Dynamic hyperspectral and polarized endoscopic imaging

The health of rich, developed nations has seen drastic improvement in the last two centuries. For it to continue improving at a similar rate new or improved diagnostic and treatment technologies are required, especially for those diseases such as cancer which are forecast to constitute the majority of disease burden in the future. Optical techniques such as microscopy have long played their part in the diagnostic process. However there are several new biophotonic modalities that aim to exploit various interactions between light and tissue to provide enhanced diagnostic information. Many of these show promise in a laboratory setting but few have progressed to a clinical setting. We have designed and constructed a flexible, multi-modal, multi-spectral laparoscopic imaging system that could be used to demonstrate several different techniques in a clinical setting. The core of this system is a dynamic hyperspectral illumination system based around a supercontinuum laser and Digital Micromirror Device that can provide specified excitation light in the visible and near infra-red ranges. This is a powerful tool for spectroscopic techniques as it is not limited to interrogating a fixed range of wavelengths and can switch between excitation bands instantaneously. The excitation spectra can be customised to match particular fluorophores or absorption features, introducing new possibilities for spectral imaging. A standard 10 mm diameter rigid endoscope was incorporated into the system to reduce cost and demonstrate compatibility with existing equipment. The polarization properties of two commercial endoscopes were characterised and found to be unsuited to current polarization imaging techniques as birefringent materials used in their construction introduce complex, spatially dependent transformations of the polarization state. Preliminary exemplar data from phantoms and ex vivo tissue was collected and the feasibility and accuracy of different analysis techniques demonstrated including multiple class classification algorithms. Finally, a novel visualisation method was implemented in order to display the complex hyperspectral data sets in a meaningful and intuitive way to the user.EThOS - Electronic Theses Online ServiceGBUnited Kingdo