Multivariate chemical maps from μ-FTIR and DESI-MS hyperspectral data

Many modern analytical equipments allow to record spectral information across a surface. Chemical maps of compounds of interest – spatially located within the sample area investigated – are usually re-constructed by selecting a single spectral feature regarded as a marker for the compound. Nonetheless, such a univariate approach considerably underutilises the complex information embodied in the spectra, which are usually composed by many variables characterised by peculiar inter-correlations. The present study shows how multivariate methods are suitable to account for the complete spectral – and spatial – information from the samples studied. An interactive exploratory approach, based on principal component analysis (PCA) is applied to hyperspectral data arising from two different analytical techniques and application fields: μ-FTIR mapping, for characterisation and localisation of painting compounds in paint cross-sections (1), and DESI-MS mapping of biopsied human tissues, for chemical characterisation and differentiation of tumour and normal tissues (2). After PCA, a brushing procedure was performed in order to understand the relationships between the PC space and the map space, connecting chemical and spatial information. In particular, the score plot allows a visual inspection of the pixel distribution in PC space. In the score plot, it is possible to visualise groupings that indicate similarities among pixels, on the basis of the information derived from the spectra, and which can be associated with the particular characteristics of the samples analysed. With the brushing procedure, pixels with similar chemical profiles can be manually selected from the score plot in order to identify correspondences between the groups of points in the PC score plot and particular regions of the map. Finally, a joint examination of the loading score plots allows chemical characterisation of each part of the map to be achieved