Citation

Afroze, A. Sabah and Tamilselvi, R. and Judith, J. and Beham, M. Parisa and Gayathri, M. and Senthilpari, Chinnaiyan (2025) Early Osteoarthritis Detection Using Multi-Scale Hybrid Vision Transformer and Thermal Imaging. In: 2025 Multimedia University Engineering Conference, MECON 2025, 21 July 2025 - 23 July 2025, Cyberjaya, Malaysia.

Text 64.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Osteoarthritis (OA) is a progressive joint disorder, and early detection is crucial for preventing irreversible cartilage damage. Traditional diagnostics, such as X-rays and MRIs, primarily detect structural changes in later stages. The new multi-scale hybrid vision transformer (MSHViT) is suggested to use thermal imaging to find early signs of OA. This technology can detect alterations in microcirculation and vascular inflammation before the visible degradation of cartilage. The MSHViT model should use selfattention and convolutional embeddings to find small differences in temperature that could be signs of OA before they become symptoms. The model's self-calibrating attention module enhances its ability to identify low-intensity inflammatory hotspots earlier. Also, gradient-weighted class activation mapping (Grad-CAM) gives us understandable heatmaps that show us where the temperatures aren't normal for clinical validation. MSHViT achieves 98.1% accuracy, 97.5% sensitivity, and 97.2% F1-score on a 2,500-image thermal dataset that includes people with mild symptoms but no X-raydetectable damage. This is a much higher score than traditional deep learning models. These results show that thermal imaging is a safe, low-cost way to check for early signs of OA. It could be used in telemedicine and remote healthcare. Future research will explore multispectral thermal imaging and federated learning to improve cross-clinical generalisation

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Early osteoarthritis detection, thermal imaging, vision transformer,
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:	18 Mar 2026 06:55
Last Modified:	18 Mar 2026 06:55
URII:	http://shdl.mmu.edu.my/id/eprint/15553

Downloads

Edit (login required)