Deep learning algorithm accurately predicts macular degeneration progression
DeepGAze, a highly accurate and fully automated deep learning algorithm, has been developed to predict the progression from intermediate age-related macular degeneration (iAMD) to geographic atrophy (GA) within a year based on spectral-domain optical coherence tomography (SD-OCT) scans, according to a study.
The study demonstrates that DeepGAze’s predictive capabilities are not only impressive but also autonomous, as the inclusion of expert-annotated features did not enhance its performance.
This retrospective cohort study analyzed data from participants with iAMD at baseline who either progressed or did not progress to GA within the subsequent 13 months. The study used data from 3 distinct sets. The first set comprised patients from the Age-Related Eye Disease Study 2 AREDS2, while the second and third sets encompassed patients from routine clinical care at a tertiary referral center and associated satellites.
The algorithm, which was trained and cross-validated on Bioptigen SD-OCT volumes, demonstrated exceptional predictive capabilities. In data set 1, it achieved an area under the receiver-operator characteristic curve (AUROC) of 0.94, with an area under the precision-recall curve (AUPRC) of 0.90, sensitivity of 0.88, specificity of 0.90, positive predictive value, negative predictive value, and accuracy. Further validation on two external data sets (data sets 2 and 3) using Heidelberg Spectralis SD-OCT scans maintained high performance, with AUROCs of 0.94 and AUPRCs of 0.92.
The inclusion of expert-annotated SD-OCT features did not improve the model’s performance compared to the fully autonomous version, underscoring the algorithm’s self-sufficiency and robustness.
Dow ER, Jeong HK, Katz EA, et al. A Deep-Learning Algorithm to Predict Short-Term Progression to Geographic Atrophy on Spectral-Domain Optical Coherence Tomography. JAMA Ophthalmol. 2023.doi:10.1001/jamaophthalmol.2023.4659