A Hybridized Multilevel Classifier with Improved GWO for Diagnosing Diabetic Retinopathy on Digital Fundus Images

Read full text Summarise with VIRAT
Authors: Karpagavadivu Karuppusamy, T. Baranidharan, Mettupalayam Subramaniam Kavitha, Thangavelu Sumathi
Cited by: 0
References: 41

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Diabetic retinopathy (DR) is a circumstance where a person with diabetes has abnormal blood vessels in their eyes, which can lead to fluid buildup in the macula resulting in diabetic macular edema (DME).However, the traditional method of identifying these conditions through manual analysis of photos is costly and time-consuming, and automated systems are limited in their adaptability to different illnesses and real-world situations.To address these challenges, a new system called Multilevel Classifier with Improved Grey Wolf Optimization (MC-IGWO) has been developed to diagnose and identify DR disorders.The MC-IGWO system extracts various properties of the optic cup, like blood vessels, neural tissue, and the optic disc, and uses feature clustering to find the vessel segment the fundus photograph.By combining the Spectral Classifier with Predictive Rules, the MC-IGWO framework aims to enhance the accuracy of classification for early disease identification.To locate and categorize DR lesions, the DenseNet-100 feature extraction technique is used on top of CenterNet, a one-stage detector.The proposed approach has been tested on challenging datasets such as APTOS-2019 and IDRiD and has realized high average accuracY of 97.89% and 98.1%, respectively.Moreover, the system has been validated across dataseats using EYEPACS and Diaretdb1 datasets.Our method can successfully identify DR and DME lesions, including grainy, low-intensity photos, and extract representative important points compared to previous classification approaches.Therefore, this new system can be highly beneficial in automating the detection and identification of DR and DME lesions, offering a more efficient and cost-effective solution for patients.

Fields of Study & Subjects

Retinal Imaging and Analysis Retinal Diseases and Treatments Ophthalmology and Visual Health Research

Publication details

Journal / source
Traitement du signal
Volume
43
Issue
1
Publication date
28 February 2026
Language
EN