An ExploratoryStudy Investigating the Intraoral Thermal Features in Advanced Periodontitis Using Machine Learning Techniques

Background/Objective: Periodontitis is a chronic inflammatory disease requiring reliable assessment of periodontal tissue involvement. This study evaluated whether quantitative intraoral thermal features could distinguish Stage III/IV periodontitis from periodontal health using machine-learning methods. Materials and Methods: The study included 108 participants, comprising 54 with Stage III/IV periodontitis and 54 healthy controls. Five standardized intraoral thermal views were obtained from each participant, yielding 540 thermal samples. Mean ROI temperature, maximum–minimum temperature variation, thermal asymmetry, and ROI temperature variance were extracted. Eleven machine-learning classifiers were evaluated using thermal features only. Participant-level stratified five-fold cross-validation was applied, while PPD, CAL, and SBI were analyzed separately for clinical comparison. Results: Higher thermal measurements were observed across all five intraoral views in the periodontitis group, accompanied by differences in PPD, CAL, and SBI. XGBoost achieved the highest accuracy (98.15%), followed by Random Forest (97.22%), Logistic Regression (96.30%), AdaBoost (96.30%), and LightGBM (95.37%). The extracted thermal features provided useful discriminatory information between the two groups. Conclusion: Intraoral thermal imaging captured quantitative thermal patterns associated with Stage III/IV periodontitis and showed potential as a non-invasive, non-contact adjunct to periodontal assessment. Manual ROI selection and the absence of independent thermal calibration and explicit reflection-artifact correction limit interpretation of absolute temperature values. Larger multicentre studies are warranted.