Orthodontic Treatment Planning based on Artificial Neural Networks

Area of Science:

• Dentistry
• Artificial Intelligence
• Machine Learning

Background:

• Orthodontic treatment planning involves complex decisions regarding extractions and anchorage.
• Accurate prediction of treatment outcomes is crucial for effective patient care.
• Artificial neural networks offer a potential solution for optimizing treatment planning.

Purpose of the Study:

• To develop and evaluate multilayer perceptron artificial neural networks for predicting orthodontic treatment plans.
• To assess the accuracy of the neural network models in determining extraction-nonextraction, extraction patterns, and anchorage patterns.
• To identify key predictive features influencing orthodontic treatment decisions.

Main Methods:

• Utilized multilayer perceptron artificial neural networks for treatment plan prediction.
• Evaluated model performance using accuracy, Area Under the Curve (AUC), sensitivity, and specificity.
• Compared different methods for handling missing discrete and continuous data, including average value and k-nearest neighbors (k-NN).

Main Results:

• Achieved 94.0% accuracy for extraction-nonextraction prediction (AUC 0.982, sensitivity 94.6%, specificity 93.8%).
• Attained 84.2% accuracy for extraction patterns and 92.8% for anchorage patterns.
• Identified "crowding, upper arch," "ANB," and "curve of Spee" as critical predictive features.
• Demonstrated superior performance of the average value method for discrete missing data and k-NN for continuous missing data.

Conclusions:

• Artificial neural networks provide a reliable and accurate tool for orthodontic treatment planning.
• The proposed method offers valuable guidance, particularly for less-experienced orthodontists.
• The AI-driven approach enhances decision-making flexibility and potentially improves treatment outcomes.