Hyperspectral data influenced by sample matrix: the importance of building relevant reference spectral libraries to map materials of interest

Abstract

Hyperspectral imaging (HSI) and mapping are increasingly used for visualization and identification of nanoparticles (NPs) in a variety of matrices, including aqueous suspensions and biological samples. Reference spectral libraries (RSLs) contain hyperspectral data collected from materials of known composition and are used to detect the known materials in experimental samples through a one-to-one pixel “mapping” process. In some HSI studies, RSLs created from raw NPs were used to map NPs in experimental samples in a different matrix; for example, RSLs created from NPs in suspension to map NPs in biological tissue. Others have utilized RSLs created from NPs in the same matrix. However, few studies have systematically compared hyperspectral data as a function of the matrix in which the NPs are found and its impact on mapping results. The objective of this study is to compare RSLs created from metal oxide NPs in aqueous suspensions to RSLs created from the same NPs in rat tissues following in vivo inhalation exposure, and to investigate the differences in mapping that result from the use of each RSL. Results demonstrate that the spectral profiles of these NPs are matrix dependent: RSLs created from NPs in positive control tissues mapped to experimental tissues more appropriately than RSLs created from NPs in suspension. Aqueous suspension RSLs mapped 0-602 out of 500,424 pixels per tissue image while tissue RSLs mapped 689-18,435 pixels for the same images. This study underscores the need for appropriate positive controls for the creation of RSLs for mapping NPs in experimental samples.

from #AlexandrosSfakianakis via Alexandros G.Sfakianakis on Inoreader http://ift.tt/2kKekfD
via IFTTT