AI-generated 3D models enhance CBCT interpretation of root canal anatomy among undergraduate and postgraduate students.
Authors
Affiliations (9)
Affiliations (9)
- Department of Stomatology, Public Health and Forensic Dentistry, Division of Oral Radiology, School of Dentistry of Ribeirão Preto, University of São Paulo (USP), Av. do Café, S/N, Ribeirão Preto, São Paulo, 14.040-904, Brazil. [email protected].
- OMFS IMPATH Research Group, Department of Imaging and Pathology, Faculty of Medicine, KU Leuven, Leuven, Belgium. [email protected].
- Department of Physiology, Faculty of Dentistry, Center of Excellence in Precision Medicine and Digital Health, Chulalongkorn University, Bangkok, Thailand. [email protected].
- Department of Dentistry, Universidade da Amazônia (UNAMA), Belém, Pará, Brazil.
- Department of Stomatology, Public Health and Forensic Dentistry, Division of Oral Radiology, School of Dentistry of Ribeirão Preto, University of São Paulo (USP), Av. do Café, S/N, Ribeirão Preto, São Paulo, 14.040-904, Brazil.
- Center of Excellence in Precision Medicine and Digital Health, Chulalongkorn University Implant and Esthetic Center, Faculty of Dentistry, Chulalongkorn University, Bangkok, Thailand.
- 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.
- Department of Dental Medicine, Karolinska Institutet, Stockholm, Sweden.
Abstract
To evaluate the efficacy of artificial intelligence (AI)-driven three-dimensional (3D) anatomical models as an adjunct to cone-beam computed tomography (CBCT) for root canal assessment regarding diagnostic accuracy, observer confidence, and time efficiency among undergraduate and postgraduate students. In this observational diagnostic study, 26 observers (13 undergraduates and 13 postgraduates) evaluated 22 tooth roots with complex anatomy from nine CBCT scans under two conditions: CBCT alone and CBCT supplemented with AI-generated 3D anatomical models. Observers assessed the number of roots, root canals, and apical foramina, while confidence (5-point Likert scale) and assessment time were recorded. Each observer performed 132 assessments, totaling 3,432 evaluations. A reference standard was established by consensus between two specialists. A significance level was set at 5% (α = 0.05) for all statistical analyses. Augmenting CBCT with AI-generated 3D models significantly improved diagnostic accuracy for all parameters (p < 0.001). Root detection accuracy reached 100% in both groups. Root canal detection increased from 83% to 94% among undergraduates and from 88% to 99% among postgraduates, while apical foramina detection increased to 99% in both groups. Observer confidence significantly increased (p < 0.001), reaching a median score of 5 (IQR: 5-5). Workflow efficiency also improved (p < 0.001), with median assessment time decreasing from 102 s to 39 s for undergraduates and from 97 s to 24 s for postgraduates. AI-driven 3D anatomical models used with CBCT enhance diagnostic accuracy, observer confidence, and evaluation efficiency in endodontic assessment. However, multi-centre studies with larger, more diverse samples, particularly including cases with pronounced artefacts, would further support generalisability. AI-generated 3D anatomical models derived from CBCT scans may serve as a valuable adjunct for the interpretation of complex root canal anatomy, improving diagnostic accuracy, increasing observer confidence, and reducing assessment time. These findings support their potential role not only in clinical decision-making but also as an effective educational tool for training dental students and clinicians.