Comparison of OCTA algorithms for detecting vascular and non-vascular flow signal on hyperreflective lesions in diabetic retinopathy

Abstract

Purpose : To evaluate and compare the performance of different optical coherence tomography (OCT) angiography (OCTA) algorithms on detecting flow signal on hyperreflective lesions on retina, such as microaneurysms, hard exudates and cysts.

Methods : In eyes with diabetic retinopathy, we obtained 3x3-mm macular OCTA scans (304x2x304) using a commercial 70-kHz OCT system (Optovue, Inc) and high-quality color photographs. We processed each OCTA scan with complex signal based optical microangiography (OMAG) and split spectrum amplitude decorrelation angiography (SSADA) methods. In OMAG, the bulk motion phase was accurately compensated using our standard deviation-based method. The en face maximum projection angiograms were then generated for each method using the same slabs. Color photographs were examined for presence of microaneurysms and hard exudates within the 3x3mm OCTA field of view. The superficial vascular complex (SVC) and deep capillary plexus (DCP) en face images as well as representative B-scan images were compared to color photographs and between two algorithm groups.

Results : The images from 12 eyes were studied. Both OCTA algorithms demonstrated true flow signal of microaneurysms (Fig. 1) in the 9 eyes with microaneurysms identified in color photographs. The sensitivity of algorithm reflected by the number and the size and the location of microaneurysm flow signal is comparable. OMAG shows strong flow signal associated with hard exudates without associated microaneurysms on 6 eyes. SSADA shows minimal artifactual flow signal in these areas (Fig. 2). In areas with highly reflective material within cysts in structural OCT, OMAG showed a stronger artifactual flow signal (Fig. 2).

Conclusions : SSADA computes the flow signal using decorrelation, which is variance normalized to the reflectance intensity. Because the flow signals from OMAG is not normalized to the reflectance intensity, it is more dependent on the reflectance signal, resulting in significant artifactual detection of flow signal in the areas with highly reflective material. In this study, SSADA was less prone to artifactual flow signal detection associated with hyperreflective lesions compared to OMAG.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.