A. M. I. C. K. Jayathilake, L. S. Nawarathna, P. N. P. S. Nagarathne
Abstract
Malocclusion is concerned with the relationship between the upper and lower arches. When the teeth of one arch are misplaced relative to teeth of other arches in the Antero-posterior transverse or vertical planes of space, they are referred to as malocclued teeth. There are two factors, general and local, contributing to developing the mal-alignment and malocclusion. The skeletal factor is one of the general and very important factors to identify malocclusion patterns of class I, class II and class III. It is very important to know the malocclusion/skeleton pattern of the patient before starting the treatment since the dentist decides the treatment according to the malocclusion pattern. This analysis was performed on patients who attend the clinic at the Division of Orthodontist, Faculty of Dental Sciences, University of Peradeniya, and the sample included 60 subjects, with the age of the patients ranging from 11 to 15 years. The study was carried out using the cervical vertebral maturation stage, evaluated on lateral cephalometric radiographs. Although it is a very convenient method to identify the malocclusion pattern of the patients, it is time-consuming and expensive. The main object of this study was to propose a classification model to predict the malocclusion patterns of orthodontic patients. Multinomial logistic regression model, k-NN algorithm, random forest model, neural network and naïve Bayes model were used to predict the malocclusion pattern. Accuracy of the multinomial logistic regression model, k-NN algorithm, random forest and naïve Bayes classification of malocclusion patterns were 88.89%, 83.33%, 88.89%, and 55.56% respectively. Further, Areas Under the Curve (AUC) of the multinomial logistic regression model, k-NN algorithm, random forest model and naïve Bayes classification were 0.9889, 0.6176, 0.5389 and 0.5333 respectively. The outcome of the study proposes the classification-based approach to predict the malocclusion pattern of orthodontic patients. The multinomial logistic regression model performed well in terms of both accuracy and AUC value when compared with other methods.
Keywords: k-NN algorithm, Malocclusion pattern, Multinomial logistic regression model, Naïve Bayes classification, Random forest model
**View Full Paper in the Springer Source