Accuracy of Robotic Root Canal Localization Using Optimized Point Cloud Registration Strategies: An In Vitro Study.

INTRODUCTION: Traditional registration process of robotic root canal localization remains cumbersome. This study aimed to determine optimal point cloud registration strategies for robotic root canal localization and evaluate its accuracy. METHODS: 3D-printed mandibular dental models were divided into point cloud group (PCG) and fiducial point group (FPG) based on registration method. Fiducial points and target points were prepared in FPG, while only target points were prepared in PCG. After cone-beam computed tomography imaging and intraoral scanning, registration was performed using respective methods, with PCG acquiring point clouds from different tooth surfaces. Registration accuracy was evaluated by calculating fiducial registration error and target registration error. Then the robot system performed root canal localization in each group, measuring entry, apical and angular deviations. Statistics were analyzed using independent sample t-tests and one-way ANOVA. RESULTS: PCG achieved lowest registration errors when lingual surface point clouds were acquired for anterior teeth and occlusal surface point clouds for posterior teeth (P < .05). Compared with FPG, the fiducial registration error of PCG was significantly lower (P < .05), while the target registration error was comparable (P > .05). In PCG, the robotic system demonstrated entry, apical, and angular deviations of 0.25 ± 0.08 mm, 0.29 ± 0.06 mm, and 1.34 ± 0.81°, which was comparable to FPG (P > .05). CONCLUSIONS: The registration strategy utilizing point clouds from lingual surfaces of anterior teeth and occlusal surfaces of posterior teeth achieves high accuracy, feasible for robotic root canal localization. This method may enhance registration accuracy and clinical efficiency.