Analysis of spectral changes and filter design in tagged cardiac MRI

Tagged cardiac magnetic resonance images are often pre-filtered before tag lines are tracked to reduce noise, emphasize tag lines, and, in grid tagged images, suppress tag lines in one of the two directions. Cardiac deformation causes changes in the spectrum at each time frame, and an effective pre-filter should cover as much of the deformed spectrum as possible around the peaks from a given tag line direction while minimizing the leakage from peaks in the other tag line direction. In this paper, we propose a model for spectral changes due to cardiac deformation, and use this model to motivate the design of a composite Gabor (C-Gabor) pre-filter. The performance of this filter was compared to standard Gabor and HARP filters in ten human imaging studies. Experimental results demonstrated that the C-Gabor filter was superior to the other two filters because it achieved better performance over the largest range of parameters