Comparative Evaluation of Foveal Avascular Zone on Two Optical Coherence Tomography Angiography Devices

Abstract

SIGNIFICANCE: This study compares foveal avascular zone (FAZ) geometry in healthy eyes as imaged by two commercially available optical coherence tomography angiography (OCTA) devices. Foveal avascular zone measurements are repeatable and reproducible with each OCTA device, but interdevice agreement was poor. We provide conversion factors between devices. PURPOSE: The purpose of this study was to perform comparative evaluation of FAZ geometry in healthy eyes as imaged by two commercially available OCTA devices. METHODS: Ninety-six eyes of 48 healthy subjects were imaged prospectively on each of two OCTA devices (DRI-OCT [Topcon Corporation, Tokyo, Japan]; Cirrus 5000 [Carl Zeiss Meditec Inc., Dublin, CA]). The FAZ was evaluated in the superficial capillary plexus layer of 6 × 6-mm foveal scans by two masked observers. Intraobserver and interobserver agreement was determined using intraclass correlation by using linear mixed models and Bland-Altman plots. K-means clustering was used to provide conversion values between two devices. Foveal avascular zone acircularity was calculated using scans from each device and compared. RESULTS: The intraobserver repeatability for DRI-OCT was 0.95 (95% confidence interval [CI], 0.90 to 0.98) for observer A and 0.92 (95% CI, 0.83 to 0.96) for observer B. Intraobserver repeatability for Cirrus 5000 was 0.988 (95% CI, 0.972 to 0.995) for observer A and 0.993 (95% CI, 0.983 to 0.997) for observer B. The interobserver variability between observers A and B for DRI-OCT was 0.87 (0.73 to 0.94) and for Cirrus 5000 was 0.984 (95% CI, 0.964 to 0.993). Poor interdevice agreement (0.205 [95% CI, -0.202 to 0.554]) was noted, and conversion formulas were devised to convert FAZ area measurements from one device to another. No significant correlation was found when comparing FAZ acircularity indices between devices (P = .39). CONCLUSIONS: Repeatable and reproducible FAZ area measurements were obtained with each respective OCTA device, but interdevice agreement was poor, yet quantifiable and systematic with calculable conversion factors between devices.