Robust cosmic ray spike detection in single Raman spectra and time series

Cosmic rays (CRs) are a common spectral interference in Raman spectroscopy. While they are easily identified by human eye, there is an increasing need for better automated removal methods especially in the cases of large datasets or dynamic measurements. Yet, existing methods rely either on repeated non-dynamic measurements of identical samples, simple metrics, or wide spectral features, to easily identify CRs. In this article, we propose a broadly applicable strategy incorporating continuous wavelet transform in a multi-step, progressively more discriminative CR detection algorithm that can be applied on a single spectrum, without requiring a reference spectrum. When applying it to a dataset comprised of various types of Raman acquisitions, the algorithm surpassed the established methods. Of importance, this single-spectrum method can be applied to the preprocessing of dynamic surface enhanced Raman spectroscopic measurements.