MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641072250 A) filed by Vellore Institute Of Technology on June 10, 2026, for Explainable Deep Learning System For Retinal Disease Diagnosis.
Inventors include Arivoli. A; P Sheetal Reddy; and Khushi Arora.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: ABSTRACT EXPLAINABLE DEEP LEARNING SYSTEM FOR RETINAL DISEASE DIAGNOSIS A computer-implemented method for retinal disease diagnosis comprises receiving a retinal fundus image (10) from an image acquisition device (12); preprocessing the retinal fundus image (10) to generate a preprocessed image (20) through resizing, normalization, and quality enhancement; classifying the preprocessed image (20) using a deep learning classification model (110) to generate a disease classification result (30) comprising cataract, diabetic retinopathy, glaucoma, or normal, and a confidence score (32); generating a visual explanation heatmap (40) using a gradient-weighted class activation mapping module (120); evaluating reliability of the heatmap (40) by computing a consistency metric (42) or stability metric (44); generating a semantic alignment score (50) by comparing a diagnostic explanation (52) with clinical reference information (54) using a clinical language model (130); and outputting a diagnostic report (60) comprising the disease classification result (30), confidence score (32), visual explanation heatmap (40), and semantic alignment score.
Disclaimer: Curated by HT Syndication.
