Scott test evaluation by multivariate image analysis in cocaine samples

Abstract

The Scott test is a preliminary colorimetric method to analyze cocaine. A blue color result in the final step denotes a positive indication for cocaine; however, some pharmacological products may lead to false positives when concentration is higher than 1mg. In order to eliminate these false positives, digital images derived from the Scott test for pure cocaine, adulterants, diluents, and their binary and ternary mixtures were acquired with a commercial scanner, and the histogram of such images were assessed through Principal Component Analysis (PCA), Partial Least Squares Discriminant Analysis (PLS-DA) and Support Vector Machine Discriminant Analysis (SVM-DA). PCA scores indicated a clear separation between a true positive and a false positive group, with exception for the higher concentrations of the adulterants assessed and their mixtures. In addition, we applied two supervised techniques, PLS-DA and SVM-DA, to the histograms of the digital images derived from the Scott test aimed at categorizing samples into true positive and true negative classes. PLS-DA yielded highly satisfactory results as only two samples were not correctly classified; the SVM-DA correctly classified all samples, yielding sensitivity and specificity equal to 1.