Various imaging techniques using ultrasound, fast CT, and conventional MRI have been used to observe the contraction of the cardiac left ventricular chamber. None of these techniques, however, can actually determine the way in which the muscle fibers deform during contraction. MRI tagging permits a grid pattern of tissue to be magnetically """"""""tagged"""""""" at end diastole. Since the actual molecules of tissue are """"""""tagged"""""""" in this way, as the myocardium contracts the observation of the deformations undergone by the tag lines will yield information about how the actual myocardial fibers themselves deform. We have developed an image processing technique, based on Fourier filtering of the tagged images, which can permit the tag lines to be automatically tracked over time and space. The method appears to work even in the presense of low signal to noise ratio. By Fourier filtering, the horizontal and vertical lines making up the tag grid can be separated, permitting easier tracking. The method has been tested in three ways: on images of muscle tissue in which the tags are known not to move; by artificially deforming a tagged grid pattern in a known way and measuring the degree to which the algorithm predicts the known location of the tag lines; and by comparing the algoritlhm's performance to a human observer's performance. Although the technique was developed using tag lines from an 0.5 Tesla MRI unit, such a scanner often gives poor myocardial images. It is hoped that the 1.5 T machine will ultimately be able to provide myocardial tagging as well.
