2-D segmentation of left ventricle in magnetic resonance images based on an optical flow algorithm

摘要

The optical flow enables the accurate estimation of cardiac motion. In this research, this tool is exploited to obtain the 2-D segmentation of the left ventricle in a cardiac sequence of magnetic resonance images. The proposed technique consists in taking a manually extracted left ventricle contour corresponding to an instant of the cardiac sequence. Then, the Left Ventricle contour for the next time instant is estimated based on calculation of the optical flow using the Horn and Schunck algorithm. The method is validated using two synthetic image sequences whose velocity field is known. Then, the technique is applied to a 3D magnetic resonance image sequence and it is validated by performing the comparison with respect to the manual segmentation using the Dice coefficient. The results show that the segmentation near the base, the equator and the apex have a Dice coefficient higher than 84%.