Sampling Density Affects Reproducibility of OCT 3 Optic Nerve Head Analysis

Abstract

Abstract: : Purpose: All optic nerve head (ONH) analysis parameters using optical coherence tomography (OCT) are based on the locations of the disk margin, which are automatically detected by the OCT system software as the end point of the retinal pigment epithelium and choriocapillaris layers. The purpose of this study was to evaluate the reproducibility of disk margin locations in 3 different sampling density settings and to compare them with subjectively detected disk margin locations. Methods: ONH OCT (StratusOCT; Carl Zeiss Meditec, Inc., Dublin, CA) images were obtained with 3 different sampling densities (128, 256, and 512 A–scans) on eyes of normal volunteers. Four sets of ONH scans, which consist of 6 individual radial scans centered on the ONH, were obtained for each sampling density pre and post dilation, and 3 of 4 sets were selected based on better image quality. The A–scan numbers of the automatically detected disk margin locations were recorded. Subjective disk margin locations were also recorded by an OCT expert. For reproducibility assessment, disk size was calculated by subtracting the A–scan numbers of one disk margin from the other. Results: Ten normal eyes of 10 normal volunteers were enrolled for this study. Eight of 1080 images (0.7%) experienced algorithm failure (only one or none of the disk margins were detected) and excluded from the study. The precision of recording the disk margin location procedure was 0.2 ± 2.8 A–scan lines. Intraclass correlation coefficients (ICC) were highest with 256 samplings for automatically measured disk size (0.34 (128 samplings), 0.44 (256), 0.24 (512), p=0.03, Wilcoxon Test), while there was no difference in ICCs with different sampling settings for those assessed by the expert (0.38 (128), 0.42 (256), 0.39 (512), p=0.78). ICCs were not affected by dilation when disk size was measured automatically (p=0.73), however dilation did improve ICCs with expert assessment (p=0.04). Expert intervention did not improve ICCs significantly (p=0.44). Automatically measured disk sizes were larger than those assessed by the expert (1.75 ± 0.17 mm vs 1.70 ± 0.12 mm, p<0.0001). Conclusions: 256 A–scan sampling density after dilation was the most reproducible setting for automated ONH analysis. Although disk size measurements were significantly different between the automatic and the expert assessment, measurement reproducibility showed no difference.

Keywords: imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • optic disc