应用拉曼光谱技术对伴2型糖尿病慢性牙周炎患者龈下菌斑的研究

Distinguishing subgingival plaques of patients with chronic periodontitis and type 2 diabetes mellitus based on Raman spectroscopy

目的应用拉曼光谱技术结合机器学习算法比较并区分伴或不伴2型糖尿病的慢性牙周炎患者以及健康成人的龈下菌斑。方法应用便携式拉曼光谱仪获取20例伴2型糖尿病的慢性牙周炎患者(A组)、23例单纯慢性牙周炎患者(B组)以及23例健康成人(C组)龈下菌斑的拉曼光谱图像,采用8种常见的机器学习算法构建模型,对3种类型龈下菌斑的拉曼光谱进行比较和区分。结果区分3种类型龈下菌斑拉曼光谱的最优模型是线性判别分析,区分A组和B组、A组和C组、B组和C组的最优模型分别是线性判别分析、极限树、线性判别分析。结论拉曼光谱技术结合机器学习算法构建分类模型可区分伴或不伴2型糖尿病的慢性牙周炎患者以及健康人的龈下菌斑,未来可作为筛查或诊断工具与临床实践相结合。

Objective The aim of this study is to combine Raman spectroscopy and machine learning techniques to distinguish subgingival plaques among three groups of subjects,including patients with chronic periodontitis(CP)and type 2 diabetes mellitus(T2DM),patients with CP alone,and healthy controls.Methods The Raman spectra of the subgingival plaques from 20 patients with CP and T2DM(group A),23 patients with CP alone(group B),and 23 healthy controls(group C)were obtained using a portable Raman spectrometer.Eight common machine learning algorithms were applied to build models to distinguish the Raman spectra of the three types of subgingival plaques.Results The model identified as optimal for distinguishing the three types of subgingival plaques was linear discriminant analysis(LDA).The optimal model to distinguish groups A and B is LDA,groups A and C is extra trees(ET),and groups B and C group is LDA.Conclusion The proposed classification model based on Raman spectroscopy and machine learning algorithms can distinguish subgingival plaques among patients with CP and T2DM,with CP alone,and healthy controls.This technique can be used in future clinical practice as a screening or diagnostic tool.