Human retinal oximetry using hyperspectral imaging

The aim of the work reported in this thesis was to investigate the possibility of measuring human retinal oxygen saturation using hyperspectral imaging. A direct non-invasive quantitative mapping of retinal oxygen saturation is enabled by hyperspectral imaging whereby the absorption spectra of oxygenated and deoxygenated haemoglobin are recorded and analysed. Implementation of spectral retinal imaging thus requires ophthalmic instrumentation capable of efficiently recording the requisite spectral data cube. For this purpose, a spectral retinal imager was developed for the first time by integrating a liquid crystal tuneable filter into the illumination system of a conventional fundus camera to enable the recording of narrow-band spectral images in time sequence from 400nm to 700nm. Postprocessing algorithms were developed to enable accurate exploitation of spectral retinal images and overcome the confounding problems associated with this technique due to the erratic eye motion and illumination variation. Several algorithms were developed to provide semi-quantitative and quantitative oxygen saturation measurements. Accurate quantitative measurements necessitated an optical model of light propagation into the retina that takes into account the absorption and scattering of light by red blood cells. To validate the oxygen saturation measurements and algorithms, a model eye was constructed and measurements were compared with gold-standard measurements obtained by a Co-Oximeter. The accuracy of the oxygen saturation measurements was (3.31%± 2.19) for oxygenated blood samples. Clinical trials from healthy and diseased subjects were analysed and oxygen saturation measurements were compared to establish a merit of certain retinal diseases. Oxygen saturation measurements were in agreement with clinician expectations in both veins (48%±9) and arteries (96%±5). We also present in this thesis the development of novel clinical instrument based on IRIS to perform retinal oximetry.EThOS - Electronic Theses Online ServiceAl-baath University, SyriaGBUnited Kingdo