MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641075812 A) filed by Vellore Institute Of Technology on June 19, 2026, for Federated Adaptive Multimodal Early Disease Detection System.
Inventors include Mohana Sundari L; Ms. Oorjaa Sonthalia; and Mr. Kaushik Sinha.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention discloses a fault-tolerant federated clinical decision support system for early Sepsis detection, designed for deployment across distributed hospital environments while preserving patient data privacy. The system employs a federated stacking ensemble wherein Hospital Node A (101) and Hospital Node B (102) independently process cardiac and metabolic clinical features using local XGBoost encoders (103, 104), and transmit encoded probability outputs to a central Logistic Regression meta-aggregator that generates a unified Sepsis Risk Score (106) using reliability-weighted fusion. The architecture incorporates four protective mechanisms: a missing-modality masking module (105) that ensures continuous inference under partial data loss, a fairness constraint engine (108) that enforces Equalized Odds across demographic subgroups, an adaptive explainable AI routing module (110) that dynamically selects between Rapid Triage View (111) and SHAP waterfall plot (112) explanations, and a data drift monitoring and audit ledger module (107, 109) that detects distributional shifts and triggers automated model rollback with immutable event logging. The system enhances robustness, fairness, interpretability, and reliability of Sepsis prediction in real-world clinical environments. Fig. 1
Disclaimer: Curated by HT Syndication.
