International Conference on Innovative Mechanisms for Industry Applications · Oct 2026

Privacy-Preserving Osteoporosis Prediction using Federated Learning and Explainable AI: A Survey

By Ramesh S, Saruladha K

Osteoporosis is a progressive skeletal disease that plays a major role in fragility fracture and long-term disability in the world. While dual-energy X-ray absorptiometry (DXA) is the current gold standard diagnostic method, many areas of the world lack access to this technology, and it is also costly and requires specific infrastructure which makes it difficult to be widely adopted for screening. While AI has shown great promise for predicting osteoporosis risk, issues surrounding patient data privacy and the non-interpretability of predictive models remain obstacles to its implementation in clinical practice. To achieve this, collaborative model training can be implemented using a new paradigm called federated learning (FL), where sensitive patient information is never shared between healthcare institutions, and explainable artificial intelligence (XAI) can increase transparency by creating interpretable predictions to guide clinical decisions. In this paper, we provide a thorough review of 25 peer-reviewed papers addressing AI-based osteoporosis risk prediction, explainable clinical decision support, federated healthcare analytics, and privacy-preserving techniques, such as differential privacy and cryptographic approaches. The review reveals the lack of a federated learning model that simultaneously offers good interpretability, strong privacy protection, and a prediction model for osteoporosis. Lastly, a conceptual framework and future research directions are suggested for creating secure, explainable and clinically deployable AI systems for osteoporosis diagnosis.

doi.org/10.1109/icimia69818.2026.11744774Open accessBiomarkers