Disease Prediction and Early Diagnosis Using Federated Models
Vibha Tiwari, B. K. Mishra, Nitya Hari Das, Balwinder Singh, Harmandeep Kaur · 2025
Federated learning has emerged as a promising approach to disease prediction and early diagnosis, addressing critical concerns about data privacy and security in healthcare settings. By allowing multiple healthcare institutions to collaboratively refine predictive models without compromising sensitive patient data, it ensures that privacy is maintained while improving the accuracy of diagnostic tools Recent advancements include personalized frameworks that consider the heterogeneity of medical data between hospitals, leading to better generalization and personal trade-offs. Additionally, the integration of federated learning with blockchain technology enhances security and privacy, enabling full data value mining while safeguarding local data sharing. Moreover, the application of different private stochastic gradient descent in federated averaging models has shown promising results in diabetes prediction, offering increased privacy levels without compromising model utility. These advancements collectively contribute to more efficient and secure disease prediction and early diagnosis processes in healthcare.