The Retina is a multi-layer structure which is located in the inner layer of the eye and contains several cellular layers. Age-related macular degeneration (AMD) and diabetic macular edema (DME) are the most common diseases in retina which can be manifested by fluid regions in Optical coherence tomography (OCT) images of retina. The main contribution of this thesis is fluid segmentation in AMD and DME subjects. In this work, a new method to transform images from pixel to neutrosophic domain and 2 automated methods for fluid segmentation are presented in neutrosophic domain. For fluid segmentation in DME subjects, 4 layers are segmented with graph shortest path methods in which new definitions of graph weights are proposed. In the next step, a new cost function is designed and minimized which leads to image clustering. For fluid segmentation in AMD subjects, 2 layers are segmented and flattened. Then, seed points for fluid and tissue regions are initialized automatically followed by proposing a new cost function for graph cut in kernel space. Cost function minimization leads to binary segmentation of images into fluid and tissue regions. Finally, the correlation of fluid regions with vision acuity are analyzed. Key Words Retina, OCT imaging, fluid segmentation, neutrosophic theory, graph theory, graph shortest path methods.
