Abstract:

Objective To investigate the role of Raman spectroscopy in gestational diabetes mellitus（GDM）. Methods A total of 29 non-pregnant healthy subjects,29 pregnant healthy subjects and 42 GDM patients were enrolled. Meanwhile,78 GDM patients and 70 healthy subjects were enrolled for verification. Orthogonal partial least square regression（OPLS）,principal component analysis（PCA） based method support vector machine（SVM） and latent dirichlet allocation（LDA） were used to classify Raman spectra obtained. Receiver operating characteristic（ROC）curve was used to evaluate the classification ability of the 3 models,and the optimal model was selected. The new method was compared with oral glucose tolerance test（OGTT）. Results Raman peaks appeared at 1 005.86,1 034.47,1 087.64 and 1 128.39 cm^-1 in the GDM group,non-pregnant healthy group and pregnant healthy group,respectively. The Roman peaks in the GDM group at 1 005.86,1 034.47,1 128.39 and 1 343.69 cm^-1 were significantly different from that in non-pregnant healthy group and pregnant healthy group. OPLS model had a good ability to distinguish GDM,non-pregnant healthy and pregnant healthy groups. The sensitivity and accuracy of OPLS model to distinguish GDM by Raman spectroscopy were 97.43% and 95.27%,respectively,which were higher than OGTT（91.03% and 93.92%）. However,the specificity of OPLS was lower than that of OGTT（92.86% vs 97.14%）. Conclusions Raman spectroscopy combined with OPLS model has high sensitivity,specificity and accuracy,It may be used for rapid screening of GDM.

Key words: Raman spectroscopy, Gestational diabetes mellitus, Multivariate analysis, Rapid screening