Detection of cardiac abnormalities from MRI sequences by using a deformable mesh model

Abstract

In this work, we explore the potential of estimated ventricle motion, using a deformable mesh model (DMM), to detect cardiac pathologies such as hypertension and mitral regurgitation in MRI cardiac gated image sequences. In DMM, left ventricle motion was estimated by deforming a 3D-mesh along pixel-intensity variations of combined tagged and cine MRI sequences. Next, dense motion fields obtained from DMM were used to compute 3D torsion maps for the LV using a B-Spline model. Features extracted from the torsion maps were used for detection of hypertension and mitral regurgitation by performing Fisher Discriminant Analysis. Finally, detection performance of DMM motion was compared to other known motion-tracking approaches, such as Feature Based (FB) analysis and Unwrapped Phase Strain (SUP).