Preliminary Construction Of A Prediction Model For Kinetic Food Impaction

To collect characteristic information of the main occlusal functional area and factors related to food impaction to construct a database, and build a prediction model for kinetic food impaction using machine learning algorithms. From January to November 2024, subjects meeting the inclusion and exclusion criteria were recruited. Basic information of participants and intraoral conditions of their first and second molars were obtained through questionnaires and clinical examinations. The area ratio and position distribution of the main occlusal functional area of the tested teeth were determined by the IOS supporting system method combined with Photoshop and recorded in Excel. Position distribution information and other related factors were assigned values of "1" or "0" based on their presence. Tested teeth were also assigned "1" or "0" according to food impaction. A total of 993 molar data were included to build the database. Machine learning models were developed using SVM, DT, and MLP algorithms, and model performance was analyzed by comparing test set evaluation indicators. Models constructed by the three algorithms all showed good prediction ability, with SVM having relatively better performance. This study preliminarily constructed a machine learning model with related factors, proving the feasibility of machine learning in food impaction research.