Integrated Predictive Analysis for Periodontal Disease and Malocclusion Detection in Dentistry using Deep Learning and CNN-based Decision Making

Abstract

The complete digitalization of the healthcare industry, particularly in the field of dentistry, is necessary to ensure timely and accurate diagnoses, effective patient management, and reliable predictive capabilities. Untreated oral conditions have the potential to cause significant discomfort and harm to the teeth. This study suggests employing digitalization and machine learning classifiers as a potential solution to mitigate the effects of Periodontitis. Effective decision-making is a critical aspect of dentistry, particularly in the areas of treatment planning, management, and chairside efficiency. The application of computer resources and technologies is being optimised through the gradual replacement of manual processes with computer-assisted decision-making, thereby ensuring effectiveness and efficiency. The utilisation of artificial intelligence and virtual reality has led to notable advancements in disease detection, identification, diagnosis, pre- and post-treatment planning, patient management, and computer-assisted surgeries. The present study suggests the utilisation of convolutional neural networks (CNN) as a potential solution to tackle the issue of malocclusion, a condition that results in anomalous positioning of teeth and jaws, thereby impacting the facial appearance during smiling. This work we employ a multi-factor analysis (MFA) model, cross-validation techniques, feature extraction, and ensemble learning to conduct predictive analysis on chronic localised and chronic generalised periodontitis. The study developed a dataset comprising of 1000 patients, and assessed the classification accuracies of various classifiers including Naïve Bayes, Support Vector Machine, Random Forest, Logistic Regression, K Nearest Neighbours, and Decision Tree. The obtained accuracies were 95.5%, 100%, 100%, 100%, 99.5%, and 99%, respectively. Orthodontic treatments are a viable solution for correcting misalignment. The dataset utilised in this study comprises of RGB images depicting patients' teeth exhibiting malocclusion as well as those with properly aligned teeth. The CNN algorithm was employed to differentiate between normal and malocclusion images, yielding a precision rate of 98.95%. The implementation of this technology has the potential to aid orthodontic professionals in making informed decisions and developing accurate treatment plans for the creation of aligners and the anticipation of tooth extraction requirements.