Hello,
I’m new to MRS and want to anlysis my MRS data with Osprey. Here I have serval questions about prepocecing and quality measures in Osprey:
As experts’ consensus recommendations mentioned that prerpocssing of MRS should include many steps such as motion correction, water peak removal, phase correction. I wonder which preprocessing steps are included in osprey and which functions are resposible for those prerpocessing steps (I have looked the Osprey docs but it only mentioned eddt current correction, spectral registration and alignment)?
Is there parameters can evaluate the qualtiy of acquired MRS data, and are they implented in Osprey pipeline?
What extend will influeces the fit results if using a basis set of same MRS sequence but acquired with different field strength or different TR and TE?
Osprey conforms with the experts’ consensus recommendations. There is some detail in Georg’s original Osprey publication, and the full processing pipeline can be found in the code, specifically in OspreyProcess.m. It’s fairly well commented and details each processing step taken.
There are some parameters that are consensus recommended quality indicators, such as SNR and FWHM, and these are reported by Osprey.
I would certainly recommend having field strength- and TE-specific basis sets. For J-coupled metabolites in particular, the actual spectra can look markedly different for different parameters and this will adversely affect your fit results if you don’t account for it. TR is less of an issue, as it doesn’t directly affect the profiles in any way, other than an overall T1 weighting. However, in both cases—while we do attempt to correct for T2 and T1 weighting in the water-referenced metabolite amplitudes—the relaxation constants for each metabolite are estimated from previous in-vivo studies and will be influenced by other factors, particularly if you’re dealing with pathology.
If you’re having issues in generating appropriate basis sets, I’d recommend FID-A or Steve Hui’s MRS cloud which was built on FID-A. If you’re going to use Osprey, these are convenient as they all built on the same data structures.
I would go two steps further:
a) basis sets must match TE and field strength as exactly as possible (within a low single digit percentage), this goes beyond a mere recommendation
b) I haven’t seen anyone simulating basis sets with a TR effect included, we always assume perfect T1 relaxation for the basis set, and any T1 effects are, as you say, corrected for during quantification based on literature
Thank you for the advices, that’s really helpful. I’ll try to go through the OspreyProcess pipeline and use FID-A to generate the appropriate basis sets.