So what conclusions can be made from a group comparison on fractional
anisotropy (FA)? Might you expect differences between normals and multiple
sclerosis patients, due to differences in white matter density?
----- Original Message -----
From: "Marenco, Stefano (NIMH)" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, November 01, 2002 7:30 AM
Subject: Re: FEAToutput, ANALYZE hdr and HRF parameters.
> stand for fractional anisotropy, a measure commonly used in Diffusion
tensor
> imaging (DTI). Stefano
>
> -----Original Message-----
> From: Darren Weber [mailto:[log in to unmask]]
> Sent: Thursday, October 31, 2002 3:55 PM
> To: [log in to unmask]
> Subject: Re: [FSL] FEAToutput, ANALYZE hdr and HRF parameters.
>
>
> What is "FA" data?
>
> ----- Original Message -----
> From: "Erik-Jan Vlieger" <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Friday, November 01, 2002 1:00 AM
> Subject: Re: FEAToutput, ANALYZE hdr and HRF parameters.
>
>
> > > > Hi - yes, you can easily turn the grand-mean scaling off by editing
> > > > fsl/tcl/feat.tcl and commenting-out/deleting the line
> > > > set thecommand "$thecommand -I $global_mean"
> > > > I'm not sure why you would want to turn this off though?
>
> > > We are analyzing DTI images with Feat (I know what to do with the
design
> > > matrix and the issues on pre-whitening to make it valid), and I would
> like
> > > to see the original FA values.
>
> > I would be interested in doing the same. Could you send me a set of
> > instructions on how to analyze FA data?
>
> Certainly. I would love to get some comments on this design!
>
> In the following design, you look for areas for which patients have higher
> gray-values than controls. If you want to look for areas with lower
> gray-values, in step (2), put the controls at the front, the patients at
the
> end, and correct for this when specifying the model.
>
>
> 1) Perform spatial normalizing on all the FA scans of alle the subjects (I
> use
> flirt for this, with the MNI templates. First I BET a 3D T1W scan, then
> register the B0 image to the betted 3D scan. The betted 3D scan is
> registered
> to the template. After that the combined transformation is applied to the
FA
> image).
> 2) Merge these normalized scans into a 4D data-set (avwmerge), e.g. 30
> controls and 32 patients. Put the controls at the beginning.
> 3) Use Feat for the analysis:
> Data: Set the TR to 1.0
> Pre-Stats: No slice time correction
> No motion correction
> No BET
> Quite some smoothing (I would not know an optimum)
> No intensity normalization (FA are absolute values!)
> No temporal filtering
> Stats: No film prewhitening
> Full Model setup:
> Skip : 0
> Off: 30 (controls)
> On: 32 (patients)
> Convolution: none
> NO temporal filtering
> NO temporal derivative
> Post-stats: whatever you want
> Registration: turn it all off (you already did this).
>
> GO!
>
> Erik-Jan
>
|