Artificial Intelligence–Based Prediction of Periodontal Breakdown Using Salivary Biomarkers and Radiomic Fusion

Abstract

Periodontal disease is a prevalent chronic inflammatory condition that can lead to progressive periodontal breakdown and tooth loss if not diagnosed and managed early. Conventional diagnostic approaches rely on clinical and radiographic assessments, which often detect disease at an advanced stage. Recent advances in artificial intelligence (AI) offer the potential to integrate multimodal data for predictive modeling. This study explores the use of AI algorithms to predict periodontal breakdown by combining salivary biomarkers with radiomic features extracted from dental imaging. Salivary biomarkers reflecting inflammatory and tissue-destructive processes were quantified, while radiomic analysis captured subtle structural patterns from imaging data. Machine learning models were trained on fused biomarker–radiomic datasets to assess predictive performance. Results demonstrated that multimodal AI-based prediction outperformed single-modality models, showing higher accuracy and sensitivity in identifying early periodontal deterioration. These findings suggest that AI-driven integration of biochemical and imaging data could enable early, personalized interventions, improving periodontal disease management.