Artificial Intelligence for the Prediction of Oral Health From Breath

Objectives: Currently, diagnosis of oral disease (OD) relies on traditional clinical assessments. Yet, surfacing intelligent technologies show encouraging diagnostic alternatives. Recent publications have reported emanated volatile organic compounds (VOCs) in early stages of disease from related cells and bacteria. Early detection of pathogenic activity surrounding teeth/dental implants prevents irreversible damage. We hypothesize that each OD is associated with distinct VOCs, nanosensor (NS) detection of VOCs from exhaled breath (EB) emlpoying machine learning algorithms offers great diagnostic potential. Our objectives are to identify oral disease-specific VOC profiles from EB by training, testing and cross-validating an Artificial-Intelligence based chair-side diagnostic device in a large patient cohort.
Methods: Systematically healthy patients (n=300, >18 y.o) were phenotypically divided into 5 patient groups: 1. healthy periodontium (non-implant group) 2. healthy implants 3. periodontitis 4.peri-implantitis and 5.caries. EB were sampled and VOCs identified by gas chromatography-mass spectrometry (GC-MS) and Electronic nose. Finally, a chair-side Sniffphone device was trained and tested for disease prediction models.
Results: Preliminary results (n=121) for GC-MS analysis yielded 18 significant VOCs following group comparisons. Discriminant Functional Analysis of a laboratory-based NS cross-reactive response to the collective VOCs enabled disease separation with 89.75-100% accuracy using up to 5 feature extractions.
Conclusions: The future of dentistry lies in bridging the gap between technological innovation and oral healthcare needs. Our preliminary data strongly suggest that volatile profiling is achievable, and we believe that an AI-based chair-side device can improve early diagnostic accuracy, cut costs and mitigate oral disease outcomes.