Analysis of indole and indazole amides synthetic cannabinoids by differential Raman spectroscopy based on ANN

A simple, rapid, accurate, and non-destructive method was developed for the determination of cannabinoids, combining principal component analysis and multi-layer perceptron neural network to classify indole and indazole amide synthetic cannabinoids. Under the experimental conditions of this study, 25 experimental samples were successfully classified into two categories as the final classification, which guaranteed 96% correct rate. First, the samples were manually classified and divided into two categories according to the difference in peak position and peak intensity of the differential Raman characteristic peaks at 650-540 cm^-1 , etc. Fisher's discriminant method (FDA) and principal component analysis (PCA) were used to analyze the experimental data. Fisher's discriminant analysis was used to formulate two classification functions to discriminate the results of manual classification, and the overall accuracy rate of classification reached 88%. Principal component analysis was used to reduce the dimensionality of the data, which could reduce the influence of redundant data on the experimental results. The original data, FDA-processed data and PCA-processed data, and artificial neural network algorithm (ANN-MLP/RBF) were combined to build a classification model. In the MLP model, the classification accuracy of the original data, FDA-processed data, and PCA-processed data was 80%, 92%, and 96% respectively, and the overall accuracy of the sample classification was 89.33%. In the RBF model, the accuracy of sample classification was 76%, 84%, and 92% respectively, and the overall accuracy of sample classification was 84%. Differential Raman spectroscopy could be used to distinguish 25 kinds of synthetic cannabinoids, and finally, the samples were divided into two categories. The PCA + MLP model was the best for processing spectral data. Based on the perspective of multivariate data, this study demonstrated that the method could be used for rapid and non-destructive testing of indole and indazole amide synthetic cannabinoids and that an efficient and non-destructive classification model was obtained. This method could be used for rapid detection and inspection of drugs in the field of forensic science.