27
IRUS TotalDownloads
Altmetric
Applications of advanced spectroscopic imaging to biological tissues
File | Description | Size | Format | |
---|---|---|---|---|
Song-CL-2020-PhD-Thesis.pdf | Thesis | 12.13 MB | Adobe PDF | View/Open |
Title: | Applications of advanced spectroscopic imaging to biological tissues |
Authors: | Song, Cai Li |
Item Type: | Thesis or dissertation |
Abstract: | The objectives of this research were to develop experimental approaches that can be applied to classify different stages of malignancy in routine formalin-fixed and paraffin-embedded tissues and to optimise the imaging approaches using novel implementations. It is hoped that the approach developed in this research may be applied for early cancer diagnostics in clinical settings in the future in order to increase cancer survival rates. Infrared spectroscopic imaging has recently shown to have great potential as a powerful method for the spatial visualization of biological tissues. This spectroscopic technique does not require sample labelling because its chemical specificity allows the differentiation of biocomponents to be achieved based on their chemical structures. Experiments were performed on 3-µm thick prostate and colon tissues that were deposited on 2 mm-calcium fluoride (CaF2) which were subsequently deparaffinised. The samples were measured under IR microscopes, in both transmission and attenuated total reflection (ATR) mode. In transmission, thermo-spectroscopic imaging of the prostate samples was first carried out to investigate the potential of thermography to complement the information obtained from IR spectral. Spectroscopic imaging has made the acquisition of chemical map of a sample possible within a short time span since this approach facilitates the simultaneous acquisition of thousands of spatially resolved infrared spectra. Spectral differences in the lipid region (3000 -2800 cm-1) were identified between cancer and benign regions within prostate tissues. The governing spectral band for classification was anti-symmetric stretching of CH2 (2921 cm-1) from PCA analysis. Nonetheless, the difference in tissue emissivity at room temperature was minimal, thus the contrast in the thermal image is low for intra-tissue classification. Besides, the thermal camera could only capture IR light between 3333-2000 cm-1. To record spectral data between 3900 - 900 cm-1 (mid-IR), Fourier transform infrared (FTIR) spectroscopic imaging was used to classify the different stages of colon disease. An automated processing framework was developed, that could achieve an overall classification accuracy of 92.7%. The processing steps included unsupervised k-means clustering of lipid bands, followed by Random Forest (RF) classification using the ‘fingerprint’ region of the data. The implementation of a correcting lens and the effect of the RMieS-EMSC correction on the tissue spectra were also investigated, which showed that computational RMieS-EMSC correction was more effective at removing spectral artefacts than the correcting lens. Furthermore, the effect of the fluctuations of surrounding humidity where the experiments were carried out was studied by using various supersaturated salt solutions. Significant peak changes of the phosphate band were observed, most notably the peak shift of the anti-symmetric stretching of phosphate bands from 1230 cm-1 to 1238 cm-1 was observed. By regulating and controlling humidity at its lowest, the classification accuracy of the colon specimens was improved without having to resort to alteration on the RF machine learning algorithm. In the ATR mode, additional apertures were introduced to the FTIR microscope, as a novel means of depth profiling the prostate tissue samples by changing the angle of incidence of IR light beam. Despite the successful attempts in capturing the qualitative information on the change of tissue morphology with the depth of penetration (dp), the spectral data were not suitable for further processing with machine learning as dp changes with wavelengths. Apart from the apertures, a ‘large-area’ germanium (Ge) crystal was introduced to enable simultaneous mapping and imaging of the colon tissue samples. Many advantages of this new implementation were observed, which included improvement in signal-to-noise ratio, uniform distribution, and no impression left on the sample. The research done in this thesis set a groundwork for clinical diagnosis and the novel implementations were transferable to studies of other samples. |
Content Version: | Open Access |
Issue Date: | Jun-2020 |
Date Awarded: | Jan-2021 |
URI: | http://hdl.handle.net/10044/1/101147 |
DOI: | https://doi.org/10.25560/101147 |
Copyright Statement: | Creative Commons Attribution NonCommercial Licence |
Supervisor: | Kazarian, Sergei |
Sponsor/Funder: | Imperial College London |
Department: | Chemical Engineering |
Publisher: | Imperial College London |
Qualification Level: | Doctoral |
Qualification Name: | Doctor of Philosophy (PhD) |
Appears in Collections: | Chemical Engineering PhD theses |
This item is licensed under a Creative Commons License