Public basis set (or refocusing pulse waveform) for Siemens sLASER CSI at TE=40ms

Analysis / Modeling Basis Sets
1

Hi all,

I’m processing a 3T Siemens CSI dataset (16×16 spatial matrix, single slice, 1024 pts, BW=1200Hz, TE=40ms, sLASER localization) with FSL-MRS, and I’m stuck on finding/simulating a matching basis set.

What I’ve tried:

  1. MRSCloud (shui5/MRSCloud): Attempted to simulate a Siemens sLASER TE=40ms basis set. Tracing through load_parameters.m, sLASER localization requires vendor-specific refocusing pulse files (Philips_GOIA_WURST_100pts.mat / GE_GOIA_WURST_100pts.mat), which are intentionally excluded from the public repo (per the README: “product sequences and rf waveforms are not shared”). PRESS has a public substitute pulse (univ_eddenrefo.pta) that works, but sLASER simulation seems to be a dead end without proprietary pulse files.
  2. Public basis sets: Used the closest available — a published Siemens sLASER TE=20ms basis (BW=4000Hz, 2048pts, zero-padded to 4096pts to satisfy fsl_mrsi’s coverage requirement). Also tested a GE sLASER TE=30ms alternative for comparison.

Cross-validation against an independent TARQUIN fit of the same raw data:

  • Singlet metabolites (tNAA/tCr, tCho/tCr, mI/tCr) agree well between FSL-MRS (using the TE=20 basis) and TARQUIN — reassuring that the pipeline itself is sound.
  • Glx/tCr is consistently ~2–2.5x higher with our TE=20 basis than TARQUIN’s real TE=40 fit (0.70 TARQUIN vs 1.76 ours). The GE TE=30 basis partially closes this gap (1.39), consistent with a TE-dependent J-coupling evolution effect as expected physically (Glx is J-coupled/multiplet, NAA/Cr are singlets and TE-insensitive).

My questions for the community:

  1. Does anyone have (or know of) a public Siemens (or any vendor) sLASER basis set at TE=40ms they’d be willing to share, or point me to?
  2. Alternatively, does anyone have access to the GOIA-WURST refocusing pulse waveform (or an equivalent public substitute for sLASER, similar to how PRESS has univ_eddenrefo.pta) that would let me simulate one myself in MRSCloud?
  3. Short of that, is there a recommended way to correct/bound the Glx overestimation from a TE-mismatched basis (e.g., a published correction factor, or is reporting Glx only for relative/group comparisons the accepted practice)?

Happy to share more details if useful. Thanks in advance!

2

Hi @amoir710

For MRSI, the exact pulse shapes are less important, as the spatial resolution of the readout isolates the CSDE / pulse profile induced effect that the simulation is attempting to simulate. (This rather suggests that we should be using different basis sets for voxels around the edge of the localisation volume, but no one does this and don’t worry about it). However the shortest route for you is probably to use the hosted version of MRSCloud (requires free registration). That has the Siemens sequence (plus pulses) integrated into it. The basis_tools convert command should be able to handle the output of this.

If that doesn’t seem to work, do come back and we can try to figure it out.