FAZ Information Tool (FIT); An accessible tool for analyzing the foveal avascular zone in OCTA images

Abstract

Purpose : The foveal avascular zone (FAZ) is the most prominent feature in optical coherence tomography angiography (OCTA) images of the human macula. The morphometry (i.e. size, shape, and fragmentation status) of the FAZ serve as potential biomarkers. Here, we created and validated an accessible MATLAB application to extract this information from an image.

Methods : A custom MATLAB application, “FAZ Information Tool (FIT)”, was developed based on two previously published methods.^1,2 First, the vasculature within an image is identified by applying several image processing techniques, then the center of the FAZ is estimated and further “cleaned” by the user (manually removing erroneous pixel markings). Regionprops is used to evaluate intercapillary area, and the FAZ is identified as the largest area found within 30% of the user selected approximation. Fragmentation status is automatically determined based on the ratio of the surrounding intercapillary areas to the largest area. To validate this application, 18 Optovue OCTA images from 18 eyes (image scale 1.54-1.89µm/px, 8 male, 10 female, nominal image size = 3 x 3mm) were analyzed via FIT independently by two users. Fragmentation status and FAZ area were compared to results derived from previously published manual FAZ segmentations, and results between users were compared.

Results : Fragmentation status was classified correctly across all trials (7 FAZs were fragmented, 11 were non-fragmented). Average (±SD) FAZ area for the 11 non-fragmented FAZs was 0.276 ± 0.152mm², consistent with previous literature. Average (±SD) difference between manual segmentation and FIT-derived area was 0.015 ± 0.009mm² (9.98 ± 7.92%). The average (±SD) difference between users was 5.9E-5 ± 8.7E-4mm² (0.17 ± 0.81%).

Conclusions : FIT was able to successfully extract FAZ information from OCTA images. Further work to examine the impact of image scale is needed to expand FIT’s utility with other OCTA modalities. In future iterations, accuracy relative to manual segmentation derived results could be improved by allowing the user to add pixel markings during the cleaning step. Additionally, this tool could be expanded to include other FAZ metrics (e.g. perimeter, circularity, roundness) or OCTA metrics (e.g. vessel complexity index, vessel perimeter index, fractal dimension).
1. PMID: 37531114
2. PMID: 31956075

This abstract was presented at the 2024 ARVO Imaging in the Eye Conference, held in Seattle, WA, May 4, 2024.