Realistic Three-Dimensional Segmentation of Healthy Eyes Using High Definition OCT in Combination With High-End Segmentation and Rendering Software

Abstract

Purpose:: Stratus^TM optical coherence tomography systems retinal mapping mode is based on 6 scans and is therefore highly dependent on interpolating software. The introduction of fast raster scanning high definition OCT systems (HD-OCT) has reduced this dependency. However, these systems rely on automatic segmentation algorithms, which may fail in advanced chorioretinal disease. In this study new semi automatic segmentation (SAS) software is compared to an automatic segmentation (AS) algorithm in healthy eyes. This work is crucial to support the results of studies using similar software in advanced chorioretinal disease in future.

Methods:: 20 healthy eyes were examined using a spectral domain HD-OCT system (Prototype of Cirrus^TM HD-OCT). Three-dimensional imaging was performed with an axial resolution of 6µm, a maximum scanning speed of up to 20000 A-scans/second and a resolution of 521x128x1024 voxels. Volumes of 5.8 x 5.8 x 2 mm³ were examined in macular retina. All scans were analyzed using a Matlab^TM based AS algorithm and an examiner controlled scan-by-scan SAS software (3D Doctor^TM, Able Software Corp.).Volumes of different compartments and topographic findings were correlated.

Results:: Scatter plot analysis reveals a high correlation between the AS and the SAS using 3D-Doctor^TM (p<0.01, R² >0.9). Topographic images performed with 3D-Doctor^TM show superior results regarding the integrity of the three-dimensional visualization. Line errors compromising the results of the AS in the macular region are absent in 3D-Doctor^TM segmentations. Retinal vasculature and foveal configuration can be imaged and displayed separately from the retina using the semi automatic mode and threshold analysis.

Conclusions:: This study shows a high correlation between the semi-automated analysis using 3D-Doctor^TM and an SAS. Changes in retinal reflectivity, i.e. photoreceptor layer and retinal vessels, can be isolated and visualized in a realistic three-dimensional pattern. Segmentation of intraretinal volumes using 3D-Doctor^TM is a promising tool to quantify and display changes in retinal microstructure accurately. The semi automated controlled mode allows to reduce artefacts of currently used automated analysis significantly. This might provide new parameters for the exact evaluation of retinal, subretinal and infrapigmentepithelial changes, especially in advanced chorioretinal disease, in future.

Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retina • topography