Impact of AI on the Accuracy of Diabetic Retinopathy Staging: A Retrospective Analysis in Urban Healthcare Centres
Year: 2025
Type: article
Abstract: ABSTRACTThis study aims to assess the role of artificial intelligence (AI) in accurately staging diabetic retinopathy (DR), with emphasis on early and moderate stages, as timely detection is essential to prevent progression to severe disease and irreversible blindness [4]. Research has shown that AI, particularly deep learning approaches such as convolutional neural networks (CNNs), delivers high sensitivity and specificity in DR detection, often matching or surpassing the performance of retina specialists [5]. These models not only identify DR but also stage its severity effectively, making them highly scalable for addressing the global diabetes burden. Evidence indicates that original fundus images provide better diagnostic accuracy than segmented vessel images, as complete retinal features enhance model performance. Recent advances, including spatial attention modules and improved CNN architectures, further support early lesion detection and improve staging precision. Robust validation on diverse datasets confirms the generalizability of these systems, making them suitable for clinical deployment. To ensure reliability, adjudicated consensus grading has been adopted as a reference standard, reducing variability among human graders and strengthening algorithm outcomes. Importantly, AI has demonstrated strong potential in resource-limited settings, complementing strategies such as tele ophthalmology and technician-led screening, thereby expanding access to care. In this retrospective study, the Messidor dataset of 1,744 images was used, with 1,200 selected for analysis. Images were annotated by optometrists and validated by an ophthalmologist, forming ground truth labels for CNN model training [1,3,6]. The AI system achieved 97% sensitivity and 93.5% specificity in staging early and moderate DR. These findings confirm AI’s potential as a reliable, accurate, and efficient screening tool for urban healthcare systems.
Authors: Manjira, Acharyya, Sudipta, Chatterjee
Language: English
Topic: Retinal Imaging and Analysis
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7083466814
FWCI:
Cites:
Cited by:
Related to: