Extracting robust features from cardiac magnetic resonance image contours for detecting dilated cardiomyopathy

Aim of this study was to extract robust features from cine magnetic resonance images (MRI) for phase estimation in dilated cardiomyopathy (DCM) patients. The found features may serve as additional diagnostic parameters. Previously found markers could be improved and were applied to an enlarged subject group. Four quantitative features based on the phase deviation between the left and the right ventricle (INTER, INTER-VOL) and the nonuniformity of left ventricular movements (INTRA, INTRA-VOL) were shown to be significantly different for DCM patients compared to healthy subjects. The noise-resistance of the features allows the utilization of fewer MRT slices and enables to reduce the contour detection accuracy permitting a more feasible analysis of future data.