Baseline modeling in FSL-MRS

Hi all,

I’ve obtained a few spectra with different B0 shimming algorithms in the Frontal region at a 3 T Siemens scanner (unedited PRESS sequence with TE:135). I’ve used a simulated basis set from MRScloud, removed the water peak, and used a default baseline modeling in FSL-MRS (second-order polynomial). My question is if the modeled baseline is optimal in these spectra. Regarding the fitting results what CRLB threshold is considered optimum for a given peak?

I appreciate any feedback on my data!

I think Will might be on paternity leave at the moment, so let me chime in here for a bit:

  • The use of hard cut-offs of relative CRLBs is discouraged since it biases against low concentrations (see here), but as a rule of thumb, lower CRLB is better.
  • The first (and maybe only?) thing to look at when comparing different B0 algorithms should be the linewidth estimate. Expert consensus states that anything over 0.1 ppm should be considered ‘poor linewidth’. I think that’s harsh, but probably not terribly wrong if you’re interested in the coupled signals like Glu, Gln etc.
  • Looking at these plots, number 2 and 5 are clearly the best in terms of linewidth (and, I would assume, that will translate into lower CLRB). 1 and 3 are much worse, and 4 is basically unusable. You want to look for at the very least clearly separated tCr and tCho signals, and some structure in the Glu/Gln/GSH/GABA/Asp regions (not just one big hump).
  • I wouldn’t worry too much about the baseline in this scenario. There is very little MM/Lip left at TE = 135 ms, and as you can see, the baselines are all well-behaved except in 4, which should be discarded due to poor shim anyway.
2 Likes

@admin Thank you for your response. I have another question which is about the water suppression. Is it normal to see a huge water peak as you can see in the following image in a water supposed spectra? For these, I did manual shimming and adjusted the f0, but I doubt if something was wrong with the water suppression.