Automated detection of noninvasive magnetic resonance markers

An image processing technique is developed that automatically identifies noninvasive magnetic resonance (MR) markers in tagged cardiac MR images. The algorithm uses correlation images. As the value of each pixel in the correlation image reflects the likelihood that an MR marker is present, marker locations appear as high intensity regions in the correlation image. These regions are isolated using knowledge of tag grid spacing and a threshold of lowest acceptable correlation. The accuracy of the algorithm is demonstrated using images from human volunteers and from a specially designed rotating phantom. The technique has the advantage of removing subjective bias and realizes a substantial savings in time over manual analysis methods