Question about processing MEGA-sLASER GABA data in Osprey 2.9.6 (Siemens 3T)

[georg]

Hello,

I have a question about processing MEGA-sLASER GABA data acquired in the cerebellum using Osprey 2.9.6. Our acquisition parameters are: TE = 68 ms, 128 averages, water suppression, editing pulse frequency [1] = 7.50 ppm, editing pulse frequency [2] = 1.90 ppm, and editing pulse bandwidth = 80.00 Hz.

I noticed that the overall data quality was best when I switched off both spectral and sub-spectral alignment, i.e., opts.SpecReg = 'none' AND opts.SubSpecAlignment.mets = 'none'.

With these settings, I currently have 40/58 subjects with usable spectra. The remaining subjects still have issues that I have not been able to resolve using different combinations of the available alignment options, so I wanted to ask whether there are any additional approaches I could try before excluding these datasets.

Case A.1 :- This looks like it may have an inverted GABA peak. I thought that aligning the sub-spectra before subtraction using L2Norm might correct this, but it did not (Case A.2).

Case A.2

Case B :- I am not sure whether this would still be considered a usable GABA+ peak, a few of the GABA+ peaks look like this.

Case C :- This appears to have a subtraction artifact that I could not correct with any combination of the alignment options I tried.

I would appreciate any comments or suggestions on something else I can try. At this point, I am concerned that these datasets may ultimately need to be excluded from the final analysis.

Thank you!

[Andy]

Hi,

The problem is happening because the algorithms “use” the stimulated out-of-voxel echoes for alignment together with the normal spectral peaks.
Since the data are already acquired, you cannot do anything to get rid of them.
However, I would try using RestrSpecReg alignment (and restrict the range to 2 ppm - 3.4 ppm) instead of Robust.

As you can see from the image, you need to be sure that MRSCont.opts structure contains the SpecRegRange parameter. Alternatively, it should be the fit.range structure which defines the area used for preprocessing, but I would not go for it in case of GABA, because the MM09 would be gone.

As far as I remember, I had to change the code a bit to read the SpecRegRange from the json job file.

Best,
Andrei