Comparing Coherent and Spontaneous Raman Modalities for the Investigation of Gastrointestinal Cancers.
Curtis, Kelly Marie
Thesis or dissertation
University of Exeter
The incidence of gastrointestinal (GI) cancers has been steadily increasing in the UK since the mid 1970’s. These include cancers of the colon and oesophagus. Colon cancers have a high incidence rate, being the fourth most common cancer in the UK for both men and women. Oesophageal cancers in comparison are much rarer, however they have a poor survival rate primarily due to a late diagnosis. The key to improving survival for these cancers and many others is to detect and remove the disease at the early stages, to prevent the cancer from advancing. At present the ‘gold standard’ for diagnosis is a biopsy followed by histopathology. This technique is invasive, time consuming and highly subjective. It is therefore important to look towards non-invasive methods for early and rapid diagnosis. Optical techniques have begun to show such promise. By probing the interactions of tissues with light, diagnostic information is able to be obtained non-invasively. Techniques such as Raman spectroscopy utilise inherent molecular vibrations to extract biochemical information from tissues. Raman spectroscopy, however, is currently fundamentally limited by long acquisition times, due to the inherently weak signals produced. Using coherent Raman techniques such as coherent anti-Stokes Raman scattering (CARS) and stimulated Raman scattering (SRS), the molecular vibrations are coherently driven to provide an enhancement in signal. This thesis explored spectral signatures from snap frozen oesophageal sections in the fingerprint (450 cm-1 to 1850 cm-1) and high wavenumber (2800 cm-1 to 3050 cm-1) regions using spontaneous Raman and compared with spectra from hyperspectral SRS. The diagnostic potential for each technique was assessed for four major pathology groups, normal, Barrett’s oesophagus, dysplasia and adenocarcinoma. Samples were classified using a principal component fed linear discriminant analysis (PCA-LDA) approach with a leave-one-out cross validation. Comparisons were made to haematoxylin and eosin (H&E) stained sections. Raman in the fingerprint region was found to be the most promising for diagnosis. There were minimal changes in the high wavenumber region between pathology groups which was also reflected in the SRS spectra and proved to be insufficient for classification. Further comparisons were made between spontaneous and coherent Raman techniques using frozen colon sections. The morphological and structural information available was explored using a k-means cluster analysis. Both spontaneous and coherent Raman were able to distinguish important structural features in the colon, such as the epithelial cells that form the colonic glands and surrounding connective tissue. Both are important visual markers for cancer diagnosis in the current approach. SRS demonstrated higher spatial resolution and faster acquisition times in comparison to spontaneous Raman. This work has discussed the many advantages of using coherent Raman techniques for tissue applications, but has also highlighted some of the limitations for spectral measurements, arising from the complexity of the system.
PhD in Physics