Basic Osprey Questions

Hi @admin,

I have a few basic Osprey questions now that I have it working- hope that’s okay.

  1. Is it okay if we use the basis set provided by Osprey? For example, I’m using /osprey/fit/basissets/3T/siemens/unedited/press/30/basis_siemens_press30.mat
    I made a new basis set using the code below from the basis set we had been using from LCModel, but I’m not seeing any differences in the data outputs between the two basis sets. Is that normal?
[BASIS] = io_LCMBasis(file,1,'unedited','none')
  1. When I use the newly created Osprey basis set (from my basis set from LCModel) and get on the “LC model” tab, my values on the lefthand side are really small (Glu=2.42e-04) in comparison to what I am seeing when I output through the LCModel software with the same rda files and the original LCModel basis set (Glu=7.694)- is there a reason for this?

  2. In the quantified tab, my tissue corrected water scaled values (and rawwaterscaled and csfwaterscaled) are much higher than what I was getting before when basic CSF correction and corrected for tissue and T1/T2 relaxation constants. Now, my final tissuecorrwaterscaled Glu value is 12.508 from Osprey, whereas before with Gannet and my scripts, it would have been about ~9. Is this because I am now implementing the full Gasparovic method via Osprey? I just want to make sure these are relatively normal values after tissue correction and water scaling and that there isn’t a greater problem at hand when I go to analyze the data.

  3. When looking at the top of the “Processed” tab, I see 0.036265 Hz/ppm after the tCr FWHM, is that the relative fit quality metric?

Thanks so much for all of your help, again I’m really appreciative

Just to clarify, what exactly was the ‘before’ method? Directly using LCModel? And now you’re using Osprey to fit or the LCModel module inside of Osprey?

  1. By default (if you don’t specify the basis set explicitly in the job file), Osprey will always automatically pick the one in the fit/basissets folder. So if you want to use the one you made, you’ll need to point to it.

  2. The LCModel outputs are very coarsely water-relaxation corrected (depending on the ATTH2O and WCONC values set in the LCModel control file). What you see in the Osprey fit tab is just the raw water-scaled amplitude ratio (metabolite amplitude divided by water amplitude, not scaled or corrected further). There is no real sense in comparing anything between Osprey and (native) LCModel because their internal methods of water-scaling are so different.

  3. It also doesn’t make much sense trying to compare Glu values from short-TE PRESS in Osprey with Glx values from medium-TE spectral edited experiments. Gannet only integrates the edited Glx signal, but doesn’t do a full linear-combination modeling, so it’s not remotely comparable. As long as you are doing everything consistently across your dataset, you shouldn’t worry too much about the absolute scaling. Every method that attempts water-scaling is just a simplification with lots of assumptions about relaxation times, water content etc. (FWIW, 12 mM doesn’t sound unreasonable for Glu.)

  4. No, that’s the FWHM estimate. You find the fit quality metric in MRSCont.QM.tables.relResA.

Hi @meggon,

  1. Glu=2.42e-04 also doesn’t make sense. More should have gone wrong there.

  2. look at:
    Metabolite concentrations calculated by Osprey - #11 by hraum
    also consider that some points are in molarity and some in molality. It’s very confusing.

Heiner, the value in the fit tab is just the coarse amplitude ratio (metabolite to water) before any scaling. Let’s say you have 10 mM Glu and 50000 mM water, you’ll arrive in that ballpark.

Ok, thanks. Now I think I have understood this point.