Citation

Maheshwari, M. Uma and Tamilselvi, R. and Parisabeham, M. and Shanmugapriya, K. and Senthilpari, Chinnaiyan and Liang, Lee Chu (2025) Automated Pneumonia Detection with Transformer-CNN Architecture. In: 2025 Multimedia University Engineering Conference, MECON 2025, 21 July 2025 - 23 July 2025, Cyberjaya, Malaysia.

Text IEEE Xplore Full-Text PDF_30.pdf - Published Version Restricted to Repository staff only Download (3MB)

Abstract

Pneumonia detection from chest X-rays remains a challenging task due to variations in image quality, overlapping features, and the need for precise localization of affected regions. Traditional CNN-based models, while effective, often struggle with capturing long-range dependencies and global contextual information. Transformer models, on the other hand, excel in self-attention mechanisms but require extensive data and computational resources. To address these limitations, we propose an improved hybrid Transformer-CNN model that integrates the local feature extraction capability of CNNs with the global attention mechanisms of Transformers. Multi-scale feature fusion enhances representation learning, while contrastive learning improves model robustness. Additionally, a Bayesian segmentation approach refines the localization of pneumonia-affected lung regions, improving interpretability. Knowledge distillation is employed to enable lightweight deployment without compromising performance. Experimental results on benchmark chest X-ray datasets show that our proposed model achieves a classification accuracy of 94.8%, outperforming traditional CNNs by 6.5% and standalone Transformers by 4.2%. The segmentation precision is improved by 7.3%, enabling more accurate identification of pneumoniaaffected areas. These advancements contribute to a more reliable and interpretable AI-driven pneumonia detection system.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Pneumonia detection, hybrid transformer-CNN
Subjects:	R Medicine > R Medicine (General) > R858-859.7 Computer applications to medicine. Medical informatics
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	17 Mar 2026 07:08
Last Modified:	17 Mar 2026 07:46
URII:	http://shdl.mmu.edu.my/id/eprint/15520

Downloads

Edit (login required)