MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109856 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 13, 2026, for An Artificial Intelligence-Based Diabetic Retinopathy Detection System Using Weighted Fusion Deep Learning And Retinal Image Analysis.
Inventor includes Dr. Sayeedakhanum Pathan.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: ABSTRACT [0016] The present invention provides a weighted fusion framework that integrates three advanced deep learning models—EfficientNet, Swin Transformer, and ConvNeXt—for accurate classification of diabetic retinopathy severity from high-resolution retinal fundus images. Diabetic retinopathy is a progressive condition and a leading cause of vision loss among diabetic patients, making timely and consistent severity assessment essential for effective intervention. Each model extracts complementary features: EfficientNet captures fine-grained spatial details, Swin Transformer models long-range contextual relationships, and ConvNeXt supplies hierarchical convolutional representations. Normalized test accuracies of the individual models determine performance-driven weights that are applied during fusion, ensuring balanced contributions and reducing the impact of any single architecture’s limitations. The system is trained and evaluated on the APTOS 2019 Blindness Detection dataset, which contains labeled images spanning five severity grades from no diabetic retinopathy to proliferative disease. Experimental results demonstrate that the fused model achieves a fusion accuracy of 81.31 percent and an F1-score of 79.55 percent, outperforming the standalone models and exhibiting improved generalization across severity classes. A Gradio-based interface enables real-time image upload and returns severity predictions together with class probabilities, thereby offering an automated diagnostic aid that assists clinicians in large-scale screening and early medical intervention.
Disclaimer: Curated by HT Syndication.
