Visualisation of the electrical activities of the heart in 3D

The work described in this Thesis is concerned with developing a technique to visualise the electrical activities of the heart in 3D from ECG signals taken from sensors on the surface of the body. Generating such visualisations can help cardiologists to identify abnormal electrical propagation flows non-invasively. The approach taken has been to develop models for the body and the heart, followed by the implementation of the 'forward' solution, which calculates the body surface potential for excitations within the heart. The results obtained match published results. The 'inverse' solution, which determines the heart electrophysiology from the body surface measurements, was then implemented. Values derived from the 'forward' solution are then used to confirm the accuracy of the 'inverse' solution. The study has led to three improvements to existing approaches: a more realistic model of the heart's conduction system; and more effective solutions to both the 'forward' and 'inverse' calculations used to determine heart electro physiology. The methods that were developed were based on the biological and physiological properties of the heart tissues as well as the working methodology of the Diffusion Tensor Magnetic Resonance Imaging (DT-MRI) scanner. Evaluation of the proposed techniques has been verified using seven methods, which include the location of Purkinje cells, the anatomy of the ventricle conduction system of the human heart, the Myocardium tissues to the Purkinje tissues ratio, the excitation propagation of the conduction system of the ventricles, the excitation isochrones of ventricles of the human heart, the generated body surface potential map and the generated 12 Lead ECG electrograms. A fundamentally new aspect of the work is to extract the conduction system of the ventricles from the DT-MRI providing a more realistic model for this structure, and this process has been accomplished by a semi-automatic manner, where extraction of the conduction system is accomplished with minimum manual processing and some simple image processing techniques. The Monodomain reaction diffusion equation, which is used to model the ventricle excitation propagation, has been updated to include the diffusion of the electrical stimulation in a non-uniform material, which is the more realistic case. The DT-MRI modality was employed for the first time to model the conduction system of the ventricles and to determine the relative non-uniform conductivity distribution inside the heart Myocardium. Unlike previous methods which consider an estimated conduction network for early activation points and assume the Myocardium material to be a uniform material, the new approach provide a more realistic solution for both the modelling of conduction system and the 'forward' solution. The 'inverse' solution is calculated for a localised multiple dipoles (sources) distribution inside the heart Myocardium based on transmembrance potential instead of current density, as currently used. This type of problem is highly illposed as the number of body surface readings is much fewer than the number of the heart dipoles (highly underdetermined). Employing the transmembrance potential reduces the size of the problem to a third of the current density solution and as a consequence it improves the localisation (due to the reduction of the underdeterminity) and reduces the memory usage and computational power to be 1/9 of the current density solution. Three equations have been derived to calculate the transfer matrix of the problem: the first one is for an isotropic source in a homogeneous-isotropic conductor; the second one is for an anisotropic source in a homogeneous-isotropic conductor; and the third one is for any type of source in an inhomogeneous-isotropic conductor. A low resolution version of the 'forward' model has been employed to simulate the 'forward' solution of the heart, and the body surface readings (200, 100, 64 and 32 electrodes) of that model were then used in the 'inverse' solution. Three regularisation techniques have been used to solve the Inverse Problem namely: the Minimum Norm (MN), the Weighted Minimum Norm (WMN) and the Exact Low Resolution Brain Electromagnetic Tomography technique (eLORETA). Results of these methods are compared to the original pacing points (1, 2, 3, 4, and 5 pacing points) which are used in the forward solution. It is concluded that the best results are obtained from the eLORETA method, and a large number of electrodes on the body surface and fewer number of sources leads to better results.EThOS - Electronic Theses Online ServiceGBUnited Kingdo