Research On Medical Data Classification And Regression Method Based On Raman Spectroscopy

Raman spectroscopy is an ultra-sensitive and specific detection technology for biological and chemical sensing.It has the characteristics of simple and fast,fingerprint and low detection limit.Raman spectroscopy is being widely concerned,and it is also one of the hot spots in biomedical research.In this study,human serum Raman spectrum signal is using as the research object,and combined with stoichiometric statistical methods to establish a screening model for thyroid dysfunction diseases and a regression prediction model of hepatitis B virus DNA(HBV DNA)value and hepatitis B virus.This research has mainly done the following two work:First,based on Raman spectroscopy combined with different classification algorithms for the first time in the screening of abnormal thyroid function,the Raman spectroscopy data collected from human serum was deducted from the fluorescent background to obtain the purified Raman spectroscopy signal,and used partial least square(PLS)method dimensionality reduction.Finally,a PLS combined with support vector machine(SVM)classification model is established.The accuracy of this classification model is 96.66%.The results show that serum Raman spectroscopy combined with SVM is feasible in the screening of thyroid dysfunction diseases.Then,according to the Raman spectrum has the concentration of the target substance and its corresponding characteristic Raman shift spectrum peak intensity is linear,we use Raman spectroscopy based on partial least square regression(PLS)method for HBV HBV Preliminary research on quantitative prediction of DNA value.The problem of spectroscopy adaptive iterative penalty least squares,multivariate scattering,smoothing filter and normalization chemometric methods on the processing of raw spectral data and regression model is systematically discussed.Combined with the PLS regression model on the basis of chemistry,it can accurately predict the content of HBV DNA value,which has guiding significance for later treatment.The research results in this paper show that PLS based on Raman spectroscopy signals combined with SVM model can effectively screen thyroid dysfunction diseases and provide a new method for clinical judgment of thyroid diseases.At the same time,it also shows that the method of establishing a regression model can effectively predict the content of HBV DNA,which is helpful for the quantitative analysis of HBV DNA in clinic.