Machine learning in orthodontics: Transforming Invisalign treatment planning through precision, interpretability, and ethical practices.

Abstract

The integration of artificial intelligence (AI) in healthcare, particularly in orthodontics, is evolving rapidly. This study leverages a unique dataset from Thai patients undergoing Invisalign treatment to explore the synergy between AI and clinical orthodontics. This research aims to augment the predictability and personalization of Invisalign treatment plans via advanced machine learning (ML) models. The focus is on enhancing clinical decision-making by predicting treatment outcomes, identifying key influencing factors, and improving the models' interpretability and explainability within a framework of ethical AI. We analyzed 657 de-identified patient records from five dental clinics in Thailand. ML techniques, including Decision Trees (DT), Random Forest (RF), XGBoost, and Artificial Neural Networks (ANNs) were employed. Emphasis was placed on model transparency using SHapley Additive exPlanations (SHAP), integrating clinical expertise with predictive analytics to deepen understanding of treatment dynamics. XGBoost outperformed other models in predicting Invisalign outcomes, achieving an accuracy of 93.94%, sensitivity of 97.12%, specificity of 90.00%, and F1-score of 94.39%. SHAP analysis enhanced interpretability, offering detailed insights into how clinical and demographic features influence predictions. This research advances the precision of orthodontic treatment predictions significantly and pioneers the ethical application of AI in orthodontic care. By improving model transparency and accountability, the study cultivates trust among stakeholders and enhances the overall effectiveness and satisfaction associated with treatment. This work sets a new benchmark for data-driven, patient-centric orthodontic care using a patient dataset from Thailand.