Abstract
Purpose :
Optical coherence tomography angiography (OCTA) imaging is known to be prone to artifacts. Floaters, large motion, wavefront error from dry eyes or media opacities when present during the acquisition can reduce the visibility of small capillaries resulting in an inaccurate interpretation and quantification of the images. A flow quality algorithm (AI quantification of OCTA en face image quality – Charles Wu, Luis de Sisternes, IOVS, Aug. 2019, Vol.60, PB099) was recently developed to address the need for a reliable assessment of the quality of OCTA scans. In this study, we analyze how the flow image quality index impacts the repeatability of OCTA measurement to derive a quality index threshold.
Methods :
19 diabetic retinopathy (DR) eyes and 20 heathy eyes (HE) were imaged on PLEX® Elite 9000 swept-source OCT (ZEISS, Dublin, CA) by acquiring 4 consecutive 6×6 mm angiography scans at 100 kHz A-scan rate. For the DR eye cohort, the image quality index, vessel density (VD) and perfusion density (PD) were computed over the 9 ETDRS sectors of the retinal slab. For each eye, the 4 scan values for each ETDRS sector were averaged and the standard deviation from that mean calculated. The mean VD and PD were computer as well on the heathy eye cohort to establish the measurement baseline.
Results :
Figure 1 displays the distribution in the image quality index on a 1-5, scale, 5 being the highest quality. Figures 2a and 2b show the VD and PD standard deviation scattered plots as a function of the flow quality index. These plots show a trend for improved measurement repeatability with index quality score. To better visualize this trend, the standard variation computed over a half-unit quality window is shown on Figures 2c and 2d. At a quality index of 3, the standard deviation in VD and PD of 1.3 and 0.027 respectively corresponds to less than 7% deviation from nominal values (VD = 23.2 mm-1 and PD = 0.4) obtained from healthy eyes.
Conclusions :
The measurement repeatability in vessel and perfusion densities is related to the index quality metric of the flow images. An image quality threshold of 3 is recommended to provide sufficient confidence in the reported flow quality metrics.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.