MEMPRAGE RMS (not echoes) for recon-all with fMRIprep

, ,
1

Short question:
How can I specify fMRIprep to use the MEMPRAGE RMS T1w scan for reconstruction, and not the multiple echoes?

Long version:
We collect one MEMPRAGE scan per subject, and extract it with heudiconv. The Siemens console automatically generates a root mean square (RMS) average of the 4 echoes. Doug Greve suggested using just the RMS scan for FreeSurfer recon-all:

From: Douglas Greve
Date: Thu, May 9, 2013 at 8:44 PM
Subject: Re: [Freesurfer] Working with MEMPRAGE
To: freesurfer at nmr.mgh.harvard.edu
the dcmunpack command is right. You should make sure to analyze only the
RMS in recon-all (ie , that is the only volume that goes into mri/orig

Heudiconv produces these five files in the anat directory:

sub-MOB1023_ses-01_acq-MEMPRAGE_run-1_echo-1_T1w.nii.gz
sub-MOB1023_ses-01_acq-MEMPRAGE_run-1_echo-2_T1w.nii.gz
sub-MOB1023_ses-01_acq-MEMPRAGE_run-1_echo-3_T1w.nii.gz
sub-MOB1023_ses-01_acq-MEMPRAGE_run-1_echo-4_T1w.nii.gz
sub-MOB1023_ses-01_acq-MEMPRAGErms_run-1_T1w.nii.gz

(I noticed the RMS JSON has copied the echo time of the first echo, but that’s an issue for another day.)

By default, fMRIprep uses all 5 MEMPRAGE scans. How can I specify it to just use the RMS for reconstruction?

Even longer version:
The multiple echoes can be used to remove dura and improve the cortical surface reconstruction, using the -dura flag:
https://doi.org/10.1016/j.neuroimage.2007.12.025

Bruce Fischl Fri, 10 May 2013 09:42:02 -0700

did you run recon-all -all all the way through? you should name your echoes something different like echo1.mgz, echo2.mgz… then specify -dura echo%d.mgz 4

Can the -dura flag with the four files be used with fMRIprep?

Finally, we also collect T2 FLAIR and T2 CUBE scans. I’m guessing the -FLAIRpial or -T2pial flags do a better job than the multi-echo dura improvement, but can they both be used? Did the Freesurfer team ever complete the pial refinement in FS v6.0? This thread indicated there was still work to be done: https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2017-February/050019.html

On Tue, 7 Feb 2017
yes, V6 was not meant for T2/FLAIR and we didn’t test it extensively. We
have been optimizing things in the past week for T2/FLAIR and I expect 6.1
will be better for them. We can give you a beta soon if you like.
cheers
Bruce

2

Hi Kayle,

How can I specify fMRIprep to use the MEMPRAGE RMS T1w scan for reconstruction, and not the multiple echoes?

This capability doesn’t currently exist. Right now the main way I can think to do it is to temporarily remove the *echo-* files when running fMRIPrep. But we should make it easier to do that. Please open an issue on sMRIPrep (we’ve split out the structural components of the workflow) and we can work out the best syntax for this. It will probably need to be something like --acq MEMPRAGErms.

@Gilles_de_Hollander might have more useful thoughts on MEMPRAGE, though…

(I noticed the RMS JSON has copied the echo time of the first echo, but that’s an issue for another day.)

We don’t currently use echo time for analyzing anatomical images, so that’s at least functionally irrelevant. I would probably manually remove it for correctness, though.

Can the -dura flag with the four files be used with fMRIprep?

Not right now, but it would be a lovely contribution! If you’re feeling up to it/have a lab member who wants some Python experience, we can certainly walk through it. In the short term, however, probably the best way to do this would be to run FreeSurfer separately with the options you want, and put the subjects directory in the fMRIPrep output directory, and let us detect and use it.

This won’t resolve the problem of us using all of the echoes for fMRIPrep T1w preprocessing, though, so let’s definitely fix that.

Finally, we also collect T2 FLAIR and T2 CUBE scans. I’m guessing the -FLAIRpial or -T2pial flags do a better job than the multi-echo dura improvement, but can they both be used?

fMRIPrep will use -T2pial/-FLAIRpial if we find T2w/FLAIR files, but I know nothing about -dura and its interactions. I would ping the list again. (And feel free to report back, so others can find the answer more easily.)

1 Like