Get started
Work

Artificial Intelligence-Assisted Screening for Early Detection of Diabetic Retinopathy: A Diagnostic Accuracy Study from a Tertiary Care Teaching Hospital in Kota, Rajasthan

Year: 2026
Type: article
Study design: Observational Study
Abstract: Background: Diabetic retinopathy (DR) is a preventable cause of visual impairment, but screening coverage remains inadequate in many Indian settings because of high diabetes burden, limited retina-specialist availability and delayed referral. Objective: To evaluate the diagnostic effectiveness of an AI-assisted fundus-image screening system for early detection of DR among adults with type 2 diabetes mellitus attending a tertiary-care teaching hospital in Kota, Rajasthan. Methods: A prospective diagnostic-accuracy study was conducted from January 2025 to May 2026 at the Department of Ophthalmology, Sudha Medical College, Kota. 210 adults with type 2 diabetes were recruited after reducing the originally planned sample by approximately 50% while retaining acceptable precision for the primary screening outcome. Two-field, non-mydriatic fundus photographs were obtained and analysed by an AI system. Results: Of 210 screened participants, 201 (95.7%) had gradable paired images and were included in the final analysis. Mean age was 56.7 +/- 10.5 years; 111 (55.2%) were male. Retina-specialist grading identified any DR in 64 participants (31.8%), referable DR in 26 (12.9%) and vision-threatening DR in 12 (6.0%). For any DR, AI achieved sensitivity 95.3%, specificity 92.7%, positive predictive value 85.9% and negative predictive value 97.7%. For referable DR, sensitivity was 92.3%, specificity 94.9%, accuracy 94.5% and kappa 0.784. Conclusion: AI-assisted retinal screening demonstrated high sensitivity, specificity and negative predictive value for early DR detection in this tertiary-care setting in Kota.
Language: English
Field: Medicine
Open Access status: diamond
OpenAlex ID: W7216679665
FWCI:
Cites:
Cited by:
Related to: 10