Electroretinograms are an important diagnostic tool to measure retinal electrical activity. However, their interpretation, done by sub-specialised ophthalmologists, can be not only time consuming but also challenging to obtain due to availability. In recent years, studies have investigated the use of artificial intelligence in the analysis of electroretinograms. This systematic review summarises the accuracy of artificial intelligence in interpreting electroretinograms and appraises the studies included. The review comprises primary, peer-reviewed published studies that determined accuracy of artificial intelligence by comparison to an expert ophthalmologist. In the 14 studies retrieved from databases and published between 2006 and 2025, machine learning was the most widely used artificial intelligence, with an accuracy rate between 39.3% and 100%. Overall, the “artificial neural network” machine learning tool was the most accurate. Quality assessment of the studies demonstrated high bias in patient selection but robustness in the methodology for the reference standard, flow and timing. The results revealed potential benefits in the real-world use of artificial intelligence in ophthalmic diagnostic testing; however, the variability in results suggests a requirement for further investigation prior to clinical implementation.
Hegde et al. (Tue,) studied this question.