Related Experiment Video
Updated: Jun 6, 2026
A Postoperative Evaluation Guideline for Computer-Assisted Reconstruction of the Mandible
Published on: January 28, 2020
Comparative analysis of five AI platforms for mandibular canal segmentation on CBCT images
Sohaib Shujaat1, Razan Alotaibi2, Amal Aldakhil2
1King Abdullah International Medical Research Center, Department of Maxillofacial Surgery & Diagnostic Sciences, College of Dentistry, King Saud Bin Abdulaziz University for Health Sciences, Ministry of National Guard Health Affairs, Riyadh, Kingdom of Saudi Arabia.
Artificial intelligence (AI) for mandibular canal (MC) segmentation shows varied accuracy across platforms. Relu Creator and 3D Slicer offer near-expert performance, crucial for preventing nerve injury in oral surgery.
Area of Science:
- Oral and Maxillofacial Surgery
- Radiology
- Medical Imaging Analysis
Background:
- Accurate identification of the mandibular canal (MC) on cone-beam computed tomography (CBCT) is critical for preventing inferior alveolar nerve injury during oral and maxillofacial procedures.
- Manual segmentation of the MC is time-consuming and prone to operator variability.
- Artificial intelligence (AI) presents a promising alternative for automated and reproducible MC segmentation.
Purpose of the Study:
- To compare the accuracy of automated MC segmentation across five distinct AI platforms.
- To evaluate both quantitative and qualitative performance metrics of different AI segmentation tools.
- To provide evidence-based guidance on the reliability of AI for MC identification in surgical planning.
Main Methods:
- Analysis of 120 anonymized CBCT scans (240 MCs) using five automated AI platforms: Atomica, BlueSkyPlan, Craniocatch, 3D Slicer, and Relu Creator.
- Quantification of accuracy using unsigned mean surface deviation, with categorization into optimal (<0.5 mm), acceptable (0.5-2.0 mm), and unacceptable (>2.0 mm) ranges.
- Qualitative assessment using a five-point anatomical fidelity scale, alongside segment-wise, laterality, and scanner-wise effect analysis.
Main Results:
- Significant performance disparities were noted among the AI platforms (p < 0.001).
- Relu Creator and 3D Slicer demonstrated the highest accuracy (≈0.5 mm), with no deviations exceeding 2.0 mm.
- Atomica and BlueSkyPlan showed greater variability and more unacceptable deviations, while Craniocatch exhibited moderate accuracy.
Conclusions:
- AI-based MC segmentation accuracy is platform-dependent.
- Relu Creator and 3D Slicer demonstrate clinical suitability, while other platforms necessitate expert review.
- Independent benchmarking and multi-scanner validation are crucial for the safe clinical adoption of AI tools for MC segmentation.
