Diagnosis of Diabetic Retinopathy by Ultra-Wide Filed Image and Artificial Intelligence: A Systematic Review and Meta-Analysis

Abstract

Purpose : To clarify the current status and performance of diabetic retinopathy (DR) screening using ultra-wide field (UWF) images and their subsequent Artificial Intelligence (AI)-driven analysis.

Methods : We conducted a systematic review and meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guidelines. We systematically searched all articles published in two major electronic bibliographic databases (PubMed and EMBASE) up to July 7, 2022, using the most appropriate combination of terms (“diabetic retinopathy” OR “proliferative retinopathy” OR “retinal disorders” OR “diabetic eye disease”) AND (“artificial intelligence” OR “deep learning” OR “transfer learning” OR “machine learning” OR “automated detection” OR “Convolutional Neural Network”) AND (UWF OR wide field OR fundus photo OR retinal image OR fundus image). Of the eligible studies, the diagnostic performance results from studies that aimed to screen for DR were pooled for meta-analysis.

Results : Of the identified 1077 studies, 9 studies were included in the systematic review, of which 4 studies were included in the meta-analysis. The studies included in the systematic review were published between 2018 and 2022 and all used images taken with Optos®. The median sample size is 1,177 (interquartile range [IQR]; 3,289, range; 224-13,271). The meta-analysis revealed that the diagnostic performance of AI-driven DR screening using UWF images was 85.0%, 72.5%, and 0.870 (0.838-0.897) for sensitivity, specificity, and area under the receiver operating characteristics curve, respectively.

Conclusions : AI-driven evaluation of UWF images for DR screening may have sufficient sensitivity but insufficient specificity based on previous standards for AI models using color fundus photo images. However, with a more robust UWF images dataset, the wider coverage of the peripheral retina in UWF image could be advantageous in determining DR severity, which could have implications for timely interventions and improved visual outcomes for the patients with diabetes.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.