Automated Ten Boundary Detection in Intra- and Outer Retinal Area of Three-Dimensional Optical Coherence Tomography

Abstract

Purpose: : To present an automatic OCT image segmentation algorithm and evaluate the repeatability of outer retinal boundary detection for exploring possible choroid structure change in diseases.

Methods: : A fast and fully automated segmentation method has been developed based on a customized edge detector, in which an edge refinery scheme utilizes the shortest-path search. The introduced edge refinery scheme suppresses the false edges by integrating both intensity and edge information. The algorithm is capable of segmenting up to 10 boundaries, including the inner sclera border, in retinal images. Three-dimensional (3D) macular scans of 20 eyes were utilized to evaluate outer retinal boundary repeatability. Each scan had three repetitions. The 3D scans were preprocessed with a proprietary registration and averaging algorithm to get corresponding 3D scan images with improved signal-to-noise ratio.

Results: : 10 boundaries were detected with the automated segmentation algorithm. The standard deviation and correlation coefficient of the thickness from the outer border of the retinal pigment epithelium to the inner sclera border were calculated to evaluate the repeatability. The newly developed segmentation method demonstrated the ability to reliably segment intra- and outer retinal boundaries with good repeatability. The segmented boundaries were found to be in agreement with the known anatomical locations.

Conclusions: : The proposed segmentation method utilizes the unique edge information and incorporates intensity information to produce robust results. The automated segmentation algorithm was able to produce reliable results for the segmentation of 10 boundaries including the outer retinal area, which is promising towards further investigation of the intra- and outer retinal structure changes in diseased eyes.

Keywords: image processing • choroid • imaging/image analysis: non-clinical