Artificial intelligence detects retinal fluid on OCT in AMD better than retinal specialists
An artificial intelligence algorithm was able to detect retinal fluid on OCT scans with a higher level of accuracy than retinal specialists, according to a presentation at the 2020 ASRS Virtual Meeting.
In this prospective study, SD-OCT scans from 651 participants (1,127 eyes) were graded for intraretinal and subretinal fluid by artificial intelligence (AI)-based Notal OCT Analyzer (NOA) and reading center (RC) graders.
Retinal fluid was present in 370 eyes. Investigators’ accuracy was 0.805 (95% CI 0.780-0.828), sensitivity was 0.468 (0.416-0.520), and specificity of 0.970 (0.955-0.981) compared to NOA accuracy was 0.851 (0.829-0.871), sensitivity was 0.822 (0.779-0.859), and specificity of 0.865 (0.839-0.889).
Investigator performance in regard to detecting intraretinal fluid was 0.815 (0.792-0.837) for accuracy, 0.403 (0.349-0.459) for sensitivity, and 0.978 (0.966-0.987) for specificity; NOA was 0.877 (0.857-0.896), 0.763 (0.713-0.808), and 0.922 (0.902-0.940), respectively.
The authors concluded: “This software tool could assist physicians in detecting retinal fluid, which is important for retreatment and prognostic tasks in AMD.”
Elman M, et al. Artificial intelligence algorithm may improve performance of retinal specialists in detection of retinal fluid on OCT in AMD: AREDS2 10-year study. Presented at: 2020 ASRS Virtual Meeting.