Abstract
Purpose :
Most commercial OCT Angiography (OCTA) instruments currently lack native software for quantifying vasculature morphology. Due to manufacturers’ proprietary processing methods, comparing OCTA metrics across devices is currently not feasible. We have created an OCTA analysis tool kit (OAT) which generates metrics from images using a standardized approach and in this study attempt to determine the repeatability and reproducibility of its measurements for inter-device comparisons.
Methods :
Five subjects without ocular pathology were imaged using the Topcon DRI OCT Triton, Zeiss Angioplex and the Optovue RTVue-XR Avanti. Three OCTA images were acquired on each device for each subject. All images were exported for analysis by OAT. OCTA metrics generated for analysis included: foveal avascular zone (FAZ) area, vessel index, vessel length index and vessel diameter index. Repeatability was assessed by looking at the average relative size of difference in standard deviation across the devices, and reproducibility was assessed by analyzing measurements generated by repeat analysis of a single image three times.
Results :
For the Avanti device, the average relative size of the standard deviation was determined to be 4.1% for the vessel index, 3.0% for vessel length index and 2.1% for vessel density measurements. The Zeiss analysis yielded an average of 4.9% in vessel index, 3.0% in vessel length and 2.1% in vessel density in relative standard deviation size. Evaluation of the Topcon data showed a relative deviation of 5.4%, 2.0% and 5.1% respectively. Repeated evaluation of one image using the OAT showed identical values with each analysis for each parameter excluding the FAZ measurements on all OCTA machines.
Conclusions :
The small and consistent size of standard deviation relative to the mean across the devices indicates a high level of repeatability for measurements using the OAT software. Identical values on repeat analysis of the same image validates the software’s reproducibility. A larger study with inclusion of more subjects is needed for further validation.
This is a 2020 ARVO Annual Meeting abstract.