要旨
Introduction:
Since 2D and 3D MRSI examinations typically last excessively
long, the use of acceleration techniques is of great importance. Here, we use a SENSE approach [1,2,3]. The Spatial Response Function (SRF) indicates which spatial areas contribute to the final voxel content, and it typically exhibits noticeable side lobes. In parallel MRSI, where the number of k-space sampling points is extremely low, this detrimental effect is particularly pronounced. As a result, voxel spectra often contain unwanted contributions by signal originating elsewhere, e.g. subcra
nial fat signal in brain matter voxels.
Methods:
In our method [4], the SENSE reconstruction matrix is calculated by minimizing, for each voxel, a custom-tailored cost function containing an SNR term and the deviation of the resulting SRF from an initially chosen target T. We use
T=Boxcar to mimic standard SENSE reconstruction, T
=Gaussian for a more realistic approximation to the actual SRF, and T=|Sinc| for comparison, all centered on the voxel of interest. Reconstruction is performed on two-fold undersampled transversal slice SELOVS MRSI [5] of a volunteer brain acquired on a 3T MR system (Philips Medical Systems, Best, The Netherlands) with an 8-channel head coil with OVS and VAPOR water suppression (Fig. 1).
Results:
Fig. 2
shows a comparison of reconstructed spectra from a voxel
in the center of the brain. Despite OVS, fat contamination peaks are clearly visible next to the NAA peak. They are
partly frequency-shifted due to shim imperfections at their original location. The use of Gaussian-shaped SRF targets results in largely reduced fat artifacts. In all cases, the remaining metabolite peaks can be clearly identified. Examples of the resulting SRFs are shown in Fig. 3. In the standard SENSE case, considerable side lobes remain.
A Gaussian target leads to a somewhat wider, but well-localized single peak.
Discussion/Conclusion:
We demonstrate that fine-tuning of SENSE recon-
struction parameters enables the suppression of residual fat fold-over artifacts. In particular, using Gaussian-sha
ped SRF targets was found to lead to spectra of superior quality compared to standard reconstruction.
