A team of researchers recently found that a deep learning (DL) system performs comparably to human experts for the joint detection of diabetic retinopathy (DR) and age‐related macular degeneration (AMD).

An ophthalmologist and four independent observers graded 600 images containing referable and non‐referable cases of both DR and AMD to establish a reference standard (RS). Validation was further assessed with the Messidor (1,200 images) for referable DR and the Age‐Related Eye Disease Study (AREDS, 133,821 images) for referable AMD against the corresponding RS.

The investigators discovered that the system achieved areas under the curve (AUC) of 95.1% for the detection of referable DR and 94.9% for referable AMD. The average human performance, on the other hand, had a sensitivity of 61.5% and specificity of 97.8% for DR and sensitivity of 76.5% and specificity of 96.1% for AMD. Further validating these results, the Messidor yielded an AUC of 97.5% for DR and the AREDS provided an AUC of 92.7% for AMD.

“This shows that DL systems can facilitate access to joint screening of eye diseases and become a quick and reliable support for ophthalmological experts,” the study authors wrote in their paper.

González-Gonzalo C, Sánchez-Gutiérrez V, Hernández-Martínez P, et al. Evaluation of a deep learning system for the joint automated detection of diabetic retinopathy and age-related macular degeneration. November 26, 2019. [Epub ahead of print].