Non-Invasive Periodontal Disease Classification Using Thermograpy and Machine Learning: A Clinical Decision Support Approach

Periodontal diseases, including gingivitis and periodontitis, remain prevalent health issues requiring improved early detection strategies. This study aimed to develop and evaluate a non-invasive diagnostic support system combining infrared thermography and clinical features, powered by machine learning, for the classification of periodontal health status. A cross-sectional study was conducted on 91 subjects categorized as healthy, gingivitis, or periodontitis. Gingival temperature features were extracted from thermographic images taken from three facial views, complemented with clinical variables such as plaque index, age, sex, smoking status, and systemic diseases. Multiple machine learning algorithms were trained and evaluated using 10-fold cross-validation, with and without dimensionality reduction. A two-phase classification strategy yielded the best performance: logistic regression identified periodontitis cases, and XGBoost distinguished gingivitis from healthy subjects. The combined thermal and clinical feature model achieved an accuracy of 94.51% and an F1-score of 94.49%, while relying solely on thermal features reduced accuracy to 75.82%. The results highlight the strong potential of gingival thermography, supplemented by clinical data, in supporting periodontal disease classification. This study demonstrates the feasibility of AI-assisted thermographic screening as a non-invasive, accurate tool to enhance diagnostic precision and facilitate timely, personalized treatment decisions in dental practice.