Dear Ankit,
it is very hard to know what might have gone wrong from your
description below.
One thing that comes to mind is
convolution: Gaussian
which will be very different from the HRF that SPM uses. So you may
want to change that to e.g. "double gamma HRF", which will be more
similar to what SPM does.
Another possibility is your choice of high-pass filter, which seems a
little short and might therefore potentially remove some of your
activations (depending on their frequency content).
If you really want to get to the bottom of the differnces I would
recommend you extract the resulting design matrices from FSL and SPM,
import them into Matlab (the SPM one will already be there I guess)
and compare the regressors that FSL has given you to those that SPM
gives. That way you'll be able to spot any differences in timing, HRF
shape, etc that might cause this difference.
Good Luck Jesper
> I have tried almost every possible option of the GUI to see if there
> is
> something i am messing up but nothing really seems to help. I am
> trying to
> run Feat for one of our event related study with several EV's. When
> i run
> the same experiment with the same set up in SPM, SPM does show me
> activations however FSL doesnt !!! I am really confused and any help
> will be
> greatly appreciated. I have posted my issues earlier and yet for
> some reason
> or the other i just dont get to see the activations.
>
> there are a total of 14 EV's in the design
> TR=1.5 sec,
> FWHM=9mm,
> 486 volumes, 6 volumes to be deleted,
> High pass filter = 30,
> I am using the option of full model set up and my EV's follow a 3
> column
> format,
> convolution : Gaussian
>
> I am trying to look at the uncorrected voxel p threshold 0.05; Z
> threshold =
> 1.64
>
> My data is 4d
> Original onset files were in terms of volumes so in order to get them
> converted into time in seconds i have multiplied each of the values
> in those
> files with TR.
>
> e.g one of the onset files in a three column format
>
> 78.1965 2 1
> 91.2405 2 1
> 104.284 2 1
> 130.338 2 1
> 143.365 2 1
> 182.46 2 1
> 208.53 2 1
> 404.025 2 1
> 547.365 2 1
> 612.525 2 1
> 703.74 2 1
>
> I already ran MELODIC on 15 subjects and also on just one subject to
> check
> the number of independent components and everytime i end up getting
> some 200
> to 240 components. I am not even sure if this is a realistic number
> for the
> components. Could it be that my data is messed up ?
>
> Thank you in advance
>
> sincerely,
> Ankit Master
>
|