Comparison of OCT Angiography Review Strategies to Identify Vascular Abnormalities in the AVATAR Study.

Purpose: To compare review strategies for optical coherence tomography angiography (OCT-A) for multiple disease features found in common diseases of the choroid and retina.

Design: Prospective, observational study.

Participants: Patients with macular disease undergoing routine spectral-domain optical coherence tomography (SDOCT).

Methods: Eyes were imaged with the Avanti RTVue XR HD (Optovue, Fremont, CA), and the split-spectrum amplitude decorrelation angiography (SSADA) algorithm software was utilized for OCT-A performance. Scans were reviewed by 2 masked expert reviewers. A third masked reviewer was utilized in cases of reviewer disagreement. A single report using automated segmentation within the Avanti software to represent the superficial retina capillary plexus, deep retina capillary plexus, outer retina, and choroid capillary layer was generated. A continuous slab descent video export was also reviewed for each OCT-A scan. This video consisted of a continuous (e.g., line-by-line) review of the OCT- data. Each dataset was reviewed for the presence of three pathologic features: choroidal neovascularization, microaneurysms, and macular ischemia.

Main Outcome Measures: Comparison of identification rates of retinal and choroidal microvascular abnormalities using different review strategies.

Results: Four hundred twenty-one eyes were included in the study. Of those, 350 eyes had reports that were deemed sufficient quality for interpretation and analysis by both reviewers. Identification rates of choroidal neovascularization, microaneurysms, and macular ischemia on the report were 90.5%, 84.5%, and 95.4% respectively compared to the overall presence. Likewise, rates of identification in the continuous slab descent review video were 88.1%, 96.4%, and 95.4% for choroidal neovascularization, microaneurysms, and macular ischemia respectively compared to the overall presence. Cohen's kappa values ranged from 0.80 to 0.96, corresponding to very good agreement between the report and continuous slab descent review for each variable.

Conclusions: Defining an optimal reporting strategy for OCT-A is important for diagnostic accuracy and optimizing workflow in retina clinics. In this study, OCT-A report using automated segmentation was comparable to continuous slab descent review for identifying microvascular abnormalities of the retina and choroid.