Developing Predictive Models for Periodontitis Progression Using Artificial Intelligence: A Longitudinal Cohort Study

Area of Science:

• Periodontology
• Biomarkers
• Machine Learning

Background:

• Periodontitis is a common inflammatory disease affecting the gums and supporting bone.
• Early detection and prediction of periodontitis progression are crucial for effective management.
• Machine learning offers potential for developing advanced predictive models.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting periodontitis progression.
• To compare the performance of different machine learning algorithms (LR, MLP, PGM).
• To identify key clinical and salivary factors influencing periodontitis progression.

Main Methods:

• Utilized data from a 12-month multi-center longitudinal study of periodontally healthy and periodontitis participants.
• Collected clinical, demographic, and salivary analyte data (10 analytes).
• Applied Logistic Regression (LR), Multi-Layer Perceptron (MLP), and Probabilistic Graphic Models (PGM); assessed performance using AUROC and SHAP values.

Main Results:

• The PGM model, incorporating clinical measures, saliva IL-1β, age, and sex, achieved the highest performance (AUROC = 0.88).
• PGM demonstrated balanced sensitivity (0.55) and specificity (0.81), outperforming LR (AUROC = 0.72) and MLP (AUROC = 0.58).
• Feature importance analysis identified the number of deep periodontal pockets as a significant predictor in PGM and MLP models.

Conclusions:

• Machine learning models effectively predict periodontitis progression, supporting early detection strategies.
• Integrating clinical data with salivary biomarkers, such as IL-1β, enhances predictive accuracy.
• The PGM approach shows promise for clinical application in periodontitis management.