CorneAI app shows high accuracy in diagnosing corneal and cataract conditions
The CorneAI app for iOS demonstrates high accuracy and reliability in classifying various corneal and cataract conditions across diverse imaging conditions and patient demographics, according to a study.
The app categorizes conditions into 9 distinct groups: normal, infectious keratitis, non-infection keratitis, scar, tumor, deposit, acute primary angle closure, lens opacity, and bullous keratopathy.
The evaluation of CorneAI’s performance on images revealed a positive predictive value (PPV) of 0.75 for the top classification with the highest predictive score. The PPV for the top 3 classifications exceeded 0.80. For specific conditions, the highest PPVs were 0.91 for infectious keratitis, 0.73 for normal, 0.42 for non-infection keratitis, 0.72 for scar, 0.77 for tumor, and 0.55 for deposit.
The study also reported the area under the receiver operating characteristic (ROC) curve for various conditions, highlighting CorneAI’s robust performance. The area under the ROC curve was 0.78 for normal (95% confidence interval [CI] 0.5-1.0), 0.76 for infectious keratitis (95% CI 0.67-0.85), 0.81 for non-infection keratitis (95% CI 0.64-0.97), 0.55 for scar (95% CI 0.41-0.69), 0.62 for tumor (95% CI 0.27-0.97), and 0.71 for deposit (95% CI 0.53-0.89).
Reference
Taki Y, Ueno Y, Oda M, et al. Analysis of the performance of the CorneAI for iOS in the classification of corneal diseases and cataracts based on journal photographs. Sci Rep. 2024;14(1):15517. doi: 10.1038/s41598-024-66296-3. PMID: 38969757; PMCID: PMC11226423.
