Study On Screening Method Of Hyperspectral And Raman Spectroscopy For Prohibited Additives In Antirheumatic Health-care Products

The scale of antirheumatic health-care products market in China continues to expand,and the occurrence of illegal addition of drugs in antirheumatic health-care products is common.Long-term using those health-care products will damage the body of consumers and affect their health.At present,the main detection methods for prohibited additives in health-care products are liquid chromatography and mass spectrometry.Most of these methods require complex pretreatment and time-consuming.Spectral detection method can realize the purpose of nondestructive and rapid identification of prohibited additives through the establishment of spectral library and the establishment of model combined with algorithm and so on.Therefore,in this paper,the near-infrared hyperspectral technology and thin layer chromatography-Surface-Enhanced Raman Spectrometry(TLC-SERS)method were used to establish a quantitative and qualitative screening method for the possible existence of prohibited added drugs in anti-rheumatic health-care products.The main contents of research are as follows:A quantitative analysis method for diclofenac sodium,which may be prohibited added in antirheumatic health-care products,was developed based on near infrared hyperspectral imaging.First of all,a series of antirheumatic health-care products containing diclofenac were prepared.Scanning at the wavelength of 1000-2500 nm,the average near infrared spectrum of the Region of Interest(ROI)was analyzed,extracted and calculated by using ENVI software.The effects of eight spectral pretreatment methods and three stoichiometric models on the accuracy of the predicted values were compared and analyzed.The results show that the multivariate linear regression model established by the pretreatment method of standard normal variables using the?-coefficient method to select the optimal band as the independent variable has good accuracy and predictive ability.The model predicted the minimum limit of diclofenac sodium was 0.05%,the coefficient of determination(R²)between the predicted value and the measured value was 0.993,and the root mean square error of prediction was 0.005.Based on TLC-SERS,a rapid screening and discrimination model was established for diclofenac sodium and 11 illegal chemical drugs with the same efficacy that might be added in antirheumatic health-care products and Chinese patent medicine.Firstly,the theoretical Raman peaks of 11 chemical drugs were calculated by using Guassian software and the peak attribution was identified.Then,comparing different development agents and TLC plates,it was found that petroleum ether-trichloromethane-ethyl acetate-glacial acetic acid(15:15:15:1.5,V/V/V/V)could effectively separate 11 chemical drugs.The volume of the gold colloidal solution and the integral time of Raman spectroscopy detection were optimized.By comparison,the conditions of Raman detection were determined by adding 4?L of gold colloidal solution and selecting the integration time of 5 s.The results show that the minimum detection limit of each drugs was0.05%-0.10%.Then,7 kinds of spectral pretreatment methods were used to process SERS spectra,and principal component analysis was carried out on SERS spectra,and the first 10characteristic principal component numbers were extracted,which would be used to establish and compare four discriminant models,such as Principal Component-Linear Discrimination,Principal Component-K Nearest Neighbor and Principal Component-Support Vector Machine.The results show that the prediction accuracy of the PCA-Linear Discriminant model is up to100%after the Gapsegment second derivative pretreatment.Two groups of drugs with similar Rf values,namely acetylsalicylic acid and indomethacin,diclofenac sodium and naproxen,were found when 11 mixed chemical drugs were separated by TLC at the same time.The chapter trys to establish models to further distinguish.The Raman spectra of acetylsalicylic acid and indomethacin and their mixture were compared with the Raman spectra of 200-2000 cm^-1 and the wave number selected by Competitive Adaptive Reweighting Algorithm(CARS)as independent variables.The results show that the Random Forest discriminant model was established by the pretreatment of CARS wavenumber with SG-smoothing and Standard Normal Variables(SNV),and the correct recognition rate was 67.17%.Based on the Raman spectra of diclofenac sodium,naproxen and their mixture,the kernel function support vector machine model was established after the Gapsegemnt second derivation combined with SNV pretreatment of the wavenumber screened by CARS.The recognition accuracy was 72.90%,but the pure samples could be identified 100%correctly.