Segmentation of Human Muscles of Mastication from Magnetic Resonance Images

Abstract

This thesis presents methods for segmenting the human masticatory muscles from magnetic resonance images that are used in maxillofacial surgeries. Human masticatory muscles directly affect oneb s ability to chew effectively and efficiently. Segmenting them is a challenging task due to close proximity between the muscles and surrounding soft tissue, and the complicated structure of the muscles. An improved watershed segmentation algorithm with unsupervised clustering was introduced to address the drawbacks of the conventional watershed algorithm. Good initializations to the GVF snake were provided automatically in a proposed model-based method while adaptive morphology was introduced in another proposed method to preserve the muscle structure. Dominant slices which together best capture the shape and area features of the muscles were determined and patient-specific muscles models were built using them. These models serve as coarse segmentations which are refined by matching distributions of pixels intensity values within muscles volumes and on muscles boundaries.