Discrepancies of SNR between MGH and MNI

[jcohenadad]

Given the restriction of the refV during the acquisition of the SNR map at the MNI, the SNR is downward biased, which is inconsistent with results from MGH which used a very similar coil:

In contrast, the B1+ maps are extremely similar between sites using the same coils:

• CRMBM vs. UCL
• MPI vs. NTNU
• MGH vs. MNI

This inconsistency for the SNR stands out, therefore I suggest we re-acquire the SNR map at the MNI, now that the refV restriction is not required anymore.

[jcohenadad]

Here is the updated figure when we normalize with the total B1+ along the cord:

[jcohenadad]

Alternatively, we should be able to estimate the requested FA a the MNI site based on the maxVoltage allowed.

[jcohenadad]

With https://github.com/spinal-cord-7t/coil-qc-code/pull/111, the discrepancies between MGH and MNI should have been addressed, but it was not. So we need to dig deeper.

Looking at the individual coil element GRE magnitude signal, we can see that the distribution and relative signal is very similar between the two coils:

[evaalonsoortiz]

Additional info: MGH and MNI reference voltages used for this scan were very similar (464@MNI vs 419@MGH). I believe that the initial limit that had been imposed on us was 400V. This was later changed to ~750V I think. I don't think the MNI scan was impacted by reference voltage limitations.

[evaalonsoortiz]

this is the figure that we currently generate in the notebook:

[evaalonsoortiz]

NTNU and MPI have a similar offset between them (compared to the offset between MGH and MNI), but the same trend along the "cord".

[jcohenadad]

MGH and MNI reference voltages used for this scan were very similar (464@MNI vs 419@MGH)

This info was retrieved from the JSON sidecar. I am wondering if this is 100% true. For example, I am wondering if the pulse Tx voltage was manually changed (that is: not the refV, but the actual excitation Tx pulse of the GRE sequence).

[jcohenadad]

NTNU and MPI have a similar offset between them (compared to the offset between MGH and MNI), but the same trend along the "cord".

Yes, very good point. Yet, we would also expect values to be 'closer' to each other.

[evaalonsoortiz]

For the MNI/MGH coil: the distance between the anterior part of the coil and the posterior casing can be adjusted (which would impact SNR). Perhaps this was adjusted differently at the two sites.

[evaalonsoortiz]

Although, looking at the images, it appears that SNR mostly differs around the posterior casing.

[jcohenadad]

Also interesting to note, is that in-vivo SNR does not differ much between MNI/MGH, which is why I suspect there is something that we are missing that could explain the discrepancies.

[evaalonsoortiz]

MGH and MNI reference voltages used for this scan were very similar (464@MNI vs 419@MGH)

This info was retrieved from the JSON sidecar. I am wondering if this is 100% true. For example, I am wondering if the pulse Tx voltage was manually changed (that is: not the refV, but the actual excitation Tx pulse of the GRE sequence).

This wouldn't be following our SOP, and I think we can be confident that scan operators followed it (I do not know who actually acquired this scan - but I could do some digging).

[evaalonsoortiz]

Also interesting to note, is that in-vivo SNR does not differ much between MNI/MGH, which is why I suspect there is something that we are missing that could explain the discrepancies.

Maybe the phantom positioning in the coil? There's a high density of Rx elements around the posterior part of the "neck". Placing the phantom a bit lower/higher could conceivably change the SNR profile.

[jcohenadad]

Another hypothesis: phantom temperature change between MNI and MGH. Although it is rather unlikely, because we're observing a ratio of about 0.7 between SNR_MGH and SNR_MNI, and given that $SNR \propto 1/{\sqrt{T}}$, a 0.7 SNR ratio, assuming 20°C at MNI (=293K) would be equivalent to a phantom temperature of -130°C at MGH ((0.7 * np.sqrt(293)) ** 2), so not likely...

[jcohenadad]

Maybe the phantom positioning in the coil? There's a high density of Rx elements around the posterior part of the "neck". Placing the phantom a bit lower/higher could conceivably change the SNR profile.

I would be very surprised if that would lead to such a dramatic change. Moreover the phantom is pretty big and tightly fits the coil, so there is no much room for large variations in terms of phantom positioning.

[jcohenadad]

Another hypothesis: The Tx mode (eg: CP mode, pTx mode, etc.) used for the SNR scan was different between the TFL scan and the GRE scan. Although this would be highly surprising, because the mode is imposed by the VOP files, which were supposedly the same between the TFL and SNR scan (ie: same session).

[evaalonsoortiz]

Another hypothesis: phantom temperature change between MNI and MGH. Although it is rather unlikely, because we're observing a ratio of about 0.7 between SNR_MGH and SNR_MNI, and given that S N R ∝ 1 / T , a 0.7 SNR ratio, assuming 20°C at MNI (=293K) would be equivalent to a phantom temperature of -130°C at MGH ((0.7 * np.sqrt(293)) ** 2), so not likely...

Wouldn't it have been nice if that math had worked out for that!