Diagnostic Accuracy of Oral Images, OPGs, and Questionnaires vs. Clinical Assessment for Periodontal Disease

Diagnostic Accuracy of Oral Images, Orthopantomographs (OPGs) and Self-Reported Questionnaires vs. Clinical Assessment for Detecting Periodontal Health and Disease: a Multi-center Diagnostic Study

This is a multi-center, cross-sectional diagnostic study aimed at evaluating the accuracy of various non-invasive methods-including self-reported questionnaires, intra-oral photographs, smartphone images, intraoral scans (IOS), and orthopantomographs (OPGs)-in detecting periodontal health and disease, compared to clinical periodontal examination as the gold standard. The study will enroll 2,000 subjects across five centers, representing the full spectrum of periodontal conditions (health, gingivitis, and periodontitis stages I-IV). Participants will undergo a standardized clinical examination, radiographic imaging, and complete validated questionnaires. Machine learning models (e.g., HC-Net+ for OPGs and DLM for oral image) will be used to analyze images and integrate data domains. The primary outcome is the diagnostic accuracy (sensitivity, specificity, AUROC) of each method alone and in combination for classifying periodontal status. The study aims to validate and refine AI-based tools for scalable, efficient periodontal screening in clinical and community settings.

This is a multi-center, cross-sectional diagnostic accuracy study. The study aims to validate and compare the performance of multiple index tests against a clinical reference standard for the detection of periodontal health and disease.

The reference standard for periodontal diagnosis will be a comprehensive full-mouth periodontal examination conducted by trained and calibrated examiners. Diagnoses (periodontal health, gingivitis, periodontitis stages I-IV) will be assigned based on the integration of clinical, radiographic, and demographic data according to the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions. The decision-making algorithms proposed by Tonetti and Sanz (2019) will be applied.

The index tests under investigation include:

1. A set of self-reported questionnaires, including a modified CDC-AAP questionnaire.
2. Intra-oral clinical photographs captured with a professional camera and a smartphone.
3. A self-performed intra-oral photograph ("selfie").
4. Digital orthopantomographs (OPGs).
5. Intraoral scans (IOS). Data from the index tests will be analyzed using previously developed and validated machine learning models (e.g., HC-Net+ for OPG analysis, a deep learning model for single frontal view images). The data collected in this study will also be used to further refine these models, particularly to improve the differentiation between gingivitis/stage I periodontitis and health/stage II-IV periodontitis.

The primary analytical method will involve assessing the diagnostic accuracy of each index test, both individually and in combination, by calculating sensitivity, specificity, and the area under the receiver operating characteristic curve (AUROC) against the clinical reference standard. Logistic regression and machine learning algorithms will be employed to identify the most predictive variables and optimal diagnostic sequences.

The study will be conducted in compliance with the Declaration of Helsinki, ICH-GCP guidelines, and relevant STARD and AI-specific reporting guidelines.

Inclusion Criteria:

1. Adult patients aged 18 years or older.
2. Seeking dental care at one of the participating study centers.
3. Ability to understand and willingness to provide written informed consent.

Exclusion Criteria:

1. Edentulous patients (complete tooth loss).
2. Pregnancy or lactation.
3. History of periodontal therapy (other than supragingival prophylaxis/cleaning) within the past 12 months.
4. Use of antibiotic medication within the 3 months prior to enrollment.