Abstract
Purpose :
To compare the performance in detecting peripapillary capillaries among 4 optical coherence tomography angiography (OCTA) devices.
Methods :
109 eyes from 109 healthy participants were imaged with 4 OCTA devices (Spectralis, HEYEX 6.8b; Optovue, Angiovue 2016.2.0.35; Triton, OCTARA 1.0; Cirrus, Angioplex 9.0). A 3×3mm scan pattern centered on the disc was repeated twice for each device with a minimum requirement for the signal strength. En face images of superficial capillary plexus (for Spectralis and Cirrus) or optic nerve head (for Optovue and Triton) were used and no segmentation was adjusted. Image was excluded if there’s obvious artifact or central deviation. One scan was chosen from each device and the 4 images of an eye were matched to ensure the same location was evaluated. All the images underwent a multi-scale Hessian filter, an autothresholding (Otsu), subtraction of large vessels and skeletonization. Vessel length density (VLD) was calculated on 4 equally divided parts (superior (S), nasal (N), inferior (I) and temporal (T)) of a ring between two concentric circles (1.5 and 2.25mm in diameter) manually centered on the disc (figure1&2). A general linear model (GLM) was used to test the impact of various devices and locations on the VLD. Intraclass correlation coefficient (ICC) of VLD between repeated scans was calculated for each device.
Results :
Among 218 scans, 36%(79/218), 92%(201/218), 76%(165/218) and 88%(192/218) were eligible for analysis from Spectralis, Optovue, Triton and Cirrus, respectively. The mean VLD (N=39) of S, N, I and T was 20±3, 25±3, 19±3 and 23±4 mm-1 for Spectralis, 19±2, 21±2, 19±2 and 23±2 mm-1 for Optovue, 22±3, 23±3, 22±2 and 24±3mm-1 for Triton and 17±2, 21±2, 17±2 and 20±4 mm-1 for Cirrus. ICC (N=20) was 0.94(95%confidence interval (CI), 0.86-0.98), 0.90(95%CI, 0.74-0.96), 0.84(95%CI, 0.59-0.94) and 0.87 (95%CI, 0.67-0.95) for the 4 devices, respectively. GLM showed both device and location had significant impact on the VLD (P<0.001). Pairwise comparison showed Triton=Spectralis>Optovue>Cirrus and T=N>S=I in terms of the amount of detected skeletonized capillaries.
Conclusions :
Triton and Spectralis detected more peripapillary capillaries, while Spectralis achieved fewest adequate images mainly from artifacts. The repeatability was similar among the devices.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.