MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202621063002 A) filed by Annapurna Karna Sahu; Ayesha Siddiqua; Chaitrali Ganeshrao Shende; Zareena S. Joya; and Ganesh. D. Kedar on May 19, 2026, for A Hybrid Framework For Automated Detection Of Diabetic Retinopathy Using Retinal Fundus Images.
Inventors include Annapurna Karna Sahu; Ayesha Siddiqua; Chaitrali Ganeshrao Shende; Zareena S. Joya; and Ganesh. D. Kedar.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Abstract: A Hybrid Framework for Automated Detection of Diabetic Retinopathy Using Retinal Fundus Images The present invention relates to a hybrid framework for automated detection and early prediction of diabetic retinopathy from retinal fundus images. The framework comprises image preprocessing, morphological enhancement, retinal vessel enhancement, lesion localization, lesion segmentation, feature extraction, and hybrid ensemble classification. Retinal images are subjected to grayscale conversion, resizing, normalization, and noise reduction to improve retinal image quality. Morphological operations including opening, reconstruction, and top-hat transformation enhance retinal structures and lesion visibility, while Sobel edge detection, morphological gradients, and fractal-Hessian filtering improve retinal vessel visualization. Lesion localization is performed using curvature instability mapping, fractal density analysis, and Lyapunov intensity dynamics to identify abnormal retinal regions. Segmentation detects microaneurysms, haemorrhages, hard exudates, and soft exudates. Extracted retinal descriptors are classified using a hybrid ensemble comprising SVM, KNN, ANN, and CNN classifiers through probabilistic fusion, thereby providing accurate, annotation-independent, and computationally efficient diabetic retinopathy screening. Fig. 9.
Disclaimer: Curated by HT Syndication.
