GMM cluster based choroid segmentation in EDI OCT images

The structure of Choroidal layer can reflect many diseases such as polypoidal choroidal vasculopathy and choroidal tumors; hence proper segmentation of this layer is a major task. Also the assessment of choroidal thickness from Enhanced Depth Imaging (EDI-OCT) image is an important basis for diagnosing. Mostly this is done by manual labeling which is a tedious and complicated process especially when the numbers of images are more and thus automatic segmentation techniques should be developed to assist the ophthalmologist in diagnosis and monitoring of eye disease effectively. The segmentation of both Bruch Membrane (BM) and the Choroid-Sclera Interface (CSI) layer can be done by Choroidal segmentation. This paper presents a Dynamic Programming based algorithm for BM that denotes the upper bound of the choroid detection and Expectation Maximization (EM) based Gaussian Mixture Model (GMM) for CSI that denotes the lower bound of the choroid segmentation