An investigation of optimal performance criteria in Electrical Impedance Tomography

S Meeson

University of Southampton, Department of Physics and Astronomy, Faculty of Science.

Imaging of tissue electrical impedance has been a subject of research since the early 1980s. Rapid data collection is possible using modest computers and low cost equipment, that is simple to use and can be miniaturised. The imaging system poses no known hazards to the patient and is suitable for monitoring patient health. In this thesis three main issues are addressed: noise performance, electrode misplacement, and image reconstruction. All of these issues are important for long term monitoring applications of Electrical Impedance Tomography (EIT).

 A performance figure of merit for EIT is established based on a model of system noise. It is proposed that the figure of merit be used to characterise the noise performance of a data acquisition unit when attached to either a test object or a human subject. The figure of merit provides a criterion for judging the quality of the data and the sharpness of the image which could be achieved in specific clinical practice.

 In long term in vivo EIT monitoring changes in body position affect electrode positioning which introduces artefacts into the images. Also measurement electrodes are difficult to position accurately, particularly on patients in intensive care beds. Two procedures are described to reduce these electrode induced artefacts. The first reduces the sensitivity of the reconstruction to the electrode positions, and provides a rational criterion for regularising the reconstruction. The main consequences are that smoother images are produced, the number of artefacts and their magnitude are generally reduced, and the region of interest measurements are more reliable. The second procedure aims to compensate for electrode movements that occur during data measurement, by utilising the data measured in vivo to estimate modifications to the reconstruction process. Computer simulation tests have shown that this modification produces improved image fidelity.

 In most practical EIT systems the imaging algorithms use a reconstruction matrix generated from a uniform conductivity distribution. An assessment is made of the potential gains from incorporating a priori data in the reconstruction process. The results show that in the absence of precise knowledge of anatomy the assumption of uniformity is the least likely to introduce artefacts.

 A major part of the thesis develops a new non-linear treatment of image reconstruction, based on the Newton-Raphson iterative procedure. A constraint, based on promoting image smoothness, is used to guide the minimisation of the mean square difference between the experimental and the modelled data. By emphasising the character of the images achieved in the reconstructed solutions this work has not only achieved the best possible EIT images with available hardware, but also holds prospects for future hardware developments.

Graduation photographs

Last modified 11th November 2009

Stuart Meeson