Geometric CBCT reconstruction and scanbody pose estimation for automated dental implant planning

Cone Beam Computed Tomography (CBCT) is a key imaging modality in dental implantology; however, its automated processing remains challenging due to noise, limited contrast, anatomical variability, and metal artifacts. This work presents a fully automated computational pipeline for three-dimensional reconstruction, geometric dental implant detection, and scanbody pose estimation directly from real clinical CBCT data. In contrast to learning-based approaches, the proposed method relies on geometric and structural analysis, improving interpretability and reducing the dependence on annotated training data. The methodology was evaluated on 26 real CBCT studies acquired under routine clinical conditions, comprising 203 annotated implants. The proposed pipeline achieved a precision of 1.000, a recall of 0.941, and an F1-score of 0.970 for implant detection. For scanbody pose estimation, valid alignments were obtained in 22 out of 26 cases, with a mean alignment error of 116.9 ± 41.2 μ m . The total processing time per CBCT study remained below two minutes, demonstrating computational feasibility without specialized hardware. These results suggest that the proposed approach is a robust and efficient alternative for automated CBCT processing and may support its integration into computer-assisted implant planning workflows.