Artificial intelligence as a prediction tool for orthognathic surgery assessment

Area of Science:

• Orthodontics and Dentofacial Orthopedics
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Accurate diagnosis of dental and skeletal discrepancies is crucial for ideal orthodontic treatment.
• Distinguishing between orthodontics and orthognathic surgery can be challenging, especially for borderline cases.
• Technological advancements and big data are driving the integration of artificial intelligence (AI) into clinical decision-making.

Purpose of the Study:

• To evaluate the predictive capacity of various machine learning (ML) models in determining the necessity of orthognathic surgery versus conventional orthodontics.
• To utilize soft and hard tissue cephalometric data for treatment prediction.
• To compare ML model performance against expert clinical judgment.

Main Methods:

• A dataset of 920 lateral cephalograms from patients treated with orthodontics or combined orthognathic surgery was analyzed.
• Thirty-two cephalometric measurements were extracted from each radiograph.
• Ten ML models were trained, validated, and tested on the data, with performance compared to a panel of four experts.

Main Results:

• The combined prediction from 10 ML models demonstrated high accuracy, F1-score, and AUC values on the test dataset.
• Performance was notable across the entire sample (Accuracy: 0.707, F1: 0.706, AUC: 0.791), Class II patients (Accuracy: 0.759, F1: 0.758, AUC: 0.824), and Class III patients (Accuracy: 0.822, F1: 0.807, AUC: 0.89).

Conclusions:

• The integrated 10 ML model effectively predicted the requirement for orthognathic surgery.
• The model exhibited superior performance in predicting treatment needs for Class III malocclusions.