Efficacy of artificial intelligence in detecting diabetic retinopathy from retinal fundus images — a systematic review

Abstract

Background: Diabetic retinopathy (DR) is a microvascular disorder that damages the retina’s blood vessels. This review aims to evaluate scientific literature about the efficacy of artificial intelligence (AI) in detecting diabetic retinopathy from retinal fundus images. Material and methods: Systematic literature review was carried out following preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines. Selected Medical Subject Headings (MeSH) keywords were used to search “PubMed” and “ScienceDirect” databases. Inclusion and exclusion criteria were developed using the patient, intervention, comparison, outcome (PICO) tool. Quality assessment of selected studies was done using a modified seven-item checklist based on the methodological index for nonrandomized studies (MINORS) criteria. Results: 15 studies from 14 scientific publications were included in this systematic review. AI algorithms analyzed a total of 150179 images. The AI-based algorithm’s average sensitivity (Se) was 92.58 %, ranging from 76.2% to 100%. The average specificity (Sp) was 87.22%, with the lowest of 53.16% and the highest of 98.5%. The average area under the receiver operating characteristic (AUROC) curve was 0.937, with a variation of 0.843 to 0.9905. Conclusion: Our results show that AI-based algorithms can accurately detect DR in retinal fundus images. These systems should be considered of use in clinical practice to save time and reduce costs.