Prediction of skeletal maturity using machine learning based on multiple biological indicators.

Area of Science:

• Orthodontics
• Radiology
• Machine Learning

Background:

• Accurate skeletal maturity assessment is crucial for orthodontic treatment planning and growth modulation.
• Traditional methods like hand-wrist radiographs can be time-consuming and may not always be practical.
• Developing efficient and accurate prediction models is essential for optimizing patient care.

Purpose of the Study:

• To develop a comprehensive machine learning (ML) model for predicting skeletal maturation stages.
• To evaluate the predictive performance of ML models integrating cervical vertebral morphology, dental maturation stage (DMS), gender, and age.
• To compare the efficacy of different ML algorithms for skeletal maturity prediction.

Main Methods:

• Utilized a dataset of 860 patients with lateral cephalograms, panoramic, and hand-wrist radiographs.
• Compared a baseline model with enhanced models incorporating DMS, gender, and age.
• Evaluated six ML algorithms, with CatBoost demonstrating superior performance, and analyzed feature importance.

Main Results:

• Incorporating DMS, gender, and age significantly improved predictive accuracy compared to cervical morphology alone.
• The CatBoost model achieved high performance metrics, including an AUC of 0.924 and an F1-score of 0.752.
• Dental maturation stage, particularly the mandibular second molar, was identified as a strong predictor of skeletal maturity.

Conclusions:

• A novel and clinically practical ML model for skeletal maturity prediction was developed using routinely acquired orthodontic records.
• Combining dental maturation stage with cervical morphology, age, and gender enhances prediction accuracy.
• This ML model provides a reliable and efficient alternative to traditional hand-wrist assessments for determining skeletal maturity.