Modelling scattering and absorption in the vibrational spectroscopy of cells and tissues

Abstract

Infrared spectroscopy has in the past decades grown to become one of the most widely used analytical techniques for biochemical characterization of biological materials. Since the technique is non-destructive, it allows for investigating cells and tissue in its native form. The infrared absorbance spectrum quantifies the wavenumber dependent absorption properties of the samples, and is therefore considered a molecular fingerprint. However, in infrared microscopy, scattering effects which disrupt the absorbance spectra have been a recurring problem. The scattering is especially strong for micrometer-sized samples, such a biological cells. The scattering from spherical biological samples has been identified as Mie scattering, and leads to significant baseline distortions in the infrared spectra. For thin film samples, scattering effects may also be prominent in the absorbance spectra. Multiple internal reflection inside the film gives rise to sine wave shaped fringes in the spectrum. During the past two decades, pioneering work has been done to develop methods for correcting scatter effects in infrared spectroscopy. Models which build on extended multiplicative signal correction (EMSC) is considered state-of-the-art for modelling and correcting a wide range of scatter effects, such as Mie scattering and interference fringes, as well as chemical variability. Central to these models is the reference spectrum, which the measured spectrum is modelled around. However, the role of the reference spectrum and other chemical constituent spectra is not very well described in the literature. Further, the EMSC based models for handling Mie scattering and interference fringes suffers both from lack of transferability to new datasets and are generally not very user-friendly. This makes the models less applicable to large infrared datasets, such as infrared hyperspectral images. In order to develop stable algorithms for preprocessing infrared spectra, the scattering phenomena must be understood in depth. For estimating and removing Mie scattering from infrared absorbance spectra, an isolated sphere is used as model system. However, biological cells are often not expected to be perfectly spherical. They also frequently appear in clusters of more than one cell, which raise the question whether an isolated sphere is an appropriate model system, or if coupling effects arise which need to be taken into account. Another coupling effect which is highly relevant is the coupling between the sample and the infrared microscope slide. For developing appropriate models, a better understanding of the scattering systems which are frequently encountered in infrared microscopy is needed. This thesis considers scattering and absorption in infrared microscopy, with the intention to develop stable and user-friendly preprocessing techniques for these data. The thesis is going in depth of scattering phenomena, especially Mie scattering and multiple internal reflection, and prepares a foundation for understanding and handling these phenomena. The thesis also seeks to develop robust methods that can be used on large datasets of spectra such as hyperspectral infrared images and to make algorithms available in user friendly and open-source platforms for the research community.