Hyperspectral data processing for chemoselective multiplex coherent anti-Stokes Raman scattering microscopy of unknown samples

Multiplex coherent anti-Stokes Raman scattering (MCARS) provides labeling free and fast characterization of materials and biological samples in nonlinear microscopy. In spite of its success, remaining challenges regarding the data analysis for chemoselective imaging still have to be solved. In general, image contrast has been realized by using only one spectral feature directly taken from the unprocessed raw data. This procedure is limited to strong and well separated Raman resonances like the saturated CH-stretching vibration of lipids in the case of biological samples. In order to overcome this limitation, we present a new method of MCARS data processing that exploits the whole measured spectrum to disentangle overlapping contributions of different (bio-) chemical components. Our "two-step" approach is based on the combination of imaginary part extraction followed by global fitting of the hyperspectral data set. Previous knowledge about the sample, e.g., pure spectra of the individual components is no longer necessary. The result is a highly contrasted image, where the patterns and differences between the sample components can be represented in different colors. We successfully applied this method to complex structured polymer samples and biological tissues.