Automatic registration of myocardial perfusion images

Myocardial perfusion is an indicator of heart health that can be used to locate damaged areas of the heart. It can be measured by an imaging modality such as magnetic resonance imaging (MRI). When measuring myocardial perfusion in a living human we must align images over time to correct for motion caused by breathing. In this thesis we will introduce a version of the normalized gradient fields similarity method, modifed to improve noise robustness and decrease computational complexity. Furthermore we will use a segmentation in the reference time-frame as a priori information. The results of this thesis is the development of a registration method that aligns the myocardium in perfusion MRI images. The myocardial alignment, measured by the Dice coefficient, increased from 0.79+-0.11 before registration to 0.89+-0.028 after registration. The resulting perfusion curves looks like we would expect from previous studies of myocardial perfusion