ฐานข้อมูลและระบบคาดการณ์สุขภาพช่องปากและแนวทางในการรักษาทางทันตกรรมในผู้ป่วยสูงอายุ

Oral health care for elderly patients is complicated. Before treatment, it is necessary to evaluate various factors to guide the proper treatment planning. This study is an introduction to Bayesian networks for manages this problems. An objective in this study was developing a database to diagnose oral health in elderly patients using Bayesian networks and guide for dental treatment in elderly patients. This study is a pilot research, which is the use of artificial intelligence technology in dentistry. Research design, which is a retrospective, was using data from patient record and data operations as program development with a convolution neural network in collaboration with the Bayesian networks. Programmer was write code using python. The researchers set out the framework of the different variables that must be used as a diagnostic basis for oral health in elderly patients. Collect data set by information from patient record in General Dentistry Department, Faculty of Dentistry, Srinakharinwirot University. 400 record files with panoramic radiography, then 80 percent of the samples were made into a program training dataset, and 20 percent of the samples were used as precision test data set by developing computer programs for diagnosing periodontal diseases from panoramic film using convolution neural network. Multifactor affecting oral health plans in the elderly was entering on, such as systemic diseases, medications, etc. Oral health information including caries, retained root, and dental treatment, including preventive dentistry, risk of systemic diseases, chief complaint treatment, tooth extraction or oral surgery, prosthodontic treatment, periodontal treatment, to construct Bayesian networks model. In order to analyze the accuracy of model, five-fold cross-validation is used, with the metric for analysis being average cross validation score. For evaluate the model with test sets is the validation precision. The results show that to develop of The Bayesian networks model. The factors have been affecting oral health treatment planning in elderly patients include gender, age, socioeconomic, functional status, smoking, drinking, systemic diseases, medications. The primary factors were including chief complaint, oral hygiene, diet, dry mouth, etc. Oral health information including caries, caries exposed pulp, retained root, periodontitis, and reduce posterior support. Adjust all data to the form of a Bayesian network structure. The probability of oral health treatment planning in elderly patients has been analyzed. The five-fold cross-validation showed that the development of model for diagnosing periodontal disease from panoramic film using convolution neural network was 80.5 percent of the validation value. While as Bayesian network model, the validation score was 87.5 percent, and finally, when this model is tested on a test data set, the validation precision value was 71.4 percent.