Hi
It seems from your description that you have a translation/rotation component between the two runs. Depending on the b-value of your acquisition, eddy_correct could do a reasonable job, so the pre-processing options you describe make sense (A,B). I wouldn't analyse the data separately.
Cheers
Stam
On 14 Jan 2013, at 04:17, M Kumar wrote:
> Dear FSL Group, Prof.Smith
>
> I will be much obliged if you could weigh in on the following question.
>
> In practice there are subtle variations in acquiring DWI data. Examples include different combinations :
>
> 1. Short DWI scans (30 directions) about 6 mins.
> 2. Longer duration (say 128 directions) about 18 mins.
> 3. Multiple averages acquired sequentially with a break in between.
> 4. Repetitions (averages) part of the same pulse sequence.
>
> Specific Condition:
> My question pertains to post-processing one such combination: Two independent (but identical for averaging), long duration (15 min) scans with a break in between to check on the subject.
>
> The subjects (patient population) likely readjusted in-between the scans as I talked to them between the two scans, even though we had cautioned them against moving their heads a priori. This in my opinion, is potentially a significant, source of motion in addition to the expected within scan motion.
>
> Options from this list :
> I have carefully studied both : Option (A) averaging the two acquisitons following 'eddy_correction' and Option (B) bvec rotation on eddy corrected concatenated data. These options appear to be suited for short duration, repetitions with limited patient motion between acquisitions.
>
> Question :
> My question is given my specific DWI acquisition variation, perhaps it is better to treat the two scans independently (as they really are), by eddy_correcting them separately. I could them compute DTI measures (FA, MD etc.) for each run and then combine them by co-registering and averaging them for each subject.
>
> Thank you in advance for your time,
>
> M Kumar
>
|