I have tried the FID-A modeling of the PRESS sequence with a couple of coupled systems.
I like the results, but i wish there was a way to incorporate T1 and T2 relaxation. I could fudge T1 by adjusting scale factors, but T2 during precession is trickier. The Hamiltonian variable has a lot of components referring to the X and Y magnetization, but they are normalized an I am not sure which I could scale with an exp(-t/T2) factor to simulate relaxation.
Ideally this would also be done with a Hamiltonian. The sys structures would have to get an additional T1 and T2 field. For now I would be happy with a shortcut that gives the correct result for T2 relaxation to simulate a TE series. The T1 is fairly easy to do with the scale factor and density.
Has anyone tried this?
Hi Ronald,
AFAIK this will be a little tricky to implement within FID-A; for example Cr is not broken down into the 3.02 ppm and 3.9 ppm signals, which will experience different T2 relaxation. I agree that additional fields within the sys structure would be helpful to incorporate it, but the spin system definitions would have to be changed as well to allow different relaxation rates to apply to different parts of the spin system.
It’s probably worth opening an issue over at the FID-A repository - I’ll also shoot Jamie an e-mail so he can respond to this thread here.
Best wishes,
Georg
Hi Georg,
Thanks for your response. I agree: the relaxation parameters in a spin system should be linked to the chemical shifts of the protons. That is really where the problem lies. I want the analyze series of spectra with different echo times (and repetition times too eventually) in a way that uses the common factors in the spectra as prior knowledge. The T* or B0 field inhomogeneity is common to all resonances in all spectra, the chemical shifts for each peak should be constant through the TE series and the T2 is a coherence along TE and ties in with T2* (linewidth). By analyzing spectra individually we throw all that away. Most important issue is: the metabolite amplitudes found for the TE series rarely follow an exponential decay. J-coupling is the worst culprit (this is why we need good simulations). Multi-compartments is another. Look at the fit residuals as a function of TE…They should be random about zero with no correlation to TE. Often they are not, even though your fit Rsquared looks fine.
So, can we use basis sets created for each TE and then fit the whole series as one? No, not if the amplitude scaling for a component (like Cr) is not the same for all resonances in the system. We could simulate Cr and then split the output spectra into two components.
When you think about this it will be needed for a realistic basis set for all spectra with longer TE, not just my TE series.
I too hope we can have a or FID-A or MRS community conversation going to generate ideas to solve these problems.
Regards,
Ronald