Segmentation of 3D brain MRI images using active contours and shape priors

Image processing is an important tool in research and treatment design in the medical industry. One area of image processing is segmentation: the process of finding objects in an image. This thesis develops a segmentation algorithm for finding the eyeballs in 3D MRI human brain images. An integral part of this is the prior knowledge of the general position, shape, and size of the eyeballs. The segmentation is done in two stages: first, the most likely position of the eyeballs is found using a shape prior (a representation of two spherical objects with specific size and distance apart), followed by a stage of more precise active contour segmentation in the relevant area. The design process of this algorithm can be used to develop a similar segmentation procedure to look for different objects in the human brain, or in other types of images.