Pre-processing and Source Separation methods for Raman spectra analysis of biomedical samples

Raman spectroscopy is a useful tool to investigate the molecular composition of biological samples. Source separation methods can be used to efficiently separate dense informations recorded by Raman spectra. Distorting effects such as fluorescence background, peak misalignment or peak width heterogeneity break the linear and instantaneous generative model needed by the source separation methods. Preprocessing steps are required to compensate these deforming effects. We show in this paper how efficiency of source separation methods is deeply dependent on preprocessing steps. Resulting improvements are illustrated through the study of the numerical dewaxing of Raman signal of a human skin biopsy. The applied source separation methods are a classical ICA algorithm named JADE and two positive source separation methods called NMF and MLPSS.