基于表面增强拉曼技术的口腔微生物检测分析  

Oral Microbial Detection and Analysis Based on Surface-Enhanced Raman Spectroscopy

作　　者：石嘉欣 黄靖 王佳睿 黄炜雅 刘刚 胡亚 陈帅[2] 黄祖芳 SHI Jiaxin;HUANG Jing;WANG Jiarui;HUANG Weiya;LIU Gang;HU Ya;CHEN Shuai;HUANG Zufang(College of Photonic and Electronic Engineering,Fujian Normal University,Fuzhou 350117,China;Fujian Key Laboratory of Oral Disease Research,Fujian Engineering Research Center for Oral Biomaterials,Fujian Key Laboratory of Oral Medicine in Higher Education Institutions,School of Stomatology,Fujian Medical University,Fuzhou 350002,China;College of Life Sciences,Fujian Normal University,Fuzhou 350117,China)

机构地区：[1]福建师范大学光电与信息工程学院,福建福州350117 [2]福建省口腔疾病研究重点实验室,福建省口腔生物材料工程技术研究中心,福建省高校口腔医学重点实验室,福建医科大学口腔医学院,福建福州350002 [3]福建师范大学生命科学学院,福建福州350117

出　　处：《福建师范大学学报(自然科学版)》2025年第2期110-116,共7页Journal of Fujian Normal University:Natural Science Edition

基　　金：国家自然科学基金项目(62275049);福建省自然科学基金项目(2022J02024、2021J01278)。

摘　　要：口腔微生物检测在口腔疾病预防、诊断和个性化治疗中具有重要作用。常规检测方法存在操作耗时、繁杂等问题,难以满足对口腔微生物的快速、精准检测需求。利用表面增强拉曼散射(surface-enhanced Raman scattering,SERS)技术,开展口腔常见微生物的高效、高灵敏度检测。通过优化SERS检测参数,获得金黄色葡萄球菌(Staphylococcus aureus,S.aureus)、变异链球菌(Streptococcus mutans,S.mutans)和粪肠球菌(Enterococcus faecalis,E.faecalis)高灵敏度检测。详细分析比较了3种细菌的SERS谱峰归属及光谱响应差异,进而评估传统机器学习与一维卷积神经网络模型在细菌分类中的表现。结果显示,决策树分类器的准确率为93%,梯度提升分类器提高至97%,而一维卷积神经网络模型准确率达100%。The detection of oral microorganisms plays a critical role in the prevention,diagnosis,and personalized treatment of oral diseases.Conventional detection methods are often time-consuming and labor-intensive,making it challenging to meet the demand for rapid and accurate identification of oral microorganisms.This study aims to leverage surface-enhanced Raman scattering(SERS)technology for efficient and highly sensitive detection of common oral microorganisms.By optimizing SERS detection parameters,we achieved high-sensitivity detection of Staphylococcus aureus,Streptococcus mutans,and Enterococcus faecalis.We thoroughly analyzed and compared the SERS spectral peak assignments and spectral response differences among these three bacterial species and evaluated the performance of traditional machine learning models and a one-dimensional convolutional neural network(1D-CNN)in bacterial classification.The results demonstrated that the decision tree classifier achieved an accuracy of 93%,while the gradient boosting classifier increased the accuracy to 97%.Remarkably,the 1D-CNN model reached a classification accuracy of 100%.

关 键 词：拉曼光谱 金黄色葡萄球菌 变异链球菌 粪肠球菌 机器学习 深度学习 

分 类 号：Q433.4[生物学—生理学] Q939.9