Modeling and correction of Raman and Rayleigh scatter in fluorescence landscapes

• Rayleigh and Raman scatter in fluorescence landscapes disturb signal modeling.
• Local modeling the scatter with smooth curves allow accurate signal modeling.
• No specific assumptions about the fluorescence landscapes need to be made.
• The procedure can be seen as a necessary preprocessing for multiway modeling.

Rayleigh and Raman scatter in fluorescence (excitation–emission) landscapes are a nuisance in two-way and three-way data modeling. We provide a method to clean individual emission spectra. The scatter can be represented accurately by Gaussian peaks, characterized by location, width and height. The analytic signal of interest effectively acts as a background to the scatter peaks. Modeling it locally as a smooth curve, using penalized least squares, allows accurate estimation of the parameters of scatter peaks. Once the peaks are modeled, they can be subtracted from the spectrum, almost completely removing the artifacts. Apart from local smoothness, no assumptions are made about the fluorescence spectra.