If you have not done so already, you will really benefit from going
through the Feat practicals (http://www.fmrib.ox.ac.uk/fslcourse)
You want to establish if your data is alright, if the pre-processing
went alright, and if your EVs/model is set up alright.
The EV you are using is a boxcar with period of 32 secs is quite crude
- you may need to consider more subtle modelling using 3 column format
EVs (see the Feat prac)
Also you can look at Melodic results to see if there is any plausible
activation, what timings that activation might have, and if there are
any nasty artefacts in the data.
The best way to qualitatively assess the straightforward voxelwise GLM
stats is to go into the stats directory inside the feat directory, and
take a look at the zstat files in Fslview.
To look for negative activations you need an extra contrast that is
[-1] ; presumably the one you have at the moment is just .
Dr Mark Woolrich
EPSRC Advanced Research Fellow University Research Lecturer
Oxford University Centre for Functional MRI of the Brain (FMRIB),
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
Tel: (+44)1865-222782 Homepage: http://www.fmrib.ox.ac.uk/~woolrich
On 10 Aug 2009, at 03:12, Eric Zalusky wrote:
> I am new to FSL and functional analysis and would like to see if Iím
> up my analysis correctly in FEAT. Briefly the experiment uses two
> paradigms - each of which are scanned during two separate
> sequences. The
> first sequence alternates between thinking and passive activities.
> passive activity last 16 seconds and requires the subject to view 2
> rows of
> asterisks for 3 seconds, followed by a 7 second pause. Then the
> word yes or
> no appears and the subject clicks the correct hand clicker. Next
> there is a
> 6 second pause then a thinking activity occurs. The subject is then
> given a
> row of letters to memorize for 3 seconds. This is followed by a 7
> pause and then a single letter appears and the subject must click
> yes/no if
> the letter was in the previous row of letters. Then there is a 6
> pause before a new set appears.
> The second paradigm uses the same timing but a different task - this
> is a
> separate scan.
> When I use FEAT this is how I set up the analysis.
> select analyze data
> TR =2
> High pass = 100
> Volume = filled in automatically
> slice timing correction = interleaved
> nothing else changed
> Full Model Set Up:
> Number of original EVs = 1
> Basic Shape = square
> Skip = 0
> Off = 16
> On = 16
> Phase = 0
> Stop after = -1
> Convolution = Gamma
> Phase = 0
> Stddev = 3
> Mean lag = 6
> Yes to temporal derivative
> Yes to Apply temporal filtering
> Contrasts = 1
> F-tests = 0
> I change nothing. I use cluster thresholding, z=2.3 p=0.05
> I do not use the initial structural image
> For the main structural image I use a high res T1 after BET at 12 DOF.
> The standard space is the default MNI152 T1 2mm 12 DOF.
> Then I hit go and after the analysis is done, I go under Utils in
> FEAT and
> upsample+overlay using the FEAT directory that was created and I
> upsample to
> standard with the background image as the main structural.
> I do this the same for both paradigms that we scanned.
> When I go into FSL Viewer I am getting some weird activation. I
> tried to
> attach some pictures of what Iím seeing, but the files were too big,
> when I zipped them. So I'll try to describe what I see in the viewer
> After I open my background image and add the activation, the
> contrast/brightness is already set and there is significant activation
> outside of the brain. I can manually adjust this, but I can never
> seem to
> get the image the same as my rendered thresh image.
> My questions are:
> Am I setting up the analysis correctly in FEAT?
> Under the model set up, would I create more EVs, contrasts, or F-test?
> How do I correctly change the contrast and brightness to give me
> images that
> are the same to the rendered thresh images?
> Is there a way to see negative activation? I am able to see this on
> Siemens scanner. I would like to be able to measure both the
> positive and
> negative activation.
> Eric Zalusky