Improving Harmonic Phase Imaging by the Windowed Fourier Transform

This work proposes a new methodology for the extraction of Harmonic Phase images in cardiac tagged Magnetic Resonance Imaging. The procedure draws upon the use of the Windowed Fourier Transform, which provides a spatially varying representation of the signal spectra. The spectral peaks of the local Fourier domain are then extracted by a conventional Harmonic Phase recovering technique in such a way that the peak position is separately estimated for each spatial location. Finally, the phase of the complex image is reconstructed by inverting the transformation. It turns out that the proposed methodology substantially outperforms the classical Harmonic Phase analysis on a simulated dataset and seems to be more accurate for determining the tag intersections in a real dataset.