Hyperspectral Raman images of human prostatic cells have been collected and analysed with several approaches to reveal differences among normal and tumor cell lines. The objective of the study was to test the potential of different chemometric methods in providing diagnostic responses. We focused our analysis on the ν(CH) region (2800-3100cm-1) owing to its optimal Signal-to-Noise ratio and because the main differences between the spectra of the two cell lines were observed in this frequency range. Multivariate analysis identified two principal components, which were positively recognized as due to the protein and the lipid fractions, respectively. The tumor cells exhibited a modified distribution of the cytoplasmatic lipid fraction (mainly localized alongside the cell boundary) which may result very useful for a preliminary screening. Principal Component analysis was found to provide high contrast and to be well suited for image-processing purposes. Self-Modelling Curve Resolution made available meaningful spectra and relative-concentration values; it revealed a 97% increase of the lipid fraction in the tumor cell with respect to the control. Finally, a univariate approach confirmed significant and reproducible differences between normal and tumor cells.
Keywords: Chemometrics; Prostate cancer; Raman imaging; Raman spectroscopy; Single cell spectroscopy.
Copyright © 2016 Elsevier B.V. All rights reserved.