Hi there,
I agree, I would go for your option 1. To learn more about potential bias see the part on "interpolation asymmetries" of the intro of the following paper: "Reuter M, Schmansky NJ, Rosas HD, Fischl B. Within-subject template estimation for unbiased longitudinal image analysis. Neuroimage. 2012 in press". That paper describes issues with volume and thickness measures of t1 images in longitudinal analysis, but will also apply to diffusion data.
Best,
Koene
On Wed, 25 Apr 2012 10:23:47 +0100, Stephen Smith <[log in to unmask]> wrote:
>Hi - the simplest thing to do for now is your option 1) - this should be robust and unbiased. In the future we will probably have an option in TBSS to carry out a potentially slightly more accurate pipeline for longitudinal analyses.
>
>Cheers.
>
>
>
>On 23 Apr 2012, at 08:12, Bonnie Y K Lam wrote:
>
>> Hi FSL users,
>>
>>
>> I am doing a cross-sectional and a longitudinal study on white matter of the brain in degenerative diseases. From what I learnt from the discussion forum is that there are two ways to compare longitudinal changes (please correct me if I get it wrong):
>>
>> 1) All diffusion data are registered to standard space, create mean FA and mean FA skeleton in TBSS and register all FA images onto the mean FA skeleton. Compare results with a paired t-test.
>>
>> 2) All diffusion data are registered to standard space. Then create a averaged group template of baseline images and register follow-up images onto baseline images using FLIRT. Voxelwise comparison done with GLM.
>>
>> My questions are as follow:
>>
>> a) Which of the above method will be more suitable or accurate for a longitudinal study of 40 people (three disease groups (n=11,7,7) and controls (n=15)) scanned at two different time points?
>>
>> For method 2:
>> b)How do I create a white matter group average template for baseline images? (tbss_2_reg -n ?)
>>
>> c) As in actual command lines, how should I incorporate it into what I have ran with my cross-sectional data? So I do the same eddy current correction, bet, dtifit, etc for all subjects, then run a separate TBSS for cross-sectional and longitudinal data?
>>
>> To better illustration, these are the command lines I used for cross-sectional data, to process longitudinal data I will need to:
>>
>> eddy_correct raw_dwi.nii.gz data.nii.gz 0
>> fslroi data.nii.gz nodif.nii.gz 0 1
>> bet nodif.nii.gz nodif_brain -m -g 0.2 -f 0.3
>> fslview nodif.nii.gz nodif_brain_mask.nii.gz
>> fslmaths nodif.nii.gz -mas nodif_brain_mask.nii.gz nodif_brain.nii.gz
>> dtifit --data=data.nii.gz --out=dti --mask=nodif_brain_mask.nii.gz --bvals=bvals --bvecs=bvecs
>> fslview dti_FA.nii.gz dti_V1.nii.gz
>> (use the same as processed for cross-sectional data)
>>
>> tbss_1_preproc *.nii.gz
>> tbss_2_reg -T (change to tbss_2_reg -n ?)
>> tbss_3_postreg -S (skip)
>> cd stats
>> fslview all_FA.nii.gz
>> fslview mean_FA.nii.gz mean_FA_skeleton.nii.gz
>> tbss_4_prestats 0.2
>>
>>
>> Thank you for replying in advance !
>>
>> Regards,
>> Bonnie
>>
>
>
>---------------------------------------------------------------------------
>Stephen M. Smith, Professor of Biomedical Engineering
>Associate Director, Oxford University FMRIB Centre
>
>FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>+44 (0) 1865 222726 (fax 222717)
>[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>---------------------------------------------------------------------------
>
>
>
>
|