Flexible cellulose-based plasmonic SERS-AI nanosensor for non-invasive salivary intelligent diagnosis of periodontal diseases

Periodontal disease, one of the most common oral diseases worldwide, has drawn significant attention due to the oral and systemic health risks it poses. This study developed a diagnostic platform integrating surface-enhanced Raman scattering (SERS) with artificial intelligence (AI) to enable rapid detection of the key pathogen (Porphyromonas gingivalis) responsible for periodontal disease and staging diagnosis of the condition. The study utilized interface self-assembly technology to construct positively charged SERS sensors on a cellulose paper substrate, significantly enhancing pathogen capture efficiency through electrostatic interactions (with detection limits meeting clinical saliva sample requirements) while demonstrating excellent interference resistance (effectively distinguishing free bacteria from saliva matrix). Experimental results demonstrated the platform's outstanding performance in real-world periodontitis saliva sample testing: after optimization with machine learning algorithms, diagnostic accuracy improved from 94 % to 99 %, precision increased from 91 % to 100 %, and both F1-score and Recall-score remained consistently above 0.83. Further analysis revealed that the multi-classifier integration strategy significantly improved the discriminative efficacy of disease staging. The intelligent detection system proposed in this study, based on non-invasive saliva-based diagnosis and combined with cellulose paper-based SERS sensing technology, provides an innovative solution for periodontal pathogen screening and non-invasive early disease monitoring.