Augmented reality in cochlear implantation surgery
Loading...
Date
Journal Title
Journal ISSN
Volume Title
Publisher
SPIE, the international society for optics and photonics
Department of Engineering
Department of Clinical Neurosciences
https://doi.org/10.1117/12.3083584
Department of Engineering
Department of Clinical Neurosciences
https://doi.org/10.1117/12.3083584
Abstract
Description
Cochlear implantation requires drilling through the temporal bone and navigating narrow anatomical corridors near critical structures such as the facial nerve. While image-guided systems have the potential to improve safety and accuracy, their adoption in otologic surgery remains limited due to reliance on invasive fiducials or specialised tracking hardware. This pilot study proposes a minimally invasive, augmented reality (AR) framework designed to overlay anatomical structures, segmented from preoperative CT, in the surgical microscope’s field of view. At its core is a silhouette-based, 3D-2D registration pipeline, reformulated as a surface-to-surface alignment task. Contours of visible anatomy, such as the pinna and the incus, are traced in 2D, back-projected into 3D to form projector surfaces, and aligned with segmented CT surfaces using a modified Iterative Closest Point algorithm. To overcome local minima and improve robustness under occlusion, the framework adopts a multi-stage strategy: it begins with rough alignment based on external landmarks (e.g. the pinna) and progressively refines the registration by incorporating internal structures (e.g. the incus) as they are exposed during the procedure. The approach is validated using both synthetic renderings and real microscope images of 3D-printed models, paired with surfaces segmented from high-resolution micro-CT and pseudo-clinical CT. Registration accuracy is evaluated using anatomical structures not involved in the alignment process. The framework achieves sub-millimetre accuracy, with a maximum observed error of 0.52 mm. This compares favourably with values previously reported in the literature. These findings demonstrate the feasibility of achieving accurate AR overlay without invasive tracking. The software is publicly available at https://github.com/SunnyRT/AR_main and offers a lightweight solution for surgical guidance in otology and beyond. Ongoing work involves validation on cadaveric specimens to support clinical translation.