Unwrapping Cochlear implants by spiral CT.

Abstract

Multielectrode, intracochlear implants were designed for individuals with profound sensorineural hearing loss who derive little or no benefit form acoustic hearing aids. Determination of each electrode's position in a patient's inner ear may improve speech processor programming to maximize speech recognition. In this paper, an approach is described to use as input a volumetric spiral computed tomography (CT) image of the Nucleus electrode array (Cochlear Pty. Ltd, Lane Cove, NSW, Australia) to unwrap it, and to measure its implanted length given starting and end points. Representative curvilinear structures were digitally synthesized in image volumes of isotropic 0.1-mm voxels. The electrode array was spirally CT-scanned in vitro and in vivo, and reconstructed on an isotropic grid in 0.1-mm steps. Two algorithms were constructed to track and measure these curvilinear structures. The first algorithm is Karhunen-Loeve (K-L)-transform based, in which the K-L transform is locally applied at a current main axis position to determine the eigenvectors of the main axis voxels, the next main axis position is estimated from the current position along the principal eigendirection, adjusted to the mass center of the orthogonal cross section passing through the estimated position, and then scaled to have a prespecified step. The second algorithm is similar to the first one but avoids use of the K-L transform. In the second algorithm, the next position is directly estimated along the local direction and then processed with the same correction and scaling operations. With user-specified starting and end points as well as a local direction at the starting point, a curvilinear structure can be automatically tracked using either of the algorithms. The first algorithm is more robust, while the second one is more efficient. In the numerical and in vitro studies, the lengths of the curvilinear structures were accurately measured. Given local directions determined in the tracking process, an electrode array image can be unwrapped into a linear array with the central electrode axis as the abscissa. The unwrapping approach allows longitudinally and cross-sectionally accurate measurement and better visualization of cochlear implant images. With preimplantation knowledge of length, width, and center electrode distance, the position of individual electrodes can be estimated after unwrapping.