American Diabetes Association
Supplementary_Material_20231027.pdf (457.24 kB)

Performance of Artificial Intelligence in Detecting Diabetic Macular Edema from Fundus Photographs and Optical Coherence Tomography Images: A Systematic Review and Meta-analysis

Download (457.24 kB)
posted on 2024-01-19, 20:00 authored by Ching Lam, Yiu Lun WONG, Ziqi TangZiqi Tang, Xiaoyan HuXiaoyan Hu, Truong X. Nguyen, Yang, Dawei, Shuyi Zhang, Jennifer Ding, Simon Szeto, An Ran Ran, Carol Y. Cheung

Background: Diabetic macular edema (DME) is the leading cause of vision loss in people with diabetes. Application of artificial intelligence (AI) in interpreting fundus photographs (FPs) and optical coherence tomography (OCT) images allows prompt detection and intervention.

Purpose: To evaluate the performance of AI in detecting DME from FPs or OCT images, and identify potential factors affecting model performances. PROSPERO registration number: CRD42021276009

Data sources: We searched 7 electronic libraries up to 12th February, 2023.

Study Selection: We included studies utilizing AI to detect DME from FP or OCT images.

Data Extraction: We extracted study characteristics and performance parameters.

Data Synthesis: 53 studies were included in the meta-analysis. FP-based algorithms of 25 studies yielded pooled area under receiver operating characteristic curve (AUROC), sensitivity, and specificity of 0.964, 92.6%, and 91.1%. OCT-based algorithms of 28 studies yielded pooled AUROC, sensitivity, and specificity of 0.985, 95.9%, and 97.9%. Potential factors improving model performance included deep-learning techniques, larger size, and higher diversity in training datasets. Models demonstrated better performance when validated internally than externally, and those trained with multiple datasets showed better results upon external validation.

Limitations: Analyses were limited by unstandardized algorithm outcomes and insufficient data in patient demographics, OCT volumetric scans, and external validation.

Conclusions: This meta-analysis demonstrates satisfactory performance of AI in detecting DME from FPs or OCT images. External validation is warranted for future studies to evaluate model generalizability. Further investigations may estimate optimal sample size, effect of class balance, patient demographics, and additional benefits of OCT volumetric scans.