3D-localisation of cochlear implant electrode contacts in relation to anatomical structures from in vivo cone-beam computed tomography.

Positioning of the cochlear implant (CI) electrode in relation to the anatomical structures is a key factor for the hearing outcome and the preservation of residual hearing after cochlear implantation. Determining the exact electrode's location is therefore expected to play an important role in optimisation of the electrode design, the surgical techniques and the post-operative device fitting. The aim of this study is the development and validation of a robust and efficient computerised algorithm for three-dimensional (3D) localisation of the CI-electrode contacts with respect to the relevant cochlear structures, such as the basilar membrane and the modiolus, from modern clinical in vivo cone-beam computed tomography (CBCT). In the presented algorithm, the pre- and post-implantation CBCT are spatially aligned. To localise the anatomical structures, a cochlear microanatomical template derived from lab-based X-ray computed microtomography (µCT) measurements is warped to match the patient-specific cochlear shape acquired from pre-implantation CBCT. The electrode-contact locations, determined from the post-operative CBCT, are superimposed onto the cochlear fine-structure of the microanatomical template to localise the array. The accuracy of this method was validated in a temporal bone study by comparing the distance of the electrode contacts from the modiolar wall, as derived by the algorithm from CBCTs, with the distance determined from synchrotron-radiation (SR) µCT on the same specimens. Due to the achievable spatial resolution, good tissue contrast and limited presence of metallic artifacts, the SRµCT technique is considered to be a golden standard in the proposed approach. In contrast to other approaches, this validation method allowed for the evaluation of the final electrode-to-modiolus distance (EMD) error, and covers the error in co-alignment of the images, in the determination of the electrode contact location and in the localisation of the cochlear structures. The absolute mean error on the EMD parameter was determined at 0.11 mm (max = 0.29 mm, SD = 0.07 mm) across five samples, slightly lower than the voxel size of the CBCT-scans. In a retrospective study, the algorithm was applied to identify scalar translocations of the electrode from clinical in vivo CBCT datasets of 23 CI-recipients, which showed perfect (100%) agreement with the blinded opinion of two experienced neuroradiologists.