This study investigates a novel approach for detecting gingival inflammation using thermal imaging and artificial intelligence (AI). Thermal imaging is a technique that utilizes heat to generate detailed images, while AI assists in analyzing these images to identify patterns. Unlike traditional methods that require direct contact or visual examination, this approach is non-invasive, eliminating the need for physical interaction with the gingiva or reliance on subjective assessments. A key aspect of this study is its focus on individuals with mouth breathing, a condition that complicates gingival health monitoring. By utilizing thermal imaging, the study successfully detected and classified gingival inflammation levels (healthy, mild, moderate, or severe) based on heat distribution patterns. Additionally, specific temperature thresholds were established to differentiate between healthy and inflamed gingival tissues in this patient group, representing a novel contribution to the field. The developed AI system demonstrated high accuracy in identifying inflammation. This technology has the potential to facilitate earlier detection of gingival disease, even before clinical symptoms become evident. Furthermore, it offers a fast, painless, and reliable method for monitoring gingival health over time, enhancing accessibility and improving patient experience in dental care. These findings suggest that the integration of thermal imaging and AI could significantly improve the diagnosis and management of gingival diseases. Future research could further refine this technology by expanding the sample size and optimizing analytical models to enhance accuracy and widespread applicability.
Ethical Approval and Participant Selection This study was initiated following approval from the Ethics Committee of Gazi University (Meeting No: 13, dated 30.07.2024). Participants were selected from individuals presenting to the Department of Periodontology, Faculty of Dentistry, Gazi University. Determination of Mouth Breathing The diagnosis of mouth breathing was based on patients' medical history and clinical examination. Participants were asked about the use of oral breathing devices, whether they slept with their mouth open, and whether they experienced nocturnal awakening due to dry mouth. The following diagnostic tests were employed during the clinical examination to confirm mouth breathing: Participants were instructed to close their lips and breathe through one nostril while the other nostril was occluded. Individuals with nasal breathing exhibited effective alar muscle function, which is typically absent in mouth breathers. While participants were instructed to breathe normally, a mirror was held horizontally beneath the nostrils bilaterally. The presence of fog on the lower side of the mirror was considered an indicator of mouth breathing. After evaluating the criteria for mouth breathing, individuals meeting the criteria were enrolled in the study based on the following inclusion and exclusion parameters: Inclusion Criteria: Presence of at least 20 teeth Age between 18 and 25 years, with systemic health No history of periodontal treatment within the last six months Exclusion Criteria: Presence of any acute infection Use of systemic antibiotics or anti-inflammatory drugs within the past three months History of systemic diseases Pregnancy and/or lactation Presence of xerostomia or drug-induced gingival inflammation Current or former smoking history Acquisition of Thermal Images The thermal camera lens was fixed and focused on the gingival region being imaged throughout the recording process (Optris GmbH, PI 450, Berlin, Germany) \[5\]. After the initial thermographic image, participants were instructed to rinse with 20°C ice water for 60 seconds. Following the rinse, thermal images were captured at 1, 2, and 3 minutes. Labeling Process A total of 82 thermal images from 20 participants were resized to 640 × 480 pixels and annotated using an online licensed browser tool (Make Sense AI). A total of 867 labeled data points were generated and divided into a training set (80% of the total) and a testing set (20% of the total) . The labeling process involved tracing from the midpoint of the interdental papilla tip, following the juxta-gingival level to the midpoint of the adjacent interdental papilla, and extending vertically to the mucogingival junction. From this vertical endpoint, the boundary was traced back parallel to the starting point, returning to the projection of the initial papilla's midpoint. The boundaries were closed vertically at this projection . Assessment of Inflammation The degree of inflammation was evaluated using the Gingival Index . Images were categorized into four groups based on the highest inflammation grade observed in the region: Inflammation 0: Normal gingiva Inflammation 1: Mild inflammation Inflammation 2: Moderate inflammation Inflammation 3: Severe inflammation
Study Type
OBSERVATIONAL
Enrollment
20
The degree of inflammation was evaluated using the Gingival Index
Gazi University Faculty of Dentistry Department of Periodontology
Ankara, Çankaya, Turkey (Türkiye)
Classification Performance of AI Model in Detecting Gingival Inflammation from Thermal Imaging Data
The performance of the AI model in detecting gingival inflammation is assessed using classification metrics derived from thermal imaging data. The model's effectiveness is evaluated based on overall accuracy, precision, sensitivity (recall), specificity, and F1-score. The classification process is performed using the XGBoost algorithm, with a 5-fold cross-validation approach to ensure reliability. The final accuracy and performance metrics are calculated as the mean and standard deviation of cross-validation results.
Time frame: 6 months
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.