|M.Sc Student||Tzvi Goldman|
|Subject||MRI/Ultrasound Image Registration for Analysis of Cardiac|
|Department||Department of Biomedical Engineering||Supervisor||Professor Emeritus Adam Dan|
|Full Thesis text|
In certain clinical settings it is desired to fuse information from different types of imaging modalities. An example of such a setting is the selection of patients suffering from ischemic cardiomyopathy for Implantable Cardiac Defibrillator (ICD) implantation. Studies have shown that tissue heterogeneity at the interface between infarct tissue and preserved tissue is a predictor for ventricular tachycardia (VT). Heterogenic regions can be identified by substrate analysis or by assessment of mechanical function. While contrast-enhanced magnetic resonance imaging (MRI) is an effective modality for substrate analysis, it is not the modality of choice for the assessment of mechanical function. Conversely, echocardiography-based regional strain assessment performs well at quantifying mechanical function but suffers from a low signal-to-noise ratio (SNR). Hence, it is desired to fuse MRI-derived morphological detail with the functional information provided by the echo-based strain assessment. A necessary first step for such a fusion is the registration of the complementary image sequences. The objective of this work was to assess the feasibility of registering 2D echocardiographic images with 2D cardiographic cine-MRI using pre-segmented endocardial borders for alignment. Nearest-neighbor interpolation was used to achieve temporal registration, while a combination of Iterative Closest Point, Generalized Multi-Dimensional Scaling, and the Thin-Plate Spline transformation was used to achieve spatial registration. Multiple sanity checks were enforced to ensure that nonsensical results were rejected. Due to limitations in the image acquisitions and poor data quality, results show limited success. However, some changes to the image acquisition protocol are proposed, and it is believed that if these changes are implemented, the success rate of this method could be greatly improved.