Related Experiment Video
Updated: Apr 15, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.7K
Comparative Analysis of Third Molar Segmentation Performance Between Sexes Using Deep Learning Models
Ayşe Bulut1, Melis Büşra Aşkın2, Gökalp Çınarer3
1Department of Oral and Maxillofacial Radiology, Physiology, Faculty of Dentistry, Yozgat Bozok University, Yozgat 66100, Türkiye.
Diagnostics (Basel, Switzerland)
|April 14, 2026
Summary
Forensic identification can use artificial intelligence to determine sex from third molars in dental radiographs. This study shows that isolated third molars contain sex-related signals, offering a new method for forensic dentistry.
Area of Science:
- Forensic Odontology
- Artificial Intelligence
- Radiographic Analysis
Background:
- Sex determination is crucial in forensic identification, particularly with compromised biological evidence.
- Current AI methods often require complete dentition or craniofacial data.
- This study explores sex determination using only third molars in panoramic radiographs.
Purpose of the Study:
- To investigate the feasibility of extracting sex-based information from segmented third molars in panoramic radiographs.
- To evaluate the performance of different deep learning models for this task.
- To assess the clinical relevance of segmentation-focused AI in forensic dentistry.
Main Methods:
- A retrospective dataset of 2818 third molar annotations from 757 panoramic images was created.
- Three segmentation-based deep learning models (YOLOv12n, YOLO26n, RT-DETR v2) were evaluated.
- Detection metrics including sensitivity, recall, and mean Average Precision (mAP) were used for performance assessment.
Main Results:
- YOLOv12n achieved the highest balanced performance with an mAP@0.50 of 0.810.
- RT-DETR v2 showed higher sensitivity but longer training times and lower localization accuracy.
- Class-based analysis revealed sex-specific third molar morphology, with better detection in female samples.
Conclusions:
- Isolated third molars contain discernible sex-related information.
- Segmentation-focused AI frameworks provide an interpretable and clinically relevant approach for forensic sex determination.
- Future research should focus on larger datasets, multi-tooth integration, and explainable AI.
