Integrated Raman Spectroscopy and Supervised Imaging for Label-Free Identification and Mapping of Oral Pathogens in Complex Biofilms

Résumé fourni par la source

Although Raman spectroscopy has demonstrated excellent performance for the identification of isolated bacterial cells, its application to spatially resolved analysis of complex polymicrobial communities remains a major analytical challenge. Here, we present a transferable supervised Raman imaging workflow that integrates spontaneous Raman spectroscopy, confocal Raman microscopy, and PCA-LDA chemometric analysis for the label-free identification of clinically relevant oral bacteria. The proposed framework uses Raman spectra acquired from isolated bacterial cells to train a supervised classifier that is directly transferred to hyperspectral Raman images, enabling quantitative pixel-wise bacterial identification without fluorescent labels or molecular probes. The workflow was validated using four representative oral bacterial species (Streptococcus oralis, Actinomyces naeslundii, Fusobacterium nucleatum, and Porphyromonas gingivalis), spanning the ecological transition from oral health to periodontal disease. The PCA-LDA classifier achieved 99.8% accuracy for binary classification, maintained an overall accuracy of 93.8% after extension to four bacterial species, and successfully identified P. gingivalis at bacterial ratios as low as 1:6000 within mixed populations. Importantly, the classifier, trained exclusively on planktonic single-cell Raman spectra, was directly transferred to Raman images of mixed bacterial communities and increasingly complex biofilm models without retraining, achieving classification accuracies of approximately 98%. The Raman-based classification was independently validated by morphology-based segmentation and Gram staining. These results establish a scalable and transferable analytical framework for quantitative Raman imaging of polymicrobial communities. By combining supervised machine learning with hyperspectral Raman microscopy, the proposed workflow provides a versatile strategy for label-free bacterial identification and spatial mapping, representing an important step toward Raman-based precision microbiology and future optical diagnostics of oral infectious diseases.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.