Raman spectroscopy has been demonstrated to have diagnostic potential in areas such as urine and cervical cytology, whereby different disease groups can be classified based on subtle differences in the cell or tissue spectra using various multi-variate statistical classification tools. However, Raman scattering is an inherently weak process, which often results in low signal to noise ratios, thus limiting the method's diagnostic capabilities under certain conditions. A common approach for reducing the experimental noise is Savitzky–Golay smoothing. While this method is effective in reducing the noise signal, it has the undesirable effect of smoothing the underlying Raman features, compromising their discriminative utility. Maximum likelihood estimation is a method for estimating the parameters of a statistical model given an available dataset and a priori knowledge of the model type. In this paper, we demonstrate how Savitzky–Golay smoothing may be enhanced with maximum likelihood estimation in order to prevent significant deviation from the ‘true’ Raman signal yet retain the robust smoothing properties of the Savitzky–Golay filter. The algorithm presented here is demonstrated to have a lower impact on Raman spectral features at known spectral peaks while providing superior denoising capabilities, when compared with established smoothing algorithms; artificially noised databases and experimental data are used to evaluate and compare the performance of the algorithms in terms of the signal to noise ratio. The proposed method is demonstrated to typically provide at least a 50% increase in the signal to noise ratio when compared to the raw data, and consistently out-performs two alternative smoothing filters.

Raman spectroscopy is a powerful diagnostic tool in biomedical science, whereby different disease groups can be classified based on subtle differences in the cell or tissue spectra. A key component in the classification of Raman spectra is the application of multi-variate statistical models. However, Raman scattering is a weak process, resulting in a trade-off between acquisition times and signal-to-noise ratios, which has limited its more widespread adoption as a clinical tool. Typically denoising is applied to the Raman spectrum from a biological sample to improve the signal-to-noise ratio before application of statistical modeling. A popular method for performing this is Savitsky–Golay filtering. Such an algorithm is difficult to tailor so that it can strike a balance between denoising and excessive smoothing of spectral peaks, the characteristics of which are critically important for classification purposes. In this paper, we demonstrate how Convolutional Neural Networks may be enhanced with a non-standard loss function in order to improve the overall signal-to-noise ratio of spectra while limiting corruption of the spectral peaks. Simulated Raman spectra and experimental data are used to train and evaluate the performance of the algorithm in terms of the signal to noise ratio and peak fidelity. The proposed method is demonstrated to effectively smooth noise while preserving spectral features in low intensity spectra which is advantageous when compared with Savitzky–Golay filtering. For low intensity spectra the proposed algorithm was shown to improve the signal to noise ratios by up to 100% in terms of both local and overall signal to noise ratios, indicating that this method would be most suitable for low light or high throughput applications.

Cosmic ray artifacts may be present in all photo-electric readout systems. In spectroscopy, they present as random unidirectional sharp spikes that distort spectra and may have an affect on post-processing, possibly affecting the results of multivariate statistical classification. A number of methods have previously been proposed to remove cosmic ray artifacts from spectra but the goal of removing the artifacts while making no other change to the underlying spectrum is challenging. One of the most successful and commonly applied methods for the removal of comic ray artifacts involves the capture of two sequential spectra that are compared in order to identify spikes. The disadvantage of this approach is that at least two recordings are necessary, which may be problematic for dynamically changing spectra, and which can reduce the signal-to-noise (S/N) ratio when compared with a single recording of equivalent duration due to the inclusion of two instances of read noise. In this paper, a cosmic ray artefact removal algorithm is proposed that works in a similar way to the double acquisition method but requires only a single capture, so long as a data set of similar spectra is available. The method employs normalized covariance in order to identify a similar spectrum in the data set, from which a direct comparison reveals the presence of cosmic ray artifacts, which are then replaced with the corresponding values from the matching spectrum. The advantage of the proposed method over the double acquisition method is investigated in the context of the S/N ratio and is applied to various data sets of Raman spectra recorded from biological cells.

Thin-film gold substrates offer improved performance and cost for NIR excitation Raman spectroscopy of biological cells when compared with CaF₂. We demonstrate a 1.65 times enhancement in the signal to noise ratio with <5% of the cost.