Early childhood caries; using Machine Learning algorithms to identify leading risk factors

Background and aims: Study aimed to integrate clinical, behavioral, social and laboratory datato find leading risk factors of Early Childhood Caries in young children (۲۴-to-۳۶-month-old)living in Tehran Iran using Machine Learning models.Method: This ML analysis was carried out on the baseline results of a cluster-randomized communitytrial in Tehran, Iran, where ۲۳۹ mother-child dyads who enrolled in public health careprograms were selected through stratified cluster random sampling in ۲۰۱۲. Children’s teeth wereexamined using ICDAS-II, dmf, and PI. Mothers filled out questionnaires about their background,SES, and oral hygiene habits and the children went to oral examination. In ۲۰۲۱, data were reanalyzedusing four algorithms (Extreme Gradient Boosting /XGB, Random Forest/ RF, AdaptiveBoost /ADB, and Support Vector Machine/ SVR) to find the hierarchy percentage of the leadingfactors to ECC. Python ۳.۸.۵ and SPSS ۱۹.۰ were the soft wares to analyze the data.Results: XGB was the most appropriate algorithm for these analyses. The age of cleaning children’steeth starts and visible plaque had ۱۱.۶۱% and ۹.۹۲% fostering effect among ۶۲ risk factors,respectively, according to XGB. Heat maps of each model illustrated the correlation ofvariables with our defined target (ICDAS۰) which formulated and calculated by the authors. Itwas the proportion of the sound tooth surfaces in an individual.Conclusion: Machine Learning is recommended to design more effective and tailored oral healthpromotion interventions and to make better policies, regarding its power to find and prioritize thefostering factors of oral diseases.

کلیدواژه ها:

artificial intelligence ، machine learning ، Early Childhood Caries ، ICDAS