Tag line tracking and Cardiac Motion Modeling from Tagged MRI

Abstract

Magnetic resonance (MR) tagging magnetically labels specified regions of the myocardium, which appear in the MR images with a spatially encoded pattern of dark stripes called tag lines. The deformation of these tag lines reflects the deformation of the underlying tissue, making it possible to quantitatively evaluate the regional myocardium deformation and strain. This is particularly valuable in the diagnosis of ischemia and infarction.

In this dissertation, three new algorithms are presented to track the tag lines and reconstruct the cardiac left ventricle (LV) motion from tagged cardiac MR images. In the new statistical tag point classification algorithm, the candidate tag point positions are not pre-smoothed during tracking, allowing smoothness constraints to be applied only as the deformation model is fitted to the tag points. Two new algorithms based on three-dimensional (3-D) B-splines in prolate spheroidal coordinates are also presented in order to reconstruct the cardiac LV motion. One of these algorithms is used for 3-D LV motion reconstruction from tracked tag lines. The other combines tag tracking and motion reconstruction. These methods model the left ventricle with prolate spheroidal B-spline models, which offer several advantages. First, spatially localized B-spline basis functions offer better local control than globally-defined spherical harmonics. Second, their domain more closely matches the shape of the LV wall than Cartesian or cylindrical models. Third, the models can enforce smoothing across the apex and compute the strain at that location.

Human and animal MR studies and simulations were used to validate the effectiveness of the new methods. The experimental results verified the effectiveness of the new tag line tracking and 3-D LV motion reconstruction algorithms.