About NAA+NAAG and GABA

MRS Software LCModel
1

Snipaste_2026-02-25_09-13-40
Snipaste_2026-02-25_09-13-401920×1163 158 KB

Dear Professor,

I have a question regarding my current results quantified using LCModel. In the difference spectrum plot, both GABA and GABA/NAA are displayed.

  1. If my results have undergone water-scaling , does the label ‘GABA’ in the figure actually represent GABA/Water (absolute concentration)?
  2. Regarding the NAA+NAAG shown in the figure: Is this signal derived from the difference spectrum ? If so, I understand that this specific signal cannot be used as a reference for quantification. Could you please confirm if my understanding is correct?
  3. if the NAA+NAAG cannot use as reference, we should use the edit-off spectrum to quantify the NAA+NAAG?
2

  1. It is the GABA/water amplitude ratio, but that only very roughly translates to an estimate of concentration. That’s because LCModel cannot do tissue/relaxation correction, which requires segmentation into GM/WM/CSF. It therefore makes very crude assumptions; more specifically, it assumes pure white matter with 65% water visibility, does not perform metabolite relaxation correction, and a blanket water T_2 relaxation factor (of about 0.7) which is not adequate for medium TE. I wrote up more info in this thread.

  2. You should introduce a basis function for the co-edited MM30 macromolecules and then report GABA+. This will result in a better fit of the 3-ppm signal. More info on how to do this can be found in this thread. We also wrote a paper about this.

  3. The NAA+NAAG estimate here is indeed from the difference spectrum. There is some interesting nuance to the question whether one should report GABA+/tNAA, and it depends a lot on the context. GABA+/tCr or water-scaled (and tissue-corrected) GABA+ are likely to correlate with the amount of grey matter in the voxel (as there is more GABA in GM). If one compares datasets from participants with large variations of GM tissue fraction (say, across a large age range, or neurodegenerative disease), any effects observed on GABA+ are likely to be driven by bulk GM effects. Caldwell and Rothman proposed that, for neurodegenerative disease, GABA(+)/tNAA is actually useful because it offers an inherent normalization to healthy neuronal tissue, and actual changes to GABA(+)/tNAA would be more sensitive to actual changes of GABA(+). I find this quite intriguing.

  4. The (negative) NAA+NAAG signal in the difference spectrum is basically equivalent to the (positive) NAA+NAAG signal in the edit-OFF. Their estimates won’t be perfectly identical, but I’d expect them to correlate strongly, so if you want to use tNAA as a reference, it shouldn’t matter too much and you can just use the convenient one from the difference spectrum.

3

Thank you for your guidance. Regarding Caldwell and Rothman, I would love to delve deeper into their specific paper. My confusion stems from the fact that most literature I’ve read advises against using NAA as a reference in neurodegenerative diseases precisely because NAA levels are reduced. This perspective seems to contradict the point raised here. Could you clarify if this represents a shift in consensus or a specific context where their approach is preferred?

4

thanks a lot, your reply really cleared up a lot of my confusion.