An image processing pipeline for segmenting the retinal layers from optical coherence tomography images

Optical Coherence Tomography (OCT) is the standard clinical modality for high resolution, three dimensional imaging of the multi-layered anatomy of the retina. This paper describes an image processing pipeline we developed for automated segmentation of the retinal strata within OCT image data-sets. The pipeline consists of three independent postprocessing modules that involve combinations of noise reduction i.e. 2D log-Gabor filtering, edge detection and interpolation. Our automated method was validated against manual measurements from three anonymized clinical OCT datasets. We found that our approach was able to accurately segment the layers of the retinal ultrastructure with mean differences of 1μm compared to manual measurements. Overall, we have developed a sophisticated automated application for segmenting the highly-structured retinal anatomy. Ultimately, this tool will be used to create detailed computer meshes of the retinal ultrastructure for our future modeling.