Myocardial Scar Quantification Using SLIC Supervoxels - Parcellation Based on Tissue Characteristic Strains

Abnormal myocardial motion occurs in many cardiac pathologies, though in different ways, depending on the disease, some of which can result in negative clinical outcomes. Therefore, a better understanding of the contractile capability of the tissue is crucial in providing an improved and patient-specific clinical outcome [4]. Cardiovascular Magnetic Resonance Imaging (CMR) is considered the gold standard for the assessment of cardiac function and has the potential to also be used for routine tissue strain analysis because of its high availability in clinical practice. In this study we estimate the local strain in myocardial tissue over a cardiac cycle using cine MRI imaging to perform the analysis. To quantify the tissue displacement, we use the diffeomorphic demons registration algorithm [15] in a multi-step 3D registration, for the minimization of cumulative errors propagation. Using the displacement gradient of the deformation, individual voxel strain curves are computed. We present a novel method for parcellating the myocardium into regions based on the strain behaviour of clusters of voxels. We define the supervoxels using the Simple Linear Iterative Clustering (SLIC) algorithm [1] inside a predefined mask. The results are consistent with late gadolinium enhancement scar identification.