A Practical Algorithm to Remove Cosmic Spikes in Raman Imaging Data for Pharmaceutical Applications

Abstract

Raman dispersive microscopic imaging techniques are finding ever-increasing applications in pharmaceutical research for their ability to provide spatial and spectral information about the sample. Spectral data acquired from dispersive Raman instruments utilizing charge-coupled device detectors are characterized by occasional high intensity spikes arising from cosmic ray events. These random cosmic spikes are superimposed on chemically meaningful spectra. Due to their high intensity and potential influence on variance structures, it is often crucial to filter cosmic spikes from data prior to the use of multivariate algorithms to extract chemical information from the image cube. Some extremely challenging cosmic spikes are found to seriously interfere with multivariate data analysis for our application, e.g., spikes with bandwidth greater than the bandwidth from species of interest, spikes in neighboring image pixels occurring at the same spectral channels, spikes right on top of the band of interest, etc. A practical algorithm is proposed for semi-automated cosmic spike removal. The algorithm is computationally efficient, conceptually simple, and easy to implement. It is an alternative to methods using repetitive measurements by taking advantage of the spatial characteristic of imaging techniques and existing knowledge from the formulation. The algorithm has been shown to generate recovered spectra with negligible spectral distortion. The utility of the algorithm will be illustrated by the analysis of Raman images of pharmaceutical samples.