Avaliação da oclusão em adolescentes: estudo epidemiológico e uma proposta para classificação da gravidade da má oclusão utilizando a inteligência artificial

Abstract

Epidemiological studies in oral health showed that malocclusions represent the third oral public health problem and therefore it is essential to investigate the prevalence of such conditions before defining strategies and interventions in oral health. This study aimed to evaluate the occlusal characteristics of the adolescents of Montes Claros, MG, to estimate the prevalence of malocclusion and associated factors. In addition, to construct a model for diagnose of the malocclusion severity using the Artificial Neural Network and to test its success rate. To achieve the objectives two studies were conducted. Study 1 aimed to estimate the prevalence of occlusal changes and malocclusion of adolescents according to sex, caries experience, presence of gingival bleeding and smile dissatisfaction. For this, an epidemiological survey was conducted with 600 adolescents, with 10 to 16 years old, of both sexes. In order to collect data on oral health, intra and extraoral examinations were carried out in the school environment by three calibrated researchers with satisfactory agreement. The following aspects were considered: number of decayed, missing or filled teeth, presence of gingival bleeding and condition of dental occlusion, according to the criteria established by the World Health Organization. The prevalence of malocclusion was estimated in two ways, considering the parameters described in the indexes of dental aesthetics and Foster and Hamilton adapted and other including the evaluation of the dental middle line and the facial and smile pleasantness. Analyzes were conducted after correction by the sampling design effect, and prevalence was estimated for each occlusal condition investigated, and the differences were calculated according to sex, caries experience, presence of gingival bleeding and smile dissatisfaction. For the analysis of the data, the Poisson regression was used, with robust variance, and for the calculation of the prevalence of malocclusion, it was considered the concept that, the presence of at least one occlusal alteration, defines the existence of malocclusion. The results showed that 83.5% of the adolescents declared themselves black or brown, 61.5% were female and 54% were 14 years or less. Regarding oral health data, 52.3% had dental caries experience, 55.8% had gingival bleeding and 49.4% were dissatisfied with their smiles. The most common occlusal alteration was crowding of lower anterior teeth, with a prevalence of 58.5%, followed by the deviation of the midline (54.9%), crowding of upper anterior teeth (51.5%), deep bite (47.7%) and overjet increased (46.5%). The prevalence of malocclusion was estimated at 89.2% and 96.0% according to the two forms of evaluation. There was no significant difference in sex. Class III molar ratio, deviated medial line, anterior open bite and anterior crossbite were associated with the caries experience. The caries experience was lower in subjects who presented diastema (PR = 0.81). The prevalence of malocclusion was higher in adolescents with gingival bleeding (PR = 1.39) and anterior crowding were associated with the presence of gingival bleeding. The dissatisfaction with the teeth was associated with smiles classified as acceptable and unpleasant. Based on these results, it is concluded that malocclusion presents an expressive prevalence among the adolescents of Montes Claros / MG and its associated with the experience of caries, presence of gingival bleeding and dissatisfaction with the smile. Study 2 aimed to construct a model using the artificial neural network, a computational intelligence resource to determine the diagnosis of malocclusion severity and to test the success rate of this. For the construction of the model the data of the 20 variables that characterized the occlusion condition, of the 600 adolescents evaluated in the first study were inserted as input data. Outcome data were the four classification levels of severity of malocclusion: 1. Normal occlusion, 2. Mild malocclusion, 3. Moderate malocclusion, 4. Severe malocclusion, obtained after a thorough interpretation of the data, following a flowchart built in the study. Data from 449 adolescents were used to perform the training or learning phase, and data from the remaining 151 were used for the test phase. Data were divided in the training and test sets, prior to the study and the responsible researcher was blinding. Of the 600 adolescents examined, 24 (4%) had normal occlusion, 160 (26.6%) had mild malocclusion, 285 (47.6%) had moderate malocclusion and 131 (21.8%) had severe malocclusion. The model success rate was 99.3%. The results of this study demonstrated that the artificial neural network was effective in determining the epidemiological diagnosis of the severity of malocclusion, with 99% accuracy. Therefore, it is suggested that the use of neural artificial network resources can be useful as a new tool in scientific research, especially in epidemiological surveys wich includes evaluation of the dental occlusion.