infrared; spectroscopy; imaging; chemometrics; skin; cancer
IR spectroscopy produces spectra in which detailed information concerning chemical structure is inherent. Numerous studies have demonstrated that the most useful IR methods for analysis of biological tissues are microscopic image-based techniques in which fine-scaled spatial and high-quality spectral information is integrated. Unlike traditional visible microscopic methods, the contrast in IR imaging is gained by differences in spectra and the spatial heterogeneity of biochemical components, not by the addition of stains. In order for IR imaging to be more broadly accepted, non-subjective data processing methods are being developed to extract the most out of the large spectral images that are acquired. This paper demonstrates data processing techniques that have been extremely useful in the analysis of normal and abnormal skin. Analysis of skin specimens is of particular clinical importance due to the difficulty in rendering a differential diagnosis. Unstained frozen skin sections were mapped using an IR microscope. Functional group mapping, clustering routines and linear discriminant analysis were used to process the data. Functional group mapping and clustering routines were useful in the initial interpretation of images and to research for trends in uncharacterized spectral images. LDA was useful for differentiating normal from abnormal tissue once a well- defined training spectral set was established. Representative spectroscopic images are shown that demonstrate the power of IR imaging.
Date de publication
Vibrational Spectroscopy-based Sensor Systems : 226–229.