Forum on Efficient Privacy-Preserving Federated Learning for Osteoporosis Diagnosis and Healthcare Applications - ANSCSE 28

The 28th Annual International Symposium on Computational Science and Engineering (ANSCSE 28) hosted a special PMU-B x British Council International Research Collaboration Session on July 30, 2025, at Thammasat University (Rangsit Campus). The forum, titled 'Forum on Efficient Privacy-Preserving Federated Learning for Osteoporosis Diagnosis and Healthcare Applications,' brought together leading experts to discuss innovative approaches to osteoporosis diagnosis using advanced AI, medical imaging, and federated learning technologies. Featured speakers included medical doctors and researchers from Thailand and the UK who shared insights on current challenges in osteoporosis care, AI-powered screening methods, and the implementation of federated learning in medical imaging applications.

The session featured several notable presentations. Assoc. Prof. Natthinee Charatcharoenwitthaya, MD, discussed 'Osteoporosis in Thailand: Current Challenges and Opportunities to Improve Access to Care,' highlighting the limited awareness and lack of accessibility to standard diagnostic tools like DXA machines. She proposed the development of AI models to predict bone mineral density from plain radiographs as a promising solution to improve access to care for underserved populations. Woraphun Phutthasakda, MD, spoke on 'Understanding Osteoporosis through Imaging,' explaining that while DXA is the gold standard, other imaging modalities and advances in AI can enhance the detection and follow-up of osteoporosis. Asst. Prof. Somrudee Deepaisarn, PhD, presented on 'Efficient AI Modelling for Osteoporosis Screening Using Radiographs,' proposing a concept to use traditional X-ray images and a convolutional neural network (CNN) to predict bone mineral density (BMD) and classify at-risk patients, offering a less expensive and more widely available screening method. Asst. Prof. Nguyen Truong, PhD, addressed 'Federated Learning for Medical Imaging Application: Challenges and Solutions,' explaining how Federated Learning (FL) allows for collaborative model training without centralizing sensitive patient data, which is crucial for privacy preservation. Prof. Lunchakorn Wuttisittikulkij, PhD, showcased 'Leveraging TU Global Virtual Hospital in the Metaverse for Demonstration of Federated Learning in Osteoporosis Screening,' presenting a novel approach to health education using a virtual hospital in the metaverse to help users understand complex medical information and privacy-preserving tools.