Parameter and Structure Indentification in Optical Tomography

Optical tomography is a non-invasive imaging technique for imaging the optical properties of biological tissue, particularly the peripheral muscle, breast and the brain. Optical tomography utilizes a set of optodes placed on the surface of the tissue to deliver an input signal. A second set of optodes at different locations detect exiting photons which have propagated through the biological tissue. The distribution of photons in these boundary measurements is used to reconstruct images of internal optical absorption and scattering coefficient distributions.The resulting images support a wide range of clinical applications. These include (i) non-invasive detection of breast tumours; (ii) functional imaging of muscle and brain activities; (iii) estimation of cerebral oxygenation and haemodynamics; (iv) measurements of cytochrome oxidase and mitochondrial energetics; (v) investigation of oxidative metabolism in muscle; (vi) measurements of tissue viability in transplantationof organs; and (vii) detection of abnormalities in joints of arthritic patients.Optical tomography is faster and cheaper than alternative imaging methods. The hardware is compact, allowing use in clinical settings where other imaging modalities are impractical. However, despite these advantages, optical tomography is not yet widely used. One of the major barriers to widespread acceptance is that the image reconstruction methods are slow and inaccurate. We consider three basic problems that are at the root of this block to progress :1. Accurate modelling methods for light propagation in tissue are too slow to be used repeatedly in solving the imaging problem.2. Optical measurements are noisy and limited in number which makes the imaging problem intrinsically inaccurate.3. Identification of clinically significant objects in the reconstructed images involves processing of noisy images even though the number or tyoe of object being sought is small.We will tackle these problems with three strategies :1. The use of model reduction techniques that allow the use of relatively inaccurate (but fast )models provided that the resultant errors are correctly handled2. The use of prior knowledge in a rigorous way using statistical techniques3. The direct reconstruction of clinical objects from the data, missing out the potentially unstable step of making the images.We will undertake a rigorous development and evaluation of these methods, including validation on experimental data. Developed software will be released on the internet.