Differentiation of normal skin and melanoma using high resolution hyperspectral imaging

Abstract

We investigated the use of high resolution hyperspectral imaging microscopy to detect abnormalities in skin tissue using hematoxylin-eosin stained preparations of normal and abnormal skin, benign nevi and melanomas. A goal of this study was to provide objective data that could be utilized by any researcher; and form the beginnings of a reference spectral data base. All spectral characterizations were acquired in percent transmission, and absorption, with contiguous wavelength acquisition between 400 and 800 nm; and a spectral resolution of ~1 nm. Biopsy sections were characterized with varying sample thickness, staining and magnification in order to determine their impact on spectral characterizations. Spectra were classified using spectral waveform cross correlation analysis, an algorithm that is linearity invariant. Classified spectra were incorporated into spectral libraries; and all spectra acquired from the field of view were correlated with library spectra to a quantified, user-determined, confidence threshold (minimum correlation coefficient). The results revealed that all skin conditions in our initial data sets could be objectively differentiated providing that staining and section thickness was controlled. We also demonstrated that it is likely that a reference spectral library database could be created to include bioinformatics and cluster analysis. This would assist multiple laboratories to participate in the input and retrieval of target spectral information.