Multivariate Measurements and Chemical Imaging of Organic Compounds
Chemical imaging is an emerging analytical technique, based on the interplay between, most often, spectroscopic measurements and advanced evaluation methodology. With this we can study the distribution of chemical compounds in different compartments, such as individual cells and pharmaceutical tablets. Also, chemical imaging allows for studies of time dependent processes, such as uptake of drugs or dissolution.
Surface enhanced Raman spectroscopy (SERS) was used to acquire data in complex matrices (lymphocytes), the spectra obtained varied significantly in background and intensity. Near-infrared spectroscopy (NIR) and atomic force microscopy (AFM) were used to characterise the texture of pharmaceutical tablets. Also, the dissolution of a tablet was monitored over time.
The focus of this thesis is the data analysis, more specifically the multivariate handling of large data sets, comparisons of images, structure identification and textures. Three application areas have been more thoroughly investigated; normalisation of spectral data to obtain quantitative results during difficult statistical sampling conditions; extraction of relevant spectral and chemical information from complex matrices and measurements in noisy or high-background environments; detection of small changes in spatial features.
A normalisation method for highly variable SERS measurements, based on an internal standard self assembled monolayer was developed and studied for use with multivariate regression. It was shown that the highly variable background in SERS imaging could be reduced with wavelet filtering and modelling of the background to extract chemically interesting spectra as outliers. Together with the concept of an internal standard applied as a self assembled monolayer, this points in the direction of a solution for quantitative SERS.
The use of wavelet transformation in combination with principal components analysis, were shown to be successful in characterisation of spatial features without prior segmentation. Continuous wavelet transform (CWT) gave more consistent results in texture analysis than discrete wavelet transform (DWT), and dual-tree complex wavelet transform (DT-CWT).