Related Experiment Video
Updated: May 20, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.7K
Automatic mandibular third molar and mandibular canal relationship determination based on deep learning models for
Elham Tahsin Yasin1, Mediha Erturk2, Melek Tassoker2
1Graduate School of Natural and Applied Sciences, Department of Computer Engineering, Faculty of Technology, Selcuk University, Konya, Türkiye.
Clinical Oral Investigations
|March 25, 2025
Summary
Deep learning models accurately classify the spatial relationship between mandibular third molars and the mandibular canal using cone-beam computed tomography scans. This AI-driven approach enhances surgical planning and reduces risks like inferior alveolar nerve injury.
Area of Science:
- Dental imaging and diagnostics
- Artificial intelligence in medicine
- Surgical planning and risk assessment
Background:
- Accurate classification of the spatial relationship between mandibular third molars and the mandibular canal is crucial for preventing inferior alveolar nerve injury during extraction.
- Cone-beam computed tomography (CBCT) is a key imaging modality for visualizing this anatomical relationship.
- Manual assessment of CBCT images can be time-consuming and prone to inter-observer variability.
Purpose of the Study:
- To evaluate the efficacy of deep learning models in classifying the spatial relationship between mandibular third molars and the mandibular canal.
- To compare the performance of different convolutional neural network architectures for this classification task.
- To assess the potential of AI to improve the accuracy and consistency of preoperative assessments.
Main Methods:
- A dataset of 305 CBCT scans was curated and annotated by maxillofacial radiology experts.
- Three deep learning models (MobileNet, Xception, DenseNet201) were trained and evaluated on the dataset.
- Classification performance was assessed using standard metrics, with categories including 'not contacted', 'nearly contacted', and 'contacted'.
Main Results:
- MobileNet achieved the highest classification accuracy at 99.44%.
- Xception and DenseNet201 demonstrated high performance, with accuracies of 98.74% and 98.73%, respectively.
- All evaluated deep learning models showed strong capabilities in classifying the molar-canal spatial relationship.
Conclusions:
- Deep learning models show significant potential for automating and enhancing the accuracy of mandibular third molar and mandibular canal relationship classification.
- AI-driven classification can improve surgical risk assessment and streamline diagnostic workflows.
- Integrating these AI systems may lead to safer, more efficient dental surgical planning, especially in resource-limited settings.
