Related Experiment Video
Updated: Jun 13, 2025
10:23
Author Spotlight: Three-Dimensional Cephalometric Landmark Annotation Demonstration on Human Cone Beam Computed Tomography Scans
Published on: September 8, 2023
2.7K
Profile Photograph Classification Performance of Deep Learning Algorithms Trained Using Cephalometric Measurements: A
Duygu Nur Cesur Kocakaya1, Mehmet Birol Özel1, Sultan Büşra Ay Kartbak1
1Department of Orthodontics, Faculty of Dentistry, Kocaeli University, Kocaeli 41190, Türkiye.
Diagnostics (Basel, Switzerland)
|September 14, 2024
Summary
Deep learning algorithms can accurately classify orthodontic patient profiles using photographs, achieving over 97% accuracy. This AI approach may reduce the need for traditional cephalometric X-rays in diagnosis.
Area of Science:
- Orthodontics
- Artificial Intelligence
- Medical Imaging
Background:
- Extraoral profile photographs are vital for orthodontic diagnosis and treatment planning.
- Current methods often rely on cephalometric radiography, which involves radiation exposure.
- Developing AI-driven alternatives can enhance diagnostic efficiency and patient safety.
Purpose of the Study:
- To assess the efficacy of deep learning algorithms in classifying orthodontic patients based on extraoral photographs.
- To correlate photographic classifications with established cephalometric measurements.
- To explore AI's potential to minimize reliance on cephalometric X-rays.
Main Methods:
- Utilized cephalometric radiographs and profile photographs of 990 orthodontic patients.
- Performed cephalometric measurements (FH-NA, FH-NPog, FMA, N-A-Pog) using Webceph software.
- Trained 14 deep learning models on patient photographs grouped by cephalometric values.
Main Results:
- Achieved high accuracy rates in classifying patient images: up to 96.67% for FH-NA, 97.33% for FH-NPog, 97.67% for FMA, and 97.00% for N-A-Pog.
- Demonstrated strong predictive performance of the trained deep learning models.
- Validated the correlation between cephalometric data and AI-based photographic classification.
Conclusions:
- Deep learning models trained on cephalometric data show high accuracy in classifying orthodontic patient profiles from photographs.
- This AI approach presents a promising, potentially radiation-free alternative for orthodontic diagnosis and treatment planning.
- This study pioneers the use of AI for classifying clinical images based on objective cephalometric parameters.
