Methods for Improving Signal to Noise Ratio in Raman Spectra

Raman microspectroscopy is an optoelectronic technique based on the inelastic scattering of light. This technique has been demonstrated to have potential to identify different materials based on subtle differences in the Raman spectral profile using various multivariate statistical classification tools. However, Raman scattering is an inherently weak process. Low photon counts coupled with non-ideal collection efficiencies means that Raman spectroscopy is vulnerable to noise. This makes system optimisations, as well as efficient and reliable noise removal, a necessity in sensitive applications such as chemical classification or diagnostics. Provided in this thesis are software and experimental methodologies to evaluate system performance, predict system performance under various conditions, and to identify the optimal system configuration/set-up in order to achieve the highest possible signal to noise ratio. Modelling methodologies presented in this thesis allow the user to systematically evaluate minimum acquisition times, optimise camera read-out modes, and predict system behaviour with alternative optical elements in order to maximise signal to noise ratio. The denosing algorithms presented in this thesis have been shown to provide superior signal to noise ratio when compared with their traditional counterparts. When compared with the double acquisition method, the proposed cosmic ray removal algorithm resulted in a 10% improvement. An algorithm that enhances Savitzky-Golay smoothing with maximum likelihood estimation produced spectra with up to double the signal to noise ratio when compared to the raw spectra and consistently outperformed the algorithms it was compared to. The use of reflective substrates is also investigated and was shown to approximately triple the collected Raman scatter when compared with transparent substrates. By utilising the methodologies detailed in this thesis it is possible to improve the efficiency of the Raman system in question