Abstract
Purpose :
To compare the diagnostic accuracy of avascular areas in 6x6-mm versus 3x3-mm optical coherence tomography angiography (OCTA) for diabetic retinopathy (DR).
Methods :
A masked grader graded 7-field color fundus photographs using the Early Treatment Diabetic Retinopathy Study scale. The 6x6-mm high-definition (400x400) and 3x3-mm (304x304) OCTA scans were acquired with a commercial system (RTVue XR Avanti, Optovue). Avascular areas (AA) in the superficial vascular complex (SVC) of 6x6-mm scans and SVC, intermediate capillary plexus (ICP), deep capillary plexus (DCP) of 3x3-mm scans were quantified using a deep-learning algorithm. After excluding the central 1-mm circle, extrafoveal avascular area (EAA) was computed (Figure 1). A deep-learning algorithm identified and excluded low-signal areas due to vitreous opacity or vignetting.
Results :
We enrolled 139 diabetic patients (64 men), with a mean age of 55±15 years, including 25 healthy controls, 21 without DR, 28 mild non-proliferative DR (NPDR), 12 moderate NPDR, 17 severe NPDR, and 36 proliferative DR. The mean EAAs significantly increased with DR severity (Table 1). In 3x3-mm macular scans, the SVC EAA showed the highest diagnostic accuracy among the three plexuses. The areas under the receiver operating characteristic curve (AROC) of the SVC EAA for discriminating diabetes without DR, mild NPDR, moderate NPDR, severe NPDR and PDR from healthy control were 0.704, 0.929, 0.987, 0.998, and 0.996 for 3x3mm and 0.706, 0.946, 0.957, 1.000 and 1.000 for 6x6mm respectively. The differences between 3x3 and 6x6 were not significant (DeLong method. All P>0.05).
Conclusions :
The diagnostic accuracy of SVC EAA in diabetic retinopathy were similar between 3x3 and 6x6 mm scans. The information acquired from wider OCTA scans may not improve the diagnostic power of avascular areas in diabetic retinopathy.
This is a 2020 ARVO Annual Meeting abstract.