Related Experiment Video
Updated: Nov 28, 2025
Quantified Assessment of Infant's Gross Motor Abilities Using a Multisensor Wearable
Published on: May 17, 2024
Validation of cervical vertebral maturation stages: Artificial intelligence vs human observer visual analysis
Hakan Amasya1, Emre Cesur2, Derya Yıldırım3
1Department of Dentomaxillofacial Radiology, Faculty of Dentistry, Istanbul University-Cerrahpasa, İstanbul, Turkey; Corlu Oral and Dental Health Center, Ministry of Health, Tekirdağ, Turkey; Department of Dentomaxillofacial Radiology, Faculty of Dentistry, Suleyman Demirel University, Isparta, Turkey.
An artificial neural network (ANN) model was developed for cervical vertebral maturation (CVM) analysis. The ANN model demonstrated performance comparable to, or exceeding, human observers in CVM assessment.
Area of Science:
- Artificial Intelligence in Medical Imaging
- Radiographic Analysis
- Orthodontics
Background:
- Cervical Vertebral Maturation (CVM) analysis is crucial in orthodontics.
- Current methods rely on human observers, which can be subjective.
- Developing automated analysis tools is essential for improved accuracy and efficiency.
Purpose of the Study:
- To develop and validate an artificial neural network (ANN) model for Cervical Vertebral Maturation (CVM) analysis.
- To compare the ANN model's performance against human observers.
- To explore the potential of AI in replacing conventional CVM evaluation methods.
Main Methods:
- A dataset of 647 lateral cephalograms from patients aged 10-30 years was used.
- An ANN model was developed using 54 image features extracted from 26 marked points on each radiograph.
- The ANN model's output was validated against the assessments of four human observers using weighted kappa and Cohen's kappa coefficients.
Main Results:
- The ANN model achieved high agreement with human observers, with weighted kappa coefficients ranging from 0.80 to 0.91.
- The model's performance was comparable to, and in some cases exceeded, interobserver agreement among human evaluators.
- An average agreement of 58.3% was observed between the ANN model and human observers.
Conclusions:
- The developed ANN model shows significant potential for accurate CVM analysis.
- The AI-driven approach performed comparably to, or better than, human observers.
- Artificial intelligence may offer a future replacement for traditional CVM evaluation techniques.
