Flipped detection of psychoactive substances in complex mixtures using handheld Raman spectroscopy coupled to chemometrics

Abstract

New psychoactive substance (NPS) misuse represents a critical social and health problem. Herein, a novel flipped approach is presented for the detection of psychoactive substances in complex mixtures using portable Raman spectroscopy. This consists firstly of evaluating the spectral dissimilarities of an NPS product to its constituent adulterants followed by detection of the NPS by means of key spectral signatures. To demonstrate it, three structurally diverse NPS and four commonly used adulterants were selected. A Design-of-Experiments guided approach was employed to determine the composition of simulate street samples, ranging from binary to quinary mixtures of varying concentrations. Spectra were acquired for all mixtures using a portable Raman

A rapid and nondestructive approach for forensic identification of novel psychoactive substances using shifted‐excitation Raman difference spectroscopy and machine learning

“…[44,45] Swersky et al used a GP-based Bayesian algorithm to derive kernel functions for searching network structures in 2013, [46] and Hector Mendoza et al proposed the first neural network model for automated tuning of parameters in 2016. [47] In this paper, three optimization methods, BOA (Bayesian optimization), grid search, and random search, are used to optimize the classifier, and their specific analytical results are shown in Figures 10,11,and 12, and it can be found that each classification model is improved to different degrees after hyperparameter optimization. The accuracy of SVM, KNN, ensemble classifier, and neural network under the optimization of the three algorithms is greater than 90% in order to avoid model overfitting for 10-fold cross-validation of the models, and the improvement effect is significant.…”

“…[ 11 ] Among the above methods, LC‐MS as well as MS are highly accurate, but they need to be analyzed in the laboratory and have a long analysis time, which cannot meet the needs of field testing such as airport customs. Jesus Calvo‐Castro's team [ 12 ] examined 21 NPS samples and found that the traditional methods are subject to strong fluorescence interference and face a more serious challenge in NPS testing. Jone Omar [ 13 ] examined 30 samples using handheld Raman spectroscopy and used principal component analysis (PCA) to reduce the dimensionality and classify the samples.…”