I am currently working on processing data from a 3T MEGA-PRESS using LCModel. I have a few questions that I need to clarify before proceeding:
File Formats: Is it mandatory to input files with the .rda suffix? If I only have files with the .dat suffix, is using other software to process the data my only option?
Basis Sets for Commercial Sequences: If a sequence is a commercial one, such as the Siemens product sequence, does it imply that there is a corresponding basis set available for it?
Using FID-A with LCModel: I have tried using LCModel on a Mac with the compiled LCModel software, using “RAW” data generated by FID-A as input. Is this acceptable, or is it best to use .rda files directly from the machine for preprocessing?
LCModel on Linux: I attempted to use LCModel on Ubuntu 22.04, but encountered GUI issues, such as some files not appearing in the LCModel file directory. I have heard that it works well on CentOS 7.9. How can I resolve these issues without changing my operating system?
Many would argue that Twix (.dat) is a better starting point, since it’ll allow more flexibility in terms of preprocessing (and the possibility to use more advanced methods for alignment and artifact rejection than supported by LCModel’s builtin tools). So, this is not a problem at all – but it does mean you’ll need some extra tools to make it work. It sounds like you’re on the right track with FID-A (3).
LCModel’s import utilities are no longer maintained, so we can expect support for vendor-specific data formats to drop off as scanner software evolves and LCModel’s standard tools do not.
Not necessarily. For MEGA-PRESS, you could start with these from Purdue: Dr. Dydak's Lab - Basis Sets, or the basis sets shipped with Osprey (the latter would need conversion for LCModel). I’m not that familiar with the Siemens product sequence, so perhaps someone else can comment on which is more appropriate in this case.
In principle yes, provided you’re using an appropriate pipeline then taking FID-A generated RAW data should be fine. In any case you should inspect the results to make sure it’s reasonable, eg paying attention to any shift or stretch in the frequency dimension.
LCModel (+gui) works well for me on Debian and Ubuntu 22 systems, although I vaguely recall some initial issues using the GUI with a remote console (I think the solution involved xauth). If you can share the error message, I’ll try to give a more concrete suggestion.
Otherwise: note that the Osprey package can automate the preprocessing of Twix (.dat) files and can perform fitting with lcmodel, so you may find that this is a convenient solution to several of the points mentioned.
Thank you for your reply; you have been incredibly helpful, and I’ve gained a lot of useful information!
Regarding Question 2, I have used BrainGPS to obtain the basis set. However, I have encountered an interesting issue: our sequence (3T MEGA-PRESS, Siemens product sequence) has an edit-off frequency of 4.7 ppm, whereas most of the literature I’ve read indicates an edit-off frequency of 7.5 ppm.
Is this discrepancy normal? Can anyone provide clarification or assistance?
Looking at the parameters you provided, I think this is just a question of terminology. The “mirror frequency” isn’t the edit OFF frequency, but rather a point mid-way between edit ON and OFF.
Typically for GABA+ (TE=68 MEGA-PRESS) the editing pulses are applied symmetrically to water. So, the edit ON targets the GABA C3 resonance at 1.9 ppm, and edit OFF is the same distance on the opposite side of the 4.7ppm water peak: 4.7 + (4.7 - 1.9) = 7.5 ppm. So here it looks like the Siemens product sequence is specifying the mirror frequency (4.7 ppm, water) from which it derives the symetrically-placed edit OFF frequency (7.5 ppm).
