Research On Detection Of Hallucinofenic Drugs Based On SERS Technology Combined With Machine Learning Algorithm

In recent years,cases,where hallucinogenic drugs have been used for illegal and criminal purposes,have frequently occurred,seriously threatening people's safety.Therefore,a method for rapid detection and identification of hallucinogenic drugs is needed.The current detection methods for hallucinogenic drugs have the disadvantages of long detection time and high equipment costs.Surface-enhanced Raman spectroscopy(SERS)is widely used in the field of drug analysis and diagnosis due to its relatively portable equipment,short detection time,and relatively simple experimental operation.This paper is based on the SERS technique to detect flibanserin and tadalafil by using a highly sensitive silver sol substrate.This article completed the qualitative and quantitative analysis of the two drugs,and combined with machine learning algorithm to complete the rapid classification of flibanserin.This method is very suitable for rapid and accurate on-site detection of psychedelic drugs.First,based on the density functional theory,the molecular structure of the two drugs was simulated,and the theoretical Raman spectra of the two drugs were optimized.According to the theoretically calculated Raman spectrum compared with the actual sample Raman spectrum,the characteristic peaks and their vibration modes are assigned.Then prepare the silver sol substrate,use a high-resolution field emission scanning electron microscope and dynamic light scattering instrument to characterize the morphology,distribution,and particle size of the prepared silver sol nanoparticles.The results show that the silver sol substrate has a uniform particle size and completes its stability testing.And then,the prepared SERS substrate was used for SERS detection of flibanserin and tadalafil.The detection limit of flibanserin was 1 ?g/m L and the detection limit of tadalafil was 10 ?g/m L.For the quantitative analysis of the two drugs,the recovery rate of Flibanserin liquor solution ranged from 93.70% to 108.32%,and the recovery rate of tadalafil methanol solution ranged from 90.7% to 107.12%.Finally,both principal component analysis(PCA)combined with support vector machine(SVM)model and convolutional neural network model(CNN)were used to qualitatively and quantitatively classify flibanserin liquor,beer,and wine solutions.The algorithm reduces the judgment workload and the probability of artificial misjudgment.The PCA-SVM model's qualitative classification accuracy for Flibanserin liquor,beer,and wine spectral data sets is 100.00%,95.80%,and 92.00%,and the quantitative classification accuracy for these three data sets is 92.30%,91.70%,92.00%.The qualitative classification accuracy of the CNN model is 100.00%,94.70%,and 96.90%,and the quantitative classification accuracy is 95.00%,90.70%,and 91.10%,respectively,which verifies the advantages and feasibility of the algorithm instead of manual.The paper fully demonstrates the huge application potential of SERS technology in the rapid on-site detection of psychedelic drugs and the possibility of automatic quantitative detection of samples in the future.