Automatization of spike correction in Raman spectra of biological samples

• An automatic spike correction for large Raman spectral data sets was established.
• A quantitative marker for spike presence was established based on the response to a Laplacian.
• The identification of the threshold for spike detection was automatically determined.
• The algorithm allows for a stable detection of spikes with an estimated sensitivity of 99% for an example data set.

Raman spectroscopy as a technique has high potential for biological applications, e.g. cell and tissue analysis. In these applications, large data sets are normally recorded which require automated analysis. Unfortunately, a lot of disturbing external influences exist, which negatively affect the analysis of Raman spectra. A problematic corrupting effect in big data sets is cosmic ray noise, which usually produces intense spikes within the Raman spectra. In order to exploit Raman spectroscopy in real world applications, detection and removing of spikes should be stable, data-independent and performed without manual control. Herein, an automatic algorithm for cosmic ray noise correction is presented. The algorithm distinguishes spikes from spectra based on their response to a Laplacian, e.g. their sharpness. Manual rating of the spike presence was used as a benchmark for algorithm validation. The algorithm's sensitivity was estimated to be above 99%.