Regarding your first question – I don’t think that’s too surprising. You might want to look into the linewidth/FHWM for these subjects (I believe it’s reported in the QC parameters tab of the output) – it’s likely that these vary between subjects and between regions, and likely that in some subjects (perhaps those with broader linewidths) the algorithm will have a harder time isolating some of the more difficult components. For this reason, a number of the major metabolites which you’re most likely interested in will often be reported as an aggregate sum of related but difficult-to-separate signals, eg tNAA (NAA+NAAG), tCho (PCho+GPC), tCr (Cr+PCr), Glx (Glu+Gln+…).
Regarding your second question: hard to say, but it does look a little low to me, and (maybe correspondingly) some of your concentration estimates perhaps a little higher than expected – but not outrageously so. You could try experimenting with the baseline flexibility a bit, or see what happens if you take it away completely (--baseline_order -1), but I’m not sure you’ll get much benefit in doing so (if anything, I’d speculate that slightly over-estimated broad MM components may encourage the baseline to sag a bit… but I don’t have any good advice on how to check or tune this)
Hi @wclarke : one thing I’ve noticed in several of Verena’s plots is what looks like a single incongruous datapoint at the far left of the ppm range – I had a quick look through the code and couldn’t find an obvious cause, nor could I determine whether this was purely visual or could actually impact the fit… but this could potentially be a factor here.
Sorry for the delay, I’ve just returned from two weeks holiday.
Alex is correct on the first point, a number of peaks are highly correlated (with correlations <-0.5) and so the program can use one to fit both, I would therefore suggest using the --combine command to sum these peaks (as has been done automatically with Cr and PCr), and the to look at these combined results. Another way is to use the MCMC solver (--algo MH) which takes longer but will do a much better job of estimating the peaks together (but not at actually fitting the true underlying contributions!).
@VerenaD Do you add the MM to a separate metabolite group as described in the documentation? Otherwise maybe try reducing the baseline order as Alex suggested.
The small graphical artefact is just the edge of the optimisation zone (at 4.2 ppm), the baseline is only estimated across this and therefore can create a step. I should make this plot show only within the bounds to avoid this small tick.
thank you for your answers. @alex I experimented a bit with the baseline and we decided to leave it at the default setting. And so if we are interested in Glu we should combine it with Gln to Glx?
@wclarke I think I will try to go for the MCMC solver - thank you for that advice.
And yes I add the MM to the fsl_mrs command with --metab_groups MM09 MM12 MM14 MM17 MM21.
Hi All,
Sorry to be late to the conversation!
A few minor points that might be helpful:
The _act.SPAR and SDAT files that contain the water suppressed data will be eddy current corrected when the flag that @admin mentioned PostProc-> Spectral Correction is set to YES
Even if that flag is set to YES, the corresponding _ref.SPAR and SDAT files for the water unsuppressed data will not be eddy current corrected
To take full advantage of the processing pipeline available with fsl-MRS, if you have a research agreement with Philips, you can export the .data/.list files quite easily. There files will have each transient and each coil stored separately. In the export of these files you can choose whether or not to apply the eddy current correction in the exported data
My images were saved as .dcm files, but I think the ECC processing is transferable. I have now applied ECC to the ref data only in my processing steps.
When exporting the data, we only have the option of saving it as enhanced or classic DICOM and suppressing the patient data. At least I don’t remember seeing anything like that - unfortunately.
@erin.macmillan
I just had a look at the saving options in the scanner and we have a second tab for non-DICOM export.
Do you recommend changing all the files to _act.SPAR and SDAT files or can we leave the files as they are for our analysis ?
Hi @wclarke, I am asking my question under this thread as it’s relevant. I wanna simulate a basis set for my data (PRESS sequence, TE135 , BW :1200, Npt: 1024, 3T Siemens scanner). I could not find the PRESS.json file on the downloaded example dataset and none of the mentioned pre-existed basis sets were compatible with my data. So, my question is how can I get the RF phase and amplitude values for the mentioned sequence? I looked into twix header by spec2nii dump, but there is no such info there.
Could you help me with this?