Hi Andreas and Matt,
thanks for your help. It's very helpful indeed. I will follow your
suggestions, and get get back to you concerning the bvecs as soon as I
have made an average of the two acquisitions.
Meanwhile, please apologize my ignorance and allow a couple of more
questions.
Q1: The third line in your bash script says that avg_2_corr is to be made
of vol 40-40. Should this read 40-80?
Q2: The big4d.nii.gz would be my 80 volume raw data (included 20
b0's)concatenated into one nii.gz? In this case, should I use MRIConvert
to convert my raw data into NiftiFSL, and then avwmerge/mri_concat into
one 4D vol, or would you suggest something else? What I have done is to
mri_convert my raw .IMA-files into .bfloat (which I need to something
else), and then mri_convert the bfloats to .nii to use with Fsl.
Q3: My first thought would be - after motion correction - to average
volumes with equal directions (in my case I guess that would be vol 11 and
51, 12 and 52 etc, resulting in 30 averaged DWI's), then average all 20
b0's into one vol and then merge the resulting 31 vols into one 4D using
avwmerge or something like that to constitute my data file. My concern is
that averaging over different gradient directions and even b0 would create
a mess, but that is maybe not even what you are suggesting or what avwroi
and avwmaths_32 does?
Thanks!
Lars
> Lars,
>
> Averaging data with FSL is easy, once you make a script to do it (call your
> starting NIFTI_GZ file with both averages big4d.nii.gz):
>
> bash eddy_correct big4d big4d_corr 0 #motion correct all data to first
volume
> bash -c "avwroi big4d_corr av1_corr 0 40"
> bash -c "avwroi big4d_corr av2_corr 40 40"
> avwmaths_32R av1_corr -add av2_corr -div 2 data #average motion
corrected
> data
>
> Then preprocess for DTIFIT:
> bash -c "avwroi data nodif 0 1" #select first b0 volume
> bash -c "/usr/share/fsl/bin/bet nodif nodif_brain -m -f .2" #make mask
of
> brain from b0 volume
> dtifit -k data -o data -m nodif_brain_mask -r bvecs -b bvals #calculate DTI
> maps
>
> We can work on getting your gradient directions oriented correctly once you
> have the data averaged.
>
> Peace,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf
> Of Lars Tjelta Westlye
> Sent: Friday, August 03, 2007 5:20 PM
> To: [log in to unmask]
> Subject: [FSL] Averaging of multiple DTI acquisitions
>
> Hi, I have a 30-direction DTI data set that is repeated 2 times. The
data
> is transferred from the Siemens scanner in a mosaic format with a total
of
> 80 volumes. Vols 1-10 is b0, 11-40 is the diffusion weighted images.
This
> sequence is repeated in vols 41-80.
>
> The Fdt pipeline says the user is to manually average multiple
> acquisitions prior to Eddy current correction.
>
> Q1: How can I do this with Fsl? Or is the averaging somehow
automatically
> taken care of within Fdt stream?
>
> I have not done any averaging, but still performed Eddy current
correction
> and DTIFit on these data. I'm not sure as to the quality of the output
files. Please see the tiny screen dump from Fslview (_V1 RGB's modulated
on _FA) and the bvecs attached. I'd be very grateful if someone would
have
> a quick look and tell me if it looks reasonable and if I'm on the right
track here before rushing into BEDPOST.
>
> Thank you very much!
>
> Best,
> Lars
>
|