Automatic Segmentation of Structures and Registration of CT Images for Image-Guided Otologic Surgery and Implant Programming

Abstract

A cochlear implant (CI) is a neural prosthetic device that restores hearing by directly stimulating the auditory nerve using an electrode array surgically placed in the cochlea. Conventional CI implantation techniques require major excavation of the skull to achieve access and place an electrode array into the cochlea. After placement, CIs are typically programmed to attempt to optimize hearing outcome. Recently, image-guidance has been proposed to minimize the invasiveness of conventional CI surgery techniques as well as to design new strategies to improve CI programming outcomes. These image-guided techniques necessitate the automatic segmentation of the structures of the ear in pre- or post-implantation CTs, or the automatic registration of pre- and intra-implantation CTs. The structures of the ear include the facial nerve, the chorda tympani, the labyrinth, the ear canal, the tympani membrane, the ossicles, and the inner ear structures, which include the scala tympani, the scala vestibuli and the spiral ganglion.

In this dissertation, we present a set of innovative image processing techniques we have developed to achieve the necessary segmentation or registration tasks. The set of techniques includes methods for automatic segmentation of the structures of the ear in pediatric pre-implantation CT, a new pose-invariant pre- to intra-implantation CT registration method, new algorithms for automatic segmentation of the inner ear structures in post-unilateral-implantation CT, and novel shape library-based algorithms for automatic segmentation of the inner ear structures in post-bilateral-implantation CT.

All these techniques have been validated both qualitatively, by experts in ear anatomy, and quantitatively, by comparing the results they produce to expert generated results.