DWI comments

[neurolabusc]

In general, these sequences look very nice. I am not an MRI physicist, so take my comments with a grain of salt. On the other hand, as a user, I really appreciate these well documented sequences. I have a couple of minor comments regarding the DWI sequence

1. The DWI sequence uses 64 slices with multi-band of 4. I thought the general rule is that the number of slices divided by the SMS factor should be an odd number to get symmetric stimulation. See figure 3 of Barth et al..
2. It might be worth considering whether partial Fourier should be used. When I set up DWI sequences the MRI physicists I worked with described the tradeoffs and noted that partial Fourier is not compatible with some post processing.
3. It might be worth objectively measuring the SNR to evaluate your DWI choices. I realize there are pros and cons to many of these options, but having a measure provides some basis for preferring one setup over another. In particular, it might be worth emphasizing that clinical sequences work well on the clinical head coils, not just the high-channel count research coils. The combination of in plane (iPAT) and between slice (SMS) do demand a reasonable number of channels. In my experience, obesity and kyphosis both limit the coil selection for clinical populations.

[lrq3000]

Hello @neurolabusc , thank you very much for having a look at our protocol and your insightful comments!

I did not know about Barth et al, I will check that out, thank you!.

Anyway, for both the multi-band acceleration factor and the partial Fourier, they cannot be avoided if one wants to maintain the protocol total acquisition time below 30 minutes, in which DWI is the longuest sequence and so the major time sink. Yes, partial Fourier prevents using mrdegibbs, which is very useful to remove Gibbs artifacts but not absolutely necessary. There are also other limitations, such as no reverse acquisition (I could not program the MRI to do that, this needed developer access for which I am not qualified), I should probably add a note in the README to clarify that.

Yes for the SNR that is planned, I plan to write a paper with additional tests, but maybe I can just run off a quick SNR calculation and publish the results here meanwhile, thank you for the suggestion :-) Our protocol works well with both 20 channels (I guess this is what you mean by clinical head coil?) and 64 channels, without modifying the parameters (except of course the activation and placement of antennas, which should be automatic with this machine). We don't use the 64 channels coil because the coil head frame is too small to fit our population with spasticity, they tend to place their heads in weird positions, but if you can use this coil on your population then go ahead! It certainly provided slightly higher resolution and more powerful cortical signal (but lowered subcortical signal) when we tested this coil :-)

[neurolabusc]

1. I have not used the Siemens Vida XA series, but if it is like the Siemens VA..VE series, you can set up reversed phased encoding by clicking on the '...' symbol to the right of the Phase enc. dir. option and setting this to "180". You would set up the first scan with P>>A + 0 and the second with P>>A + 180 and use the Copy References to ensure both scans use the same slice center and angulation. Do check this out, there is a known bug in VE11C that causes the automatic setting of P>>A + 180 to revert to R>>L which I believe is fixed by a service pack. For E11C without this you have to set the +180 individually each time the sequence is planned. This is one reason my team uses the CMRR research sequences instead of the Siemens product sequences (the CMRR sequences have a 'Reverse RO/PE' checkbox in the Special tab).

2. I do really like acquiring images with both phase encoding directions and using TOPUP/Eddy to correct them. One suggestion would be to turn off the iPAT which would allow you to increase the multi-band a step. The iPAT reduces the spatial distortion at the cost of SNR. With he higher multi band would be a step to acquiring both phase encoding directions, which would allow you to have the SNR and remove the distortion in post processing.

3. Are you absolutely bound to use the Siemens gradient directions, or would you consider using custom gradient directions
   http://www.emmanuelcaruyer.com/q-space-sampling.php
   this would have two benefits:
   a. You could adapt the number of directions in each series to allow both phase encoding directions.
   b. You could use whole sphere sampling, whereas the Siemens defaults are always half sphere
   https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/eddy

I strongly suggest looking at the Eddy If you haven't already acquired your data suggestions. Again, acquiring SNR maps will allow you to empirically test these modifications.