Study On Measurement Methods Of Chinese Traditional Medicine Based On Terahertz Time-domain Spectroscopy

In recent years,traditional Chinese medicine(TCM)is being welcomed by more and more people,and it has also been paid much attention by the government.In 2016,the development of TCM was raised to one of the national strategy of China.A series of policies to standardize and revitalize TCM have been introduced,and have been guaranteed by the legal level.With the vigorous development of TCM,its quality control has become a difficult problem,and how to provide a fast and effective detection method has become a hot spot of research.Terahertz time domain spectroscopy(THz-TDS)is a new detection method.Because of its strong penetration,high resolution,convenient,efficient,safety and pollutionfree,this method is widely applied in the field of material detection,and also provides a feasible way for the detection of TCM.At present,the application of THz-TDS is still at the initial stage,and most of the related studies are focused on the detection of single component substances.There are many problems to be solved for the complex detection object like TCM.In view of the different needs in the detection,this paper analyzed the source of THz spectrum of TCM by molecular simulation,studied the classification and quantitative regression analysis model of TCM based on THz-TDS,and made a preliminary attempt to enhance the detection sensitivity.In the course of the study,the related knowledge of spectral analysis,quantum chemistry and machine learning was used.The main research contents summed up as follows:(1)Artemisinin was taken as an example for the study of the detection and identification by THz-TDS for the single ingredients of TCM.The unique absorption peak characteristic of artemisinin in the 0.2-1.6THz band absorption spectrum can distinguish it from other medicines,such like pyridoxin,benzoic acid,cinnamic acid and so on which are detected in this experiment.In order to provide explanations for the mechanism of THz characteristic absorption peak and provide evidence for the experimental results,the simulation was carried out by density functional theory(DFT).(2)Based on the THz absorption spectra of six kinds of TCM(gelsemium elegans,flos lonicerae,radix fici hirtae,herba solani,herba aristolochiae mollissimae and solanum nigrum),study of the classification and identification methods of TCM with complex components was carried out.By comparing the most widely used machine learning methods(support vector machine,back-propagation neural network and random forest),random forest has the advantages of the fastest operation speed and the insensitivity to parameter setting.However,there are often uneven class distributions(imbalance problems)in obtained datasets in real applications,resulting in significantly decreased classification accuracies of random forest.In order to solve this problem,CDKM-SMOTE random forest is obtained by introducing the synthetic minority over sampling technique(SMOTE)and the density peak clustering algorithm.CDKM-SMOTE random forest obtained better results than other models in accuracy and speed,and provides a useful tool for the classification and recognition of TCM.(3)Based on the THz absorption spectra of notoginseng and starch mixture,typha pollen and amine O mixture,and Chinese patent medicine Zhixue Ding Tong Pian,regression analysis was applied to TCM.Considering that the calculation results of the common regression model(partial least squares(PLS),interval PLS(i PLS)and backward interval PLS(bi PLS)and synergy Interval PLS(Si PLS))are not satisfactory in the experiment,a new regression method named 2DCOS-PLS was obtained by introducing a two-dimensional correlation spectroscopy(2DCOS)in PLS.2DCOS-PLS obtained more accurate prediction results and provided an ideal model for quantitative regression analysis of TCM.(4)A metamaterial absorber based on ohmic ring array was applied to enhance the detection sensitivity of THz-TDS to TCM,because of a unique property: the position of the resonant peaks in its THz transmission spectrum vary with the surface dielectric.The measured transmission spectra of the metamaterial absorber showed two resonant peaks separately at the 0.750 THz and 1.138 THz,which consistent well with the simulation.By observing the changes in the resonant peaks,different types of borneol(two natural borneol and one synthetic borneol)can be clearly classified.The change of trace content of chlorphenamine maleate,which can't be measured directly by THz-TDS,can also be reflected in the red shift of the resonant peak of the absorber.It could be concluded that metamaterial absorber can greatly improve the sensitivity of THz-TDS for the detection of TCM.