MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073775 A) filed by Vellore Institute Of Technology on June 14, 2026, for Resource-Adaptive Diagnostic Prediction Apparatus With Entropy-Gated Model Routing And Selective Explainability.
Inventors include S Vinila Jinny; and Shlok Kumar Goenka.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention relates to a computer-implemented resource-adaptive entropy-gated diagnostic prediction apparatus executed through medical-grade computing hardware. Physiological and demographic patient data acquired from clinical acquisition interfaces are normalized into diagnostic feature vectors for machine learning inference. A lightweight analytical tier generates preliminary diagnostic probability estimates, while a gating logic module computes diagnostic uncertainty and compares the uncertainty against a calibrated entropy threshold for selectively routing physiological records between heterogeneous analytical pathways. Diagnostically unambiguous records are finalized through the lightweight analytical tier without activation of a high-precision ensemble analytical tier, whereas diagnostically ambiguous records are escalated toward the ensemble tier for refined diagnostic evaluation. A Shapley-value-based explainability module generates clinician- interpretable feature attribution outputs for escalated records. The disclosed architecture reduces processor utilization, computational load, memory bandwidth consumption, thermal overhead, and inference latency while preserving clinically acceptable diagnostic accuracy across medical-grade and edge-based healthcare computing environments.
Disclaimer: Curated by HT Syndication.
