Reproducibility of Automated Retinal Layer Segmentation Using High-Definition Optical Coherence Tomography

Abstract

Purpose:: To demonstrate the reproducibility of an automated segmentation of distinct neurosensory retinal layers analyzing images obtained with three-dimensional rapid scanning, spectral-domain high-definition optical coherence tomography (HD-OCT) using a newly developed software algorithm.

Methods:: Raster-scanning macular imaging is performed using a HD-OCT with an axial resolution of ~6µm and a resolution of 512*128*1024 voxels. A novel software was developed to obtain an automated segmentation of the neurosensory retinal layers according to their histological architecture. Healthy volunteers and glaucoma patients are scanned twice on one day and a third time on another day. Repeated scans are evaluated in 4 concentric rings around the foveola up to a radius of 2.5mm, each sub-divided in 16 segments. Inter-subject and intra-subject reproducibility of the automated segmentation is tested for the retinal ganglion cell layer (RGCL), combined retinal ganglion and inner plexiform layer (RGIPL), and overall retinal thickness (RET) using analysis of variance components.

Results:: In a preliminary analysis of 15 eyes of 15 healthy volunteers, mean thickness (MT) was 27.40-41.60µm for the RGCL, 58.36-90.04µm for the RGIPL, and 264.91-311.32µm for the RET for ring 1-4. Highest values of MT were found for the second inner ring. Median coefficient of variation (CV) for repeated measurement was 6.70%-6.89% for the RGCL, 3.64%- 4.77% for the RGIPL, and 1.25%-1.44% for the RET for ring 1-4. The lowest CV for all layers was found in the segments of the innermost ring. Inter-subject differences accounted on average for 64.23%-80.62% of variance of RGCL thickness, 81.24%-87.86% of variance of RGIPL thickness, and 94.09%-96.29% of variance of RET thickness.

Conclusions:: An algorithm for automated segmentation of intra-retinal layers was developed for HD-OCT scans of the macula. The automated algorithm was able to detect and delineate distinct neurosensory retinal layers with good reproducibility in repeated scans.

Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • macula/fovea