Thanks Tibor,
I was also wondering, did you upsample (reslice) the images before co-registering and averaging? Would you recommend this?
Thanks, Vincent
-------------------------
I had similar setup and for me the best approach was to register everything to the first B0 (using the parameter settings from the eddy script), then average the three run producing one dataset with 65 volumes.
I would not recommend to register anything to an averaged reference, since they would possess different smoothness/"effective" spatial resolution.
And yes, after the registration and averaging you do not need eddy correction.
Auer, Tibor M.D. Ph.D.
MRC Cognition and Brain Sciences Unit
15 Chaucer Road
Cambridge
CB2 7EF
United Kingdom
Phone/Work:
Mobile: +44-(0)7906-86-3837
Mail: [log in to unmask]
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Vincent Koppelmans
Sent: Saturday, January 12, 2013 3:46 AM
To: [log in to unmask]
Subject: [FSL] DTI preprocessing: averaging of B0 and directions
Dear FSL experts,
I have a question on averaging a DTI dataset, in order to prepare it for further (TBSS) analysis.
I have a 4D set, containing 195 3D files, that is constructed as follows:
1 B0
64 directions
1 B0
64 directions
1 B0
64 directions
As you probably already have guessed, the three series of 64 directions reflect that every direction has been sampled three times.
I am wondering what the best approach is for averaging this dataset. I am thinking about averaging the B0 fields, averaging the sets of directions, and subsequently registering (flirt with 6 DOF) the 64 mean diffusion images to the mean B0.
In steps:
1) register B0-1 and B0-3 to B0-2 (flirt -dof 6)
2) average the three B0 images (fslmaths -add -add; fslmaths -div 3)
3) register the first set 64 diffusion images and the third set of 64 diffusion images to the second set of 64 diffusion images
4) average the 64 triplets
5) register the 64 average diffusion images to the average B0 file
Is this the right way to go?
If I am correct, this means I would not need to run FSL's eddy_correct anymore, right?
Thanks for your help,
Vincent
|