A nearest neighbor finite element method (NNFEM) for validating leftventricular regional strain from displacement encoding with stimulated echoes (DENSE) MRI, compared to tagged MRI

摘要

Cardiovascular magnetic resonance (CMR) is a magnetic resonance imaging (MRI) technique that is considered the most viable noninvasive technology for quantifying and visualizing regional myocardial function. CMR is expensive but characterized by higher spatial resolution and functional observations. The attribute of high spatial resolution allows quantitative assessment of cardiac wall motion and computation of transmural strains, allowing phenotyping cardiovascular physiopathologies [1-9]. Currently two CMR techniques are accepted as standard research practice which are 1. MRI tissue tagging (TMRI) [1-3] and 2. Stimulated echoes [6-9]. The first of the two, TMRI, is a method for tracking myocardial motion which places noninvasive markers (tags) within the tissue by locally induced perturbations of the magnetization. The altered magnetization shows as dark lines in the tagged region in successive images and myocardial deformation during the cardiac cycle is tracked [2,3]. However the intrinsic problem with tag lines is their fading after several cardiac phases. Hence, in addition to improvements toward longer tag persistence, parallel advancements in non-TMRI quantitative gradient technologies have also been made. One such technique is displacement encoding with stimulated echoes (DENSE) which directly encodes displacements into MRI phase data in three orthogonal phase encoding directions and facilitates rapid quantification of myocardial displacement through the cardiac cycle [6-9]. It is noted that while DENSE uses high displacement encoding frequencies resulting in phase wrapping, accurate measurements of displacements can be obtained using quality-guided spatio-temporal phase unwrapping algorithms [9]. The primary goal of this study was to develop an automated procedure for tracking the motion of each pixel within a trajectory by unwrapping DENSE phase data and estimating two-dimensional (2D) Lagrange strain (radial and circumferential) for soft tissue deformation [7-9]. The second goal was to validate DENSE strains against the gold standard of TMRI strains using a pre-existing phantom validation device and also human subject data [10]. This second goal required establishing a common strain computation protocol for both modalities. The common strain analysis framework in this study is a nearest-neighbor finite element method (NNFEM) which has advantages in not requiring element connectivity and background mesh generation [11-13]. The specific meshfree NNFEM technique used was a radial point interpolation (RPIM) finite technique. A third goal was to investigate the repeatability of 2D strains calculated with DENSE in human subjects.