Related Experiment Video
Updated: May 22, 2026
10:23
Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
Multi-class anatomical landmark detection in periapical radiographs with deep learning
Cansu Buyuk1, Alperen Saruhan2, Fatma Yuce3
1Department of Oral and Maxillofacial Radiology, Faculty of Dentistry, Istanbul Okan University, Aydıntepe Mah. Prof. Dr. Necmettin Erbakan Cad. No:2, Istanbul, 34947, Turkey. cansu.buyuk@okan.edu.tr.
Odontology
|May 21, 2026
Summary
This study developed a deep learning model to detect anatomical landmarks on dental X-rays. The model shows promise for improving radiographic interpretation accuracy and consistency in dental practice.
Area of Science:
- Radiology
- Artificial Intelligence
- Dental Imaging
Background:
- Accurate identification of anatomical landmarks in periapical radiographs is crucial for dental diagnosis and treatment planning.
- Variability in anatomy and projection techniques can challenge manual landmark identification.
Purpose of the Study:
- To develop and evaluate a deep learning model for automated detection and segmentation of multiple anatomical landmarks on periapical radiographs.
- To assess the model's performance across various anatomical structures and radiographic conditions.
Main Methods:
- A dataset of 1930 periapical radiographs with 21 annotated landmarks was used.
- A YOLOv8x-seg architecture was trained for multi-class detection and instance segmentation.
- Performance metrics included precision, recall, F1-score, Dice coefficient, and IoU.
Main Results:
- The model achieved an overall precision of 0.820, recall of 0.725, and F1-score of 0.769.
- High accuracy was observed for well-defined landmarks like the maxillary sinus.
- Performance was reduced for low-contrast or small structures, with optimal detection at low confidence thresholds.
Conclusions:
- Deep learning models can reliably detect key anatomical landmarks on periapical radiographs.
- The developed model supports safer, more consistent radiographic interpretation in routine dental practice.
- Further refinement is needed for small or low-contrast anatomical features.
