Chemometrics and statistical analysis in raman spectroscopy-based biological investigations

As mentioned in the chapter 1, chemometrics has become an essential tool in Raman spectroscopy-based biological investigations and significantly enhanced the sensitivity of Raman spectroscopy-based detection. However, there are some open issues on applying chemometrics in Raman spectroscopy-based biological investigations. An automatic proce- dure is needed to optimize the parameters of the mathematical baseline correction. Spectral reconstruction algorithm is required to recover a fluorescence-free Raman spectrum from the two Raman spectra measured with different excitation wavelengths for the shifted-excitation Raman difference spectroscopy (SERDS) technique. Guidelines are necessary for reliable model optimization and rigorous model evaluation to ensure high accuracy and robustness in Raman spectroscopy-based biological detection. Computational methods are required to enable a trained model to successfully predict new data that is significantly different from the training data due to inter-replicate variations. These tasks were tackled in this thesis. The related investigations were related to three main topics: baseline correction, statistical modeling, and model transfer.