Estimating and suppressing background in Raman spectra with an artificial neural network

In this report we address the problem of skin fluorescence in feature extraction from Raman spectra of skin lesions. We apply a highly automated neural network method for suppressing skin fluorescence from Raman spectrum of skin lesions before dimension reduction with principal components analysis. By applying the background suppression, the effect of outlier spectrum in the principal components analysis was greatly reduced. Introduction Studies have shown that Raman spectra of skin cancer has potential for skin cancer diagnosis [2, 4, 5, 6, 3]. Raman spectra are obtained with equipment consisting of a radiation source and a spectrometer. The laser beam excites the molecules and two scattering processes are observed, elastic and inelastic. The elastic process is the so-called Rayleigh scattering where the reflected wavelength is the same as the exited wavelength. In contrast, the inelastic scattering changes the reflected wavelength, giving a frequency shift in the reflected Raman spectra. This is called Raman scattering. The Raman spectra of two molecules are different if they have different structure, thus specific substances can be identified by their Raman spectra. The radiation source for the excitation used in this study is a neodymium doped yttrium-aluminium garnet laser emitting at 1064 nm. This frequency lies in the near infrared region which makes the spectra less vulnerable to sample fluorescence, which may severely corrupt the spectra. Nevertheless, the influence of sample fluorescence or S. Sigurdsson was supported by the Danish Research Councils through the project Signal and Image Processing for Telemedicine (SITE)