Thomas Kirchner1, Anke Henning1,
Klaas Paul Pruessmann1, Peter Boesiger1
1Institute for Biomedical
Engineering, University & ETH
In parallel 1H MRSI, artifacts detrimental to spectral quality can be due to low sampling resolution and residual aliasing. We therefore introduce a novel reconstruction method that minimizes the deviation from a predefined target shape for the spectral response function (SRF). Based on undersampled 3T in vivo MRSI data, we demonstrate that by this method, a significant reduction in spectral contamination originating from subcranial fat can be achieved. In particular, Gaussian target shapes lead to vastly improved fat artifact suppression compared to standard reconstruction.