baseline.
DOI Link
ISBN
[9798331548650]
Publisher
IEEE
First Page
587
Last Page
591
Disciplines
Computer Sciences | Medicine and Health Sciences
Keywords
Breast Cancer Detection, Health Care Informatics, Mammography Analysis
Scopus ID
Recommended Citation
Abugabah, Ahed; Al-Smadi, Ahmad Mohammad; and Alhadhrami, Ahmed, "Health Care Informatics Smart Model for Breast Cancer Classification and Detection" (2026). All Works. 8447.
https://zuscholars.zu.ac.ae/works/8447
Indexed in Scopus
yes
Open Access
no