FID modulus: a simple and efficient technique to phase and align MR spectra

The post-processing of MR spectroscopic data requires several steps more or less easy to automate, including the phase correction and the chemical shift assignment. First, since the absolute phase is unknown, one of the difficulties the MR spectroscopist has to face is the determination of the correct phase correction. When only a few spectra have to be processed, this is usually performed manually. However, this correction needs to be automated as soon as a large number of spectra is involved, like in the case of phase coherent averaging or when the signals collected with phased array coils have to be combined. A second post-processing requirement is the frequency axis assignment. In standard mono-voxel MR spectroscopy, this can also be easily performed manually, by simply assigning a frequency value to a well-known resonance (e.g. the water or NAA resonance in the case of brain spectroscopy). However, when the correction of a frequency shift is required before averaging a large amount of spectra (due to B 0 spatial inhomogeneities in chemical shift imaging, or resulting from motion for example), this post-processing definitely needs to be performed automatically.

Zero-order phase and frequency shift of a MR spectrum are linked respectively to zero-order and first-order phase variations in the corresponding free induction decay (FID) signal. One of the simplest ways to remove the phase component of a signal is to calculate the modulus of this signal: this approach is the basis of the correction technique presented here.

We show that selecting the modulus of the FID allows, under certain conditions that are detailed, to automatically phase correct and frequency align the spectra. This correction technique can be for example applied to the summation of signals acquired from combined phased array coils, to phase coherent averaging and to B 0 shift correction.

We demonstrate that working on the modulus of the FID signal is a simple and efficient way to both phase correct and frequency align MR spectra automatically. This approach is particularly well suited to brain proton MR spectroscopy.