Related Experiment Video
Updated: Aug 4, 2025
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
Evaluation of an artificial intelligence-based algorithm for automated localization of craniofacial landmarks
Friederike Maria Sophie Blum1, Stephan Christian Möhlhenrich2, Stefan Raith3
1Department of Orthodontics, University Hospital of RWTH Aachen, Pauwelsstraße 30, D-52074, Aachen, Germany. frblum@ukaachen.de.
An AI algorithm for craniofacial landmark detection in cone-beam computed tomography (CBCT) shows comparable accuracy to manual methods but is significantly faster and more reproducible. This deep learning approach offers a promising tool for efficient diagnosis and treatment planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Healthcare
- Computational Anatomy
Background:
- Digitalization necessitates automated analysis of cranial structures for efficiency and data objectification.
- Accurate craniofacial landmark detection is crucial for diagnosis and treatment planning in dentistry and orthodontics.
Purpose of the Study:
- To train and evaluate a deep learning algorithm for automated craniofacial landmark detection in cone-beam computed tomography (CBCT).
- To assess the algorithm's accuracy, speed, and reproducibility compared to manual landmark determination.
Main Methods:
- A deep learning algorithm was trained on 931 CBCT scans.
- The algorithm's performance was tested on 114 CBCT scans, with 35 landmarks identified automatically and manually by three experts.
- Analysis included measurement error, time efficiency, and intraindividual landmark localization variability.
Main Results:
- The AI algorithm demonstrated no statistically significant difference in accuracy compared to manual methods.
- The AI was 95% faster and achieved a 2.12% lower mean error (2.73 mm) than human experts.
- Superior performance was observed in the detection of bilateral cranial structures.
Conclusions:
- Automated craniofacial landmark detection using AI is accurate, reproducible, and time-efficient, falling within clinically acceptable ranges.
- The developed algorithm offers a viable alternative to manual landmark determination, reducing workload in clinical practice.
- Further development and database expansion could enable widespread adoption in routine CBCT analysis.
