Hi
>
> 1. If I have several runs for each subject and I want to do multi-session temporal concatenation , should I concatenate the runs of each subject using fslmerge and then give it as a single subject data to Melodic or I should just give all the runs of all subjects as 4D input data. Is there any point for one method to be better than the other in terms of preprocessing steps (intensity/motion) and ICA step (application/reconstruction)?
>
Given that manual concatnation can induce nasty artefacts in terms of intensity shifts I'd recommend feeding all data into melodic as separate files. If these shifts exist then yes, melodic will behave differently in the two cases as it won't be tell these discontinuities from standard variation within a time series.
> 2. Considering that what I am doing is to feed fsl, n times (n=number of runs for one subject) and then doing concatenated temporal ICA on these n images. My question is whether or not the result would be different if I first concatenate all runs into one image and finally apply a single session ICA on it?
Yes, unless you carefully concatenate semi-preprocessed data in the way this would happen internally you will end up getting different results.
>
> 3. I suppose the motion correction step in preprocessing is just to do motion correction within the runs. But what happens when you have multiple runs of one subject? I mean if I decide to concatenate several runs of one subject what should I do? Do I need to do the intensity normalization, smoothing and sinc interpolation before? I did the simplest form of concatenation but when you move from the last volume of one run to the first volume of the other one there is a shift in intensity (or may be it is some movement).
>
Just don't concatenate data to avoid all of these issues.
> 4. My primary data was in minc format and I changed it to nii format. The z steps in mnc file were negative and the image had neurological convention. The nii images have positive steps and I am not sure if they have neurological or radiological convention. I suppose they are still in neurological convention since I have left-to-right in sform_orient. So before using mnc files to do slice timing correction my choice was regular down but I am a bit confused about nii files. Should it be regular up now considering the positive steps?
>
>
This all depends on how precisely you converted your data and if the Nifti file headers get written correctly, hard to tell from your description.
hth
Christian
>
> Cheers,
> Mona
|