Development and validation of vision system for improved placement of cochlear implant electrode arrays

Vanderbilt University, Nashville TN

Investigators

Abstract

Project summary: Cochlear implants (CIs) are neuroprosthetic devices that are considered standard-of-care treatment for severe-to-profound sensory-based hearing loss. CIs restore hearing by electrically stimulating the auditory nerve with an electrode array that is implanted in the cochlea. While CIs are effective for most recipients, there is wide variability in outcomes. Research by many groups over the last few decades has shown that proper positioning of the electrode array within the cochlea is crucial for optimizing the electro-neural interface to maximize post-implantation speech recognition. Poor placement can lead to poor electrical coverage of the auditory nerve and/or excessive channel stimulation overlap. In this project, an augmented reality CI insertion guidance system will be developed and validated. The proposed system leverages modern artificial intelligence technology and does not require any expensive or exotic equipment. The system will be experimentally tested to confirm it facilitates achieving a more optimal CI electro-neural interface. System development will begin by developing reliable and accurate algorithms for scene mapping to facilitate microscope self-localization and registration of 2D microscope video capture to 3D pre-operative imaging. Methods will be developed for registering anatomy segmentations and pre-procedure plans created using CT or MR to the microscope video directly as well as solve the microscope self-localization and scene mapping problem to update the registration as the scene changes over time. Next, tool pose detection algorithms to enable active electrode placement feedback will be developed. We will develop machine learning approaches for detecting the 3D pose of CI insertion tools to detect how closely the pre-planned electrode array insertion strategy is being followed. This will facilitate providing warning in the AR environment when the insertion is sub- optimal. Finally, we will develop and validate a fully integrated and automated augmented reality interface. This system will enable visualization of co-registered segmentations and plans and feedback on adherence to the plan. Clinical translation of the system will be comprehensively tested in a large temporal bone study. If successful, the system will facilitate placing arrays with a more optimal electro-neural interface, which would ultimately lead to improved CI outcomes.