Automatic Identification of the Structures of the Ear and a New Approach for Tubular Structure Modeling and Segmentation

Abstract

This dissertation presents studies on, and from, the development of algorithms for the automatic segmentation of the structures of the ear. Many otological procedures would benefit from a system that automatically identifies anatomical structures of the ear in CT. Conventional (registration-based) segmentation techniques are suitable for identifying ear structures that have high intensity contrast in CT or those for which a high degree of accuracy is not necessary. For some ear structures, conventional techniques are inadequate, and other segmentation methods must be used or developed. In this dissertation, approaches that permit the automatic identification of ear structures will be presented. This will include the ossicles, external auditory canal, cochlea, scala tympani, scala vestibuli, facial nerve, chorda tympani, semicircular canals, and the carotid artery. Each of these sensitive structures lies within millimeters of the surgical approach for various types of procedures. Out of the above mentioned structures, six distinct tubular shaped structures are represented, each of which has unique properties that make detection difficult. This has led to the development of an algorithm for the segmentation of general tubular structures, which is also presented in this dissertation. This algorithm is validated on these ear structures as well as other tubular structures outside of the ear. The results of all methods presented in this work are analyzed, and quantitatively compared to expert drawn or expert edited segmentations, which are treated as the gold standard.