MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641106712 A) filed by S K Shreya on September 04, 2026, for Hiervision-Rx.
Inventors include S K Shreya; Prof. D Jagadeesan; and Dr. Lokesh Ravi.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: The present invention provides a computer-implemented system and method for automated multi-disease retinal diagnosis from colour fundus images. The system performs classification across thirteen retinal classes comprising twelve disease classes and one Normal Fundus class, organised into diabetic retinopathy and non- diabetic retinal disease categories. The system utilises an open-access retinal imaging dataset comprising 22,454 images across thirteen classes and performs retina- aware pre-processing and multi-scale feature extraction using a convolutional neural network and a Feature Pyramid Network, followed by lesion-aware attention and anatomical feature encoding. A hierarchical routing mechanism separates diabetic retinopathy from non-diabetic retinal diseases before fine-grained classification. The learned retinal representation is structured using classification, proxy-based and supervised contrastive learning and associated with disease-specific prototypes maintained in a prototype memory. During inference, disease prediction is obtained through similarity between a query retinal representation and the stored prototypes. Monte Carlo Dropout provides image-level uncertainty estimation and supports specialist referral for uncertain predictions. The system further incorporates Explainable Artificial Intelligence (XAI) through Prototype-Patch Activation Similarity Mapping (PPASM), which generates prototype-grounded spatial explanations by combining feature activation with multi-scale patch-to-prototype similarity. Clinical concept attribution and lesion-preserving visualisation further provide interpretable outputs associated with retinal pathological features.
Disclaimer: Curated by HT Syndication.
