Abstract:

The inverse problem of electrocardiology might provide a powerful clinical investigation method for visualising the electrical activity of the heart. To use this method one requires accurate models of the human torso and heart. The objective of this work was to create an accurate model of the human ventricles including the valves from images recorded using Magnetic Resonance Imaging (MRI). This model is used as a "generic" model, and is adapted to a given individual with a host mesh fit to spatially registered Ultrasound (US) images.

Abstract:

A generic model of the human ventricles is created from Magnetic Resonance Images (MRI) and customised to a given patients heart using Ultrasound data with a host mesh fitting procedure. The motivation for this work is to provide individual and accurate models for the solution of the inverse problem in electro- cardiology.The first objective was the creation of a surface model of the human ven- tricles. This was constrained by several factors, due to the desired application for the inverse problem. A bi-cubic Hermite basis function with quadri-lateral elements is chosen, which is an efficient shape descriptor with first and second order continuity across element boundaries. The ventricles and the valve plane is modelled with three independent surface meshes. The whole model is closed to fulfil the boundary conditions of the inverse problem.The second objective was the adaption of this surface model to a given patients heart. MRI is the most accurate, but also most expensive imaging technology. 3D Ultrasound is cheap, but some areas are barely visible, e.g. the right ventricular free wall. Thus Ultrasound alone is insufficient to develop a complete model from. The host mesh fitting procedure combines the advantages of both imaging technologies. In areas with Ultrasound data the model gets adapted to the new positions and elsewhere the generic model is used to fill in the missing features.The generic model is first embedded as a slave mesh inside the host mesh by making the nodes relative to the ξ-coordinates of the host mesh. After a gross alignment, the Ultrasound data is projected onto the surfaces of the slave mesh yielding pairs of points for the subsequent fitting procedure of the host mesh. Finally, the slave mesh is updated to the new host mesh.An additional objective of this work was the creation of a volumetric model of the ventricular walls. This uses a tri-cubic Hermite interpolation function. The main constraint of this model in the creation is in retaining consistency in its ξ-directions, leading to a complicated mesh topology.