Contour estimation in tagged cardiac magnetic resonance images

Tracking magnetic resonance tags in myocardial tissue promises to be an effective tool in the assessment of myocardial motion and deformation [1]. The amount of data acquired is very large, requiring automated tracking methods. We describe a hierarchy of image processing steps that quickly detect the contours of the myocardial boundaries of the left ventricle (LV) and the spines of tags placed in the myocardium. We are currently using our technique to process short axis (SA) images containing a radial tag pattern and long axis (LA) images containing a parallel tag pattern. Other tag patterns, such as striped tags or a rectangular grid, may be detected with minor modifications to the algorithm.