MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202621109833 A) filed by Pranita Sanjay Waikar on September 13, 2026, for Adaptive And Personalized Federated Learning For Pregnancy Risk Prediction Under Data Heterogeneity And Privacy Constraints.
Inventors include Pranita Sanjay Waikar; and Dr. Pallavi Jha.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: A computer-implemented federated processing system and method are disclosed for generating pregnancy-risk assessment outputs across a plurality of participating healthcare institutions while retaining raw patient data locally. Each client computing node trains a received global machine-learning model using locally stored pregnancy-related data and generates a local model update. At least two heterogeneity indicators are determined and combined to generate a client-specific heterogeneity score representing deviation of the local data distribution from a federated reference distribution. An aggregation computing system determines an adaptive client contribution weight based at least on the client-specific heterogeneity score and local data volume and generates an updated global model by weighted aggregation of valid client updates. In a further embodiment, a reliability measure associated with the local model update is additionally used in determining the adaptive client contribution weight. The same client-specific heterogeneity score is further used to determine a client-specific personalization parameter controlling a degree of local adaptation of the updated global model. The personalized local model generates a machine-readable pregnancy-risk assessment output without requiring transfer of the underlying patient record. The system may further provide privacy-preserving communication, local explainability, update validation, contribution bounding, missing-indicator handling, and fallback aggregation.
Disclaimer: Curated by HT Syndication.
