Evaluation of Autoregressive Models for Predicting Two-Dimensional Mandibular Landmark Displacement During Pubertal...

Area of Science:

• Craniofacial development and growth prediction.
• Biometric analysis and cephalometrics.
• Computational modeling in orthodontics.

Background:

• Accurate prediction of mandibular growth is crucial for orthodontic treatment planning.
• Traditional methods often lack precision in forecasting individual growth trajectories.
• Autoregressive modeling offers a novel approach to analyze and predict developmental changes.

Purpose of the Study:

• To evaluate the efficacy of autoregressive models in predicting mandibular landmark displacement during pubertal growth.
• To assess factors influencing prediction accuracy, including sex, prediction interval, and landmark location.

Main Methods:

• Utilized longitudinal cephalometric data from 225 subjects (ages 10-18).
• Applied autoregressive models to predict 2D coordinates of mandibular landmarks (Condylion, Gonion, Menton, Pogonion, Point B).
• Measured accuracy using mean absolute error and percentage of predictions within 2mm, analyzing effects of various parameters.

Main Results:

• Females showed higher predictive accuracy (65%) compared to males (49%).
• Accuracy decreased with longer prediction intervals (66% for 1-year, 46% for 4-year).
• Condylion (71%) was most predictable, Menton (40%) least predictable; coordinate-based assessment superior to Euclidean distance.

Conclusions:

• Autoregressive models provide clinically relevant short-term mandibular growth predictions.
• Accuracy is influenced by prediction interval, sex, and specific anatomical landmarks.
• Condylion offers the highest predictability, while Menton exhibits significant variability.