Predição de tratamentos ortodônticos como estimativa do alinhamento da dentição: uma abordagem de inteligência artificial aplicada

Abstract

Objective: The study aims to apply and compare optimized intelligent models to estimate the duration of orthodontic treatments with Invisalign removable transparent aligners. The goal is to improve the accuracy of oral care planning in nursing, optimize the resources and time of health professionals, and provide better patient follow-up. In summary, the aim is to improve precision in care planning, optimize resources and time for health professionals, and provide better patient follow-up. Methodology: We carried out a quantitative study to validate and compare the effectiveness of different predictive models in estimating the duration of orthodontic treatments with Invisalign aligners. We conducted Data analysis using cross-validation to ensure the robustness of the results. Cross-validation was implemented by dividing the data set into multiple subsamples, using each subset to train and test the model in different combinations. In addition to linear regression, we evaluated other predictive models, including logistic regression and Perceptrons with different hidden layers. The performance metric used was the root mean square error (RMSE) to quantify the accuracy of the predictions. A comparative analysis identified the most suitable model regarding predictive accuracy and practical applicability in nursing. Results: Linear Regression pointed to an MSE of 1.78, showing reasonable performance. Logistic Regression, with an MSE of 5.33, had the worst performance. The Perceptron with one hidden layer had an MSE of 1.56, showing better performance. About the Perceptron with two hidden layers, the MSE was 1.61, with a slightly lower performance than the one layer. The Perceptron performed similarly to the two-layer with three hidden layers and an EQM of 1.63. The results show that the Perceptron with one hidden layer was the most effective model for predicting the data, with the lowest mean square error. This conclusion is clear and unequivocal: the Perceptron with one hidden layer is the best model choice. Logistic Regression had the worst performance, while linear Regression showed intermediate performance. Perceptrons with multiple hidden layers showed no significant improvement over the model with one hidden layer. Conclusion: The use of artificial intelligence (AI) in the evaluation of orthodontic care, as well as in health and nursing education and research, is a significant step towards improving patient care. It allows for a more robust and detailed data analysis, leading to better-informed decisions. Predictions generated by AI can inform policy and practice, ensuring a sound scientific basis for decision-making. The prediction of orthodontic treatments with AI, aligned with the theoretical-philosophical and methodological frameworks of nursing and health management and evaluation, shows how technological innovation can be effectively integrated into health contexts, promoting significant advances in the quality of care. For this data set, the simple Perceptron model with a hidden layer was the most suitable for prediction, highlighting the importance of choosing the right model to optimize resources and improve patient follow-up in orthodontic treatments.