Dear Christian
--------Sorry for harassing you but we are not sure what mean by this
"a simple FEAT run the same input data using the same design to get a Z-stats image associated with the t-stat image. Tis z-stat image is then thresholded using the mask and rendered on top a background image (the MNI template if I remember correctly). You can then investigate the Z-stats image explicitly in fslview (i.e. move your cursor to the desired location and check the z value."
------Does this mean now we run 2 higher level FEATS? 1 for each grp? and inputs are the orginal resting.nii.gz for each subject and then when it comes to masking is it prethreshold masking with the AN_mask.nii.gz we have just created ? for grp 1?
Thanks so much
Ciara
Dr Ciara McCabe
Neuroscience Building
Dept of Psychiatry
Warneford Hospital
Oxford University
Oxford
OX3 7JX
Tel: 01865 223778
Mob: 07966035631
http://www.psychiatry.ox.ac.uk/research/researchunits/perl/perlmembers/ciaramccabe/
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Christian F. Beckmann
Sent: 16 August 2011 22:31
To: [log in to unmask]
Subject: Re: [FSL] Multi-session multi-subject group ICA
Hi
Unless the temporal response profiles are expected to be the same you should use the concat approach. In this case including the 4/92 subjects with only 2 sessions should be fine.
hth
Christian
On 16 Aug 2011, at 21:45, Edward H Patzelt wrote:
> FSL -
>
> I want to do Melodic Group ICA. I have a study with 3 groups (schizophrenic, control, relatvive). There are 3 epi session-blocks per subject (N = 92) that are randomized by task block (N=3); each task block has sequential trial order.
>
> 1. Do I want to do Tensor ICA by temporally concatenating the sessions within each subject such that the 3 blocks A,B,C are concatenated in the same order for each subject, and my input will be 1 per subject N=92? It seems from previous posts that this may not be the best approach; instead I should put each session-block as an individual input and use multi-session temporal concatenation?
>
> 2. I have 4/92 subjects that only have 2/3 sessions. Can these subjects be included? Or must all subjects have the same number of blocks for the analysis?
|