Integrating Artificial Intelligence with Salivary Biomarkers for Predictive diagnosis in Systemic Diseases

Abstract

Background: Early detection of systemic diseases is essential for preventive healthcare and improved patient outcomes. Saliva, a non-invasive biofluid, contains multiple biomarkers reflecting systemic health. Artificial intelligence (AI) offers a powerful approach to analyze complex biomarker data for predictive diagnostics.

Objective: To develop and validate an AI-driven diagnostic system using salivary biomarkers for early detection of diabetes mellitus, cardiovascular disease, and chronic kidney diseases.

Materials and Methods:A cross-sectional study enrolled 300 participants, equally divided into four groups: healthy controls, diabetes mellitus, cardiovascular disease, and chronic kidney disease. Unstimulated saliva samples were collected and analyzed for glucose, cortisol, C-reactive protein (CRP), interleukin-6 (IL-6), and creatinine using ELISA. Data preprocessing, normalization, and feature selection preceded the application of Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) models. Model performance was assessed using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC).

Results:The ANN model outperformed others, achieving 91.3% diagnostic accuracy, 89.8% precision, 92.5% recall, and 0.95 AUC-ROC. RF achieved 88.7% accuracy and 0.91 AUC-ROC, while SVM reached 85.4% accuracy and 0.88 AUC-ROC. Salivary IL-6 and CRP were strong indicators of systemic inflammation, while glucose correlated with diabetes (r = 0.78, p < 0.001).

Conclusion: AI-driven analysis of salivary biomarkers enables non-invasive, early detection of systemic diseases. Integration into clinical workflows could enhance preventive healthcare. Further validation across diverse populations is recommended.