MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641054621 A) filed by Hindusthan College Of Engineering And Technology on April 29, 2026, for Deep Learning Based Retinal Image Analysis For Diabetic Retinopathy Detection And Staging.
Inventor includes Hindusthan College Of Engineering And Technology.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: ABSTRACT The invention is an intelligent clinical decision support system for automatic detection, risk estimation ~nd sustained monitoring of Diabetic Retinopathy (DR) based on retinal fundus images. This involves the development of a deep learning model based on an EfficientNet architecture augmented with attention mechanisms and optimized with a focal loss for the problem of imbalanced data which helps in improved classi·fication of the undenepresented stages. Both channel and spatial attention are employed by the model to focus on fine-grained lesions - such as microaneurysms and hemonhages, so as to improve perf01mance. ln addition to classification, the system produces risk percentage, clinical alerts based on preset risk values, patient records, comparison with historical data and plotting of disease progression on graph, and provides a system to automatically generate clinical reports. The present invention allows for enhanced diagnostic performance, interpretability, and early diagnosis for efficiently screening diabetic retinopathy.
Disclaimer: Curated by HT Syndication.
