Hi Saad, Steve, Aga, et al.,

I've been following this discussion closely as I'm working with longitudinal DTI data and a TBSS pipeline, and I'm grateful for all of the information.  

I think the approach that Steve laid out makes a lot of sense, although I am not clear as to the benefit of using the half-way registration, rather than simply registering Time 2 (T2) to Time 1 (T1) in this scenario.  Can anybody help clarify this?  

In addition, if one was working with >2 occasions of data (e.g., DTI measured three or more times within a subject), then it seems that a half-way registration would not be feasible, and registration of all successive waves to baseline and using those transform matrices to estimate the tensor using the method Steve described seems more reasonable.  

I appreciate any thoughts you all might have on this.

thanks,
Andrew


From: Saad Jbabdi <[log in to unmask]>
Subject: Re: AW: [FSL] co-registraton of FA maps: FA values change: why and what to do?
Date: October 30, 2012 7:48:49 AM EDT

Hi Aga

I wanted to make sure I understand correctly the most straightforward procedure suggested for FSL:
1. Run DTI fit with --save_tensor for time 1 and time 2. 
QUESTION1:  Did I get it right that up sampling the images to 1x1x1mm space before warping can be skipped if we operate on tensor maps?

yes.

2. Use the warping parameters from registering nodif time2 to time1 to warp the tensor parameters maps. 
QUESTION 2: in this case, should I use the half-way registration, does it matter here?

yes you should use vecreg and the halfway matrix.

3. Estimate the FA and other maps from the warped tensor parameter maps using -tensor_decomp in fslmaths.

Is this correct?

yes

cheers
saad


Best and thanks,
Aga

Von: FSL - FMRIB's Software Library [[log in to unmask]]" im Auftrag von "Chou Paul [[log in to unmask]]
Gesendet: Montag, 29. Oktober 2012 03:15
Bis: [log in to unmask]
Betreff: Re: [FSL] co-registraton of FA maps: FA values change: why and what to do?

Dear Steve


Thanks for your kindly help, i will try it !

Cheers

Paul

Date: Fri, 26 Oct 2012 06:53:02 +0100
From: [log in to unmask]
Subject: Re: [FSL] co-registraton of FA maps: FA values change: why and what to do?
To: [log in to unmask]

Sure - if you use the --save_tensor option with dtifit you'll get the tensor components. You can then apply your transform/warp to those (no need to upsample first if you're just warping into a new space, whether at same or increased resolution) I believe using vecreg, and then recalculate the FA (etc) using -tensor_decomp option in fslmaths.

Cheers