Artificial Intelligence Assisted Thermal Imaging for Gingival Inflammation Assessment: A Novel Approach.

BACKGROUND

The integration of thermal imaging with artificial intelligence (AI) offers a novel, non-invasive approach for assessing gingival inflammation. While thermal imaging has been widely applied in other medical fields, its use in evaluating gingival health remains largely unexplored. This study is the first to utilize AI-supported analysis of thermal gingival images in patients with mouth breathing habits, aiming to detect and classify gingival inflammation severity. This research establishes specific thermal thresholds for gingival health and disease in this unique population.

METHODS

Forty participants were included, stratified according to periodontal status and clinically confirmed breathing pattern (mouth or nasal breathing), under standardized imaging conditions. From these participants, 160 images were annotated, producing 1734 labeled data points categorized according to bleeding on probing (BoP) for diagnosis, with Gingival Index (GI) applied only for secondary stratification of inflammation severity. Preprocessing included image resizing, outlier removal, and calculation of mean RGB values. The XGBoost algorithm was used for classification, with hyperparameters optimized via grid search and 5-fold cross-validation to ensure robust model performance.

RESULTS

The XGBoost Achieved Outstanding Classification Results, With an Accuracy of 92.74%, Precision of 92.95%, Sensitivity of 92.74%, and an F1 Score of 92.78%. Cross-Validation Confirmed the Model's Reliability, With Mean-Test and Validation-Scores of 88.28% and 89.43%, Respectively.

CONCLUSIONS

This study represents the first application of AI-supported thermal imaging for evaluating gingival inflammation in mouth breathers, marking a significant step forward in periodontal diagnostics. By establishing specific thermal thresholds in this unique population, it highlights the potential of this innovative approach as a non-invasive, real-time, and scalable diagnostic tool. Future research should focus on refining AI algorithms and expanding datasets to enhance clinical applicability, paving the way for advanced diagnostics and personalized care in periodontology.