Masterarbeit von Sebastian Sindelar
Models of the human heart play an essential role in today's medical research. They can improve algorithms relevant for diagnosing heart diseases. The goal of this thesis is to develop a geometry, called heart model, that precisely represents the heart anatomy. Furthermore, it should be possible to use heart models adapted to concrete CTA-datasets as the basis for evaluating registration algorithms.
This thesis introduces a novel process for creating heart models, which combines spline and mesh-based models into a two step modeling process. The resulting heart geometry covers the blood volume of all chambers, valves, the left ventricle muscle and the beginning of the aorta. The modeling process assures that each vertex of the geometry has the same position in every dataset regarding the organic structure. This is beneficial for the training of registration algorithms and for adding information to the model. The modeling process was evaluated by a user test, which showed an average point-to-point distance of around 1.1mm between test subjects. Heart models have been adapted to eleven CTA-datasets and can be used to evaluate and improve registration algorithms.