Biomedical and clinical applications of Raman spectroscopy and multivariate chemometric methods

Raman spectra of biological samples are rather complex, resulting from the superimposition of signals from lipids, proteins, nucleic acids, and carbohydrates. A chemometric approach is therefore mandatory to extract relevant information from a large amount of apparently indistinguishable spectra. Before statistical analysis, spectra appearance should be firstly harmonized by baseline correction, noise reduction, cosmic rays removal and normalization to remove any unwanted variations not due to specific discriminating features. These aspects are fundamental in the real world for the construction of models, based on Raman spectra, able to discriminate or classify among different groups in a reliable manner. This chapter focuses on the workflow and the various approaches adopted in biomedical and clinical applications of Raman spectroscopy. Instrumental acquisition parameters which are specific to biological samples analysis, spectra mathematical pretreatment and multivariate chemometrics techniques (discriminant analysis and class-modeling techniques, including deep-learning algorithms) will be discussed. A selected number of recent case studies (published after 2019) is also presented.