Automatic segmentation of closed-contour features in ophthalmic images using graph theory and dynamic programming.

This paper presents a generalized framework for segmenting closed-contour anatomical and pathological features using graph theory and dynamic programming (GTDP). More specifically, the GTDP method previously developed for quantifying retinal and corneal layer thicknesses is extended to segment objects such as cells and cysts. The presented technique relies on a transform that maps closed-contour features in the Cartesian domain into lines in the quasi-polar domain. The features of interest are then segmented as layers via GTDP. Application of this method to segment closed-contour features in several ophthalmic image types is shown. Quantitative validation experiments for retinal pigmented epithelium cell segmentation in confocal fluorescence microscopy images attests to the accuracy of the presented technique.

Duke Scholars

Author Cynthia Ann Toth Ophthalmology, Vitreoretinal Diseases & Surgery

Author Catherine Bowes Rickman Ophthalmology, Vitreoretinal Diseases & Surgery

Author Sina Farsiu Biomedical Engineering