Tagged MR image analysis for estimation and statistical analysis of myocardial motion and strain

Cardiovascular diseases (CVDs) are the leading cause of death in the western world. Deriving and comparing intra-myocardial contractile properties accurately, meaningfully and throughout the cardiac cycle using tagged Magnetic Resonance Imaging (MRI), a promising imaging modality, has proven to be a challenging problem. This thesis pr esents algorithms for the estimation and statistical analysis of myocardial motion and strain from cardiac Tagged MR Images. Towards this, we develop a semi-automated algorithmic pipeline to perform (1) automated and sub-pixel accurate myocardial motion estimation using diffeomorphic non-rigid registration, (2) dense validation of estimated motion using novel synthetic data based ground truth generation, (3) quantitative comparison of left ventricular (LV) strains across subjects via accurate inter-subject registration, and, (4) exploratory analysis of LV strains using PCA. In addition, we also present a high dimensional inter-slice interpolation algorithm to deal with low-resolution image volumes during inter-subject registration.