I am using the AMARES method from jMRUI to quantify MRS signals. I have some questions about the result.
The first image is an unsuppressed water signal from muscle which has been apodized and phase corrected. So there, as you can see, the amplitude is close to 0.300. When I quantify using the AMARES algorithm (second picture), the result shows that the amplitude is 0.0177, which is pretty different than what’s visible to the eye. Does someone know what is happening here? Why is there so much discrepancy?
Another thing is that the line width (shown as Linew (Hz)) in the result is 10.86. This is equivalent to the half-width at full max of the spectra (if I am not wrong). When I calculated it manually in the phase-corrected spectra just before performing the AMARES analysis, I found it to be very different. I simply put my cursor at the peak, calculated half of that peak, found the x-axis values at the half, and subtracted them. I got something like 50 Hz but here at the results, it’s pretty low. Does anyone know what’s the reason? What am I missing here?
Any help/ insights would be appreciated. Thanks for your time.
You’re confusing amplitude for peak height. That is not the same thing.
AMARES fits signals in the time-domain, likely with a Lorentzian, depending on what you choose. The amplitude of that model is the equivalent of the integral over the signal in the frequency domain.
For FWHM estimation, I will say that this peak is not very symmetrical. You might try allowing some phase on it, as it appears that you are currently constraining the phase to be 0.0 deg
Thanks for the reply.
I understand the amplitude part.
For FWHM, is the line width (linew) in Hz equal to the FWHM of the signal then?
On a related note:
This peak looks more symmetrical than the other. The question remains, when I manually calculate the FWHM using the points that I see by placing the cursor on the spectra, that doesn’t quite match with the number found in the results section, which is 10.86. What might be the reason for the difference?
One thing that comes to mind is that an in-vivo signal is almost never a pure Lorentzian (as you have modeled it here, indicated by the ‘L’), but rather a Gaussian/Lorentzian convolution (a so-called Voigt profile). If the lineshape isn’t modeled adequately, the linewidth estimate will be off, although in this case here the residual looks ok.
What FWHM did you estimate by eye-balling for this peak?
I can only find Lorentzian and Gaussian options when I perform quantification. Where can I find Voigt profile?
So, Here is another set of results. Eye-balling the FWHM gives me a value of 31.1547 Hz. Again, this was done by putting the cursor at half of the maximum of the spectra and then subtracting the x-axis values. I expect the FWHM to be something close to 30. But AMARES says the line width, which is the same as FWHM apparently, is 10.86 Hz, as you can see in the screenshot.
Surprisingly, this value is pretty consistent throughout all the times I have used this unsuppressed water spectrum. Can’t think of an explanation for that.