The Influence of Preprocessing Methods on Multivariate Image Analysis in High-Performance Thin-Layer Chromatography Fingerprinting

Abstract

Planar chromatography is commonly used for the quality control of herbal medicines due to its many advantages. Its combination with chemometrics was proven to be a fast and reliable tool for the extraction of even more analytical information, such as similarity or dissimilarity between samples, and the identification of marker compounds. To date, depending on image processing procedures, different variables have been obtained as input data, and thus, various preprocessing procedures have been applied. In this study, we converted the chromatogram images of high-performance thin-layer chromatography to form a data matrix, by digitization of the chromatograms. Further, principal component analysis was applied on raw data and investigated after different preprocessing techniques. The proposed preprocessing techniques were successfully applied to improve the differentiation between two types of German propolis. The best multivariate models were observed in the case of warping, standard normal variate, and mean centering/autoscaling.