2025 Poster Presentations
P374: CLINICAL EVALUATION OF SLAM-BASED TRACKERLESS SURGICAL NAVIGATION IN THE LATERAL SKULL BASE
Ryan A Bartholomew, MD1; Haoyin Zhou, PhD2; Jakob Gerstl, MBBS2; Maud Boreel, MD2; Timothy R Smith, MD, PhD, MPH2; Jeffrey P Guenette, MD2; Jayender Jagadeesan, PhD2; C. Eduardo Corrales, MD2; 1Mass Eye and Ear; 2Brigham and Women's Hospital
Surgery of both the anterior and lateral skull base requires careful surgical dissection through opaque anatomy to avoid inadvertent injury of concealed neurovascular structures. Surgical navigation systems, which align medical imaging with intraoperative patient anatomy, can serve as important adjuncts for avoidance of surgical morbidity. Our group recently developed a novel approach to surgical navigation employing simultaneous localization and mapping (SLAM) algorithms with 3D endoscopy. 3D models of the operative field surface are reconstructed in real time and registered to volumetric models segmented from pre-operative imaging, all without requiring external tracking equipment. The surgeon is thereby provided with continuous information about anatomic structures underlying the exposed tissue surface. Recent validation in cadaveric models of both anterior and lateral skull base surgery have demonstrated approximately 1 mm mean surgical navigation errors. As an early test of clinical feasibility, we are retrospectively evaluating the fidelity of SLAM-based surface reconstruction using 3D-endoscopic lateral skull base surgical video from real clinical encounters. Surface reconstruction models of the operative field are generated and co-registered to volumetric CT models using corresponding landmarks. Initial pilot data from a patient who underwent a right middle fossa craniotomy for meningoencephalocele resection and tegmen dehiscence repair yielded a RMS surface-CT registration error of 1.31 mm (Figure 1). Surface reconstruction models generated at different timepoints during surgical dissection can also be co-registered, which is necessary for maintenance of continuous and uninterrupted surgical navigation. This resultant RMS surface-surface registration error was 2.28 mm for the pilot middle fossa craniotomy case. Analysis of data from additional lateral skull base clinical cases is ongoing. With further development, a SLAM algorithm-based surgical navigation system has the potential to promote surgical efficiency, economy of motion, and safety.
Figure 1: Visualization of SLAM-based surgical navigation using pilot clinical data from a right-sided middle fossa craniotomy case. Pilot clinical data was evaluated from a case of a 43-year-old woman who underwent a right middle fossa craniotomy for resection of a meningoencephalocele and repair of a tegmen dehiscence A surface model from two separate time points during surgery (panel A and B) and segmented volumetric CT model (panel C, either pre- or post-operative more generally; post-operative for this case) was generated. The surface models obtained at two surgical steps were co-registered using corresponding landmarks (blue spheres-craniotomy plate screws). The surface model was directly registered to the CT model also using corresponding landmarks (green spheres). Following co-registration, segmented CT structures are in anatomically appropriate locations when viewed through the semi-transparent surface model (panel D). Segmented structures shown include the inner ear otic capsule (yellow), facial nerve (purple), ossicles (grey), internal auditory canal (IAC, green), screws (black), craniotomy plate (sky blue), and internal carotid artery (red).