A Multiclass Classification Model for Tooth Removal Procedures.

Area of Science:

• Biomedical Engineering
• Data Science in Dentistry

Background:

• Limited understanding of tooth extraction mechanics due to data acquisition challenges.
• Need for objective data to improve dental education and clinical practices.

Purpose of the Study:

• To develop a machine learning model for classifying tooth removal procedures.
• To analyze force, torque, and movement data for improved understanding of tooth extractions.

Main Methods:

• Utilized a robotic measurement setup to collect high-quality force, torque, and movement data during tooth removal on cadavers.
• Engineered and selected clinically relevant features from the collected data.
• Trained a Gaussian naive Bayes multiclass classification model on 110 tooth removal experiments.

Main Results:

• The classification model achieved 86% accuracy on the training set and 54% on the test set.
• Successfully classified the jaw (upper/lower) with 95% accuracy and tooth class with 88% accuracy.
• Identified 33 key features from an initial set of 75 for the classification model.

Conclusions:

• Despite a small dataset, the developed machine learning model demonstrates reasonable performance in classifying tooth removal procedures.
• This study represents a significant advancement in understanding complex tooth extraction mechanics.
• The findings have the potential to inform the creation of evidence-based educational materials and clinical guidelines for dentistry.