Abstract
Objective: This study aimed to compare two methods for measuring the foveal avascular zone (FAZ) before and after phacoemulsification: a script-based semiautomated algorithm and a machine learning (ML)-based semiautomated algorithm. Methods: Optical coherence tomography angiography (OCTA) images were obtained with a Spectralis OCTA system (Heidelberg Engineering, Germany) preoperatively and in three postoperative visits. The FAZ was measured using both methods. Results: The study analyzed 708 OCTA scans from 59 eyes. Correlation analyses showed strong agreement between the semiautomated script-based and ML-based methods in the three plexuses, with Pearson correlation coefficients of r = 0.836 (95% CI: 0.74-0.89), r = 0.646 (95% CI: 0.45-0.78), and r = 0.861 (95% CI: 0.78-0.92), respectively (p < 0.0001 for all). In longitudinal analysis, the FAZ in the SVP decreased significantly after phacoemulsification at 1 and 2 months postoperatively with both the script-based method (p = 0.017 and p = 0.039) and the ML-based method (p < 0.0001 and p = 0.004). Conclusions: Our findings suggest that the ML-based approach is a reliable method for measuring the FAZ on OCTA, comparable to the semiautomated script-based algorithm, and may serve as a practical alternative. Moreover, a significant reduction in FAZ within the SVP was observed two months after phacoemulsification.