Artificial intelligence guided occlusion reconstruction in nonoccluding CBCT: A validation study.
Authors
Affiliations (4)
Affiliations (4)
- PhD Researcher, OMFS-IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven; Department of Oral and Maxillofacial Surgery, University Hospitals Leuven, Leuven, Belgium; and Assistant Lecturer, Department of Prosthodontics, Faculty of Dentistry, Tanta University, Tanta, Egypt.
- Postdoctoral Researcher, OMFS-IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Leuven, Belgium; Department of Oral and Maxillofacial Surgery, University Hospitals Leuven, Leuven, Belgium; and Assistant Lecturer, Department of Prosthodontics, Faculty of Dentistry, Tanta University, Tanta, Egypt.
- Professor, Department of Stomatology, Public Health and Forensic Dentistry, Division of Oral Radiology, School of Dentistry of RibeirÃo Preto, University of SÃo Paulo (USP), RibeirÃo Preto, Brazil; and OMFS-IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Leuven, Belgium.
- Professor, Department of Dental Medicine, Karolinska Institute, Stockholm, Sweden; Full Professor and Head of OMFS-IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Leuven, Belgium; and Department of Oral and Maxillofacial Surgery, University Hospitals Leuven, Leuven, Belgium. Electronic address: [email protected].
Abstract
Cone beam computed tomography (CBCT) scans acquired with interocclusal separation improve stability and reduce motion artifacts but lack reliable occlusal contacts. Accurately reconstructing occlusion is critical for diagnosis, restorative planning, and implant workflows. While integrating intraoral scans (IOSs) can capture occlusion accurately, the validity of artificial intelligence (AI)-driven alignment algorithms for reconstructing occlusion in separated CBCT datasets has not been established. The purpose of this retrospective study was to validate an AI-based tool for reconstructing occlusion by aligning separated maxillary and mandibular CBCT segmentations using IOS-derived occlusal data, with IOS occlusion as the reference standard. Forty paired CBCT scans acquired with interocclusal separation and corresponding IOS datasets were uploaded into an AI-driven platform, which automatically segmented and registered the CBCT and IOS scans. The AI tool used IOS‑derived occlusal relationships to align the segmented CBCT models, generating 3 occluded models: IOS‑only, AI-fused CBCT‑IOS, and CBCT‑only. Occlusal contacts, occlusal intersections, occlusal contact surface area, and 3-dimensional (3D) occlusal surface deviations were analyzed. Statistical analyses were performed using the Friedman test, repeated-measures ANOVA, and the Mann-Whitney U test. Intra-operator reliability was assessed using weighted kappa (κ) (α=.05). No significant differences were observed between IOS and AI-fused CBCT-IOS models for occlusal contacts or contact surface area (median=46, mean ±standard deviation=256 ±170 mm²) and (median=48, mean=255 ±166 mm²), respectively. AI-driven CBCT models showed significantly fewer contacts (median=22, mean ±standard deviation=184 ±163 mm²). Median surface deviation between IOS and fused models was 0 µm, whereas CBCT-only models showed deviations of 70 to 80 µm. Dentition status did not influence outcomes. Intra-operator reliability was excellent (κ=0.86). The AI-driven fusion of IOS occlusal data with CBCT scans acquired with interocclusal separation accurately reconstructs occlusion, enabling reliable treatment planning without the need for CBCT acquisition in occlusion.