Evaluation of the Performance of an Artificial Intelligence-Based Classification Model for Pediatric Maxillofacial Morphology

<b>Background/Objectives</b>: Accurate assessment of craniofacial morphology is essential for orthodontic diagnosis and treatment planning. The Sassoni classification provides a useful framework for categorizing craniofacial morphology into nine groups but lacks standardized clinical criteria. This study developed an AI model to classify pediatric craniofacial morphology based on the Sassoni classification using lateral cephalometric radiographs and evaluated its agreement with classifications made by orthodontists. <b>Methods</b>: Data from 300 pediatric patients aged 6 to 10 years were analyzed. Nine cephalometric measurements and patient gender were used as input features. Three orthodontists classified morphology based on the Sassoni classification. Random forest (RF), logistic regression (LR), and support vector classification (SVC) models were trained and evaluated using 10-fold cross-validation. <b>Results</b>: The Random Forest (RF) model demonstrated the highest accuracy (RF: 0.907 ± 0.051, LR: 0.837 ± 0.057, SVC: 0.770 ± 0.055). It also outperformed the other two models in terms of F1 score, sensitivity, and positive predictive value, showing the best overall classification performance. The most influential feature was the ANB angle, while gender had minimal impact. <b>Conclusions</b>: The RF-based AI model demonstrated high accuracy in pediatric maxillofacial classification. Performance may be further improved with larger datasets and more balanced case distributions.