Test-retest - MEGA-PRESS - Gannet absolute conc

Hi,

I ran three consecutive MEGA-PRESS scans using the following parameter at Siemens 3T Skyra.
TR/TE=2000/ 68 ms, Average=96, voxel =3x3x3 and one non-water scan.

I processed the data using Gannet software. The stability between the second and third scans is less than 5%, and the first and second scans are around 15%, even though I used four preparation scans. Is it acceptable?
Why are there more fluctuations between the 1st and second scan compared to 2nd and third scan? Is there any reason for this?

Also, could you please tell me how to calculate the absolute concentration of GABA and Glx using Gannet?

Thanks

Hi, @narayanan76. To assess the quality/reproducibility of your data, it would be very helpful to see your GannetLoad and GannetFit outputs. If you prefer, you can email your outputs to me at mam4041[at]med.cornell.edu.

In terms of calculating absolute concentrations, this depends on what data you have acquired. Since you have an unsuppressed water scan, GannetFit will give you a water-scaled concentration that is semi-absolute but not corrected for voxel-dependent partial volume effects (i.e., not tissue corrected).

For a “fuller” quantification, you will need T1-weighted structural images. Running GannetCoRegister (after you’ve run GannetLoad and GannetFit) will first co-register the MRS voxel to the respective T1 image. Subsequently running GannetSegment will segment the T1 image and MRS voxel mask. Once you’ve done this, you can run GannetQuantify to obtain tissue-corrected (pseudo-)absolute concentrations for GABA and Glx.

More details on how Gannet calculates absolute concentrations can be found here.

Let me know if you have any questions about any of this.

–Mark

I would add that test-retest CV of GABA-edited MEGA-PRESS is usually on the order of magnitude of ~5-15%, so I don’t think the variation that you have observed is too outlandish. I would maybe expect it to be slightly better if you ran them back-to-back (without repositioning). But 96 averages isn’t that much (is that 96 in total, i.e. total acquisition time of 192 sec, or 96 ON + 96 OFF, i.e. a total acquisition time of 384 sec?), and modeling is always susceptible to noise.

As Mark suggests, it is helpful to include GannetLoad and GannetFit outputs to judge data quality.

Cheers,
Georg