Study Of Raman Spectroscopy Combined With Machine Learning In Diagnosis Of Systemic Lupus Erythematosus

Objective: The research is based on the Raman spectroscopy technology to detect the serum of Systemic lupus erythematosus(SLE)and Primary Sjogren’s syndrome(p SS)、Undifferentiated connective tissue disease(UCTD)、Healthy control(HC)groups to obtain the spectral characteristics of their different molecular substances,and combined with computer algorithms to establish the SLE classification model,so as to screen SLE patients at an early stage,aiming to provide a basis for developing a fast,accurate,and simple SLE diagnosis method.Methods: The serum of 105 patients with SLE,80 patients with p SS,31 patients with UCTD and 75 healthy control groups(HC)were detected by Raman spectroscopy.Through baseline correction,smoothing,outlier elimination and normalization operation pretreatment,the principal component analysis(PCA)was used for data dimensionality reduction,the training set and the test set were divided according to the 8:2 ratio,and the model performance was evaluated using the three-fold cross-validation method,Finally,support vector machine(SVM)and linear discriminant analysis(LDA)are established.Results:1.The RS of SLE and p SS,UCTD and HC groups were significantly different at 1326,1447,1605 and 1660 cm-1 displacement(P<0.05).2.The difference of RS relative intensity between SLE and p SS,UCTD and HC groups can reflect the difference of disease biochemical components.3.RS combines PCA-SVM and PCA-LDA algorithm to establish the classification model.The accuracy of PCA-SVM is 95.3%,the sensitivity is 83.3%,the specificity is 98.3%,and the AUC value is 0.90;The accuracy of PCA-LDA was 88.6%,the sensitivity was 60.0%,the specificity was 95.8%,and the AUC value was 0.77.Conclusion: After preliminary research,two classification models,PCA-SVM and PCA-LDA,have been established based on Raman spectroscopy technology and machine algorithms.SVM has a good classification effect.