Study of Raman spectroscopy for the early detection and classification of malignancy in oesophageal tissue

Abstract

Raman Spectroscopy for the identification and classification of malignancy in the oesophagus has been demonstrated in this thesis. The potential of Raman spectroscopy in this field is twofold; as a adjunct for the pathologist and as a biopsy targeting tool at endoscopy. This study has demonstrated the feasibility of these potential applications in vitro. Spectral diagnostic models have been developed by correlating spectral information with histopathology. This is the current 'gold standard' diagnostic method for the identification of dysplasia, the established risk factor for the development of oesophageal cancer. Histopathology is a subjective assessment and widely acknowledged to have limitations. A more rigorous gold standard was therefore developed, as part of this study, using the consensus opinion of three independent expert pathologists to train the diagnostic models. Raman spectra have been measured from oesophageal tissue covering the full spectrum of malignant disease in the oesophagus, using a near infrared Raman spectrometer customised for tissue spectral measurements. Two spectral datasets were measured with different volumes of tissue probed using twenty and eighty times magnification ultra long working distance objectives. Multivariate statistical analysis has been used to extract the required spectral information with the greatest discriminative power. Principal component fed linear discriminant spectral models have been tested with leave one out cross validation procedures. Three pathology group models have correctly classified up to 91% of spectra, and eight group models have correctly classified up to 82% of spectra. Optimisation of the spectral models by selection of significant principal components, filtering the data and using staggered models was investigated. Effort has been made to understand the findings in their clinical context, with review of patient history and clinical progress, long term follow up is required. Preliminary work projecting independent data on to the models has been encouraging with 76% of the spectra in the three group model correctly classified, approaching classification levels of the training dataset. Formalin fixed tissue models were demonstrated to perform well, with 80% of the spectra were correctly classified in the seven group model. This further demonstrates the potential of Raman spectroscopy as a pathology tool. If Raman spectroscopy is to be implemented in a clinical setting it must be transferable between different measurement systems. This has been evaluated with oesophageal tissue spectra measured on two systems using three objectives. Simple calibration has demonstrated the use of multiple systems and measurement parameters in the development and application of spectral classification models. Testing of a new design of fibre probe has provided encouraging preliminary results. There is potential for the application of Raman spectroscopy in vivo, however the technology remains immature. Spectral maps of samples taken from across the spectrum of disease have shown clear delineation of the morphological features seen on the H&E images. Furthermore the biochemical information elicited has been analysed. Initial measurements of oesophageal tissue using multiphoton imaging have demonstrated the potential of collagen autofuorescence in the diagnosis of malignant change.