IndiaHealthcare and life sciences2022 – ongoing

Postgraduate Institute of Medical Education and Research, Chandigarh

AI-assisted diabetic retinopathy screening in Punjab

PGIMER's screening pilot detected referable retinopathy with 78.9% sensitivity, with weaker macular-edema detection

WorkingIndependent study

Problem

Limited specialist capacity made diabetic eye screening difficult to provide in public health facilities.

Approach

PGIMER researchers evaluated commercial screening algorithms and integrated the selected system with a low-cost retinal camera. A trained optometrist used it at CHC Badhani Kalan from February to June 2022. Researchers compared its classifications with human grading.

Results

78.9%sensitivity for referable diabetic retinopathy during implementationIndependent study[1]

98.1%specificity for referable diabetic retinopathy during implementationIndependent study[1]

26.5%sensitivity for diabetic macular edema during implementationIndependent study[1]

Takeaway

Performance differed substantially by condition. Macular-edema sensitivity remained low despite stronger results for referable retinopathy.

Sources

Real-World Evaluation of AI-Driven Diabetic Retinopathy Screening in Public Health Settings: Validation and Implementation Study
JMIR Medical Informatics9 Sep 2025Research paper or report

Source summaries

[1]JMIR Medical Informatics9 Sep 2025

Researchers reported a public-clinic implementation involving 343 participants. Referable-retinopathy sensitivity was 78.9% and specificity was 98.1%; macular-edema sensitivity was 26.5%.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026