Hi,
Thank you very much for this detailed explanation. I am looking forward to
seeing TFCE option in FEAT.
Cheers,
Auer, Tibor M.D.
Biomedizinische NMR Forschungs GmbH
am Max-Planck Institut für Biophysikalische Chemie
Am Fassberg 11
37077 Göttingen
Germany
Phone/Work: +49-(0)551-201-1725
Phone/Home: +49-(0)551-250-4576
Mobile: +49-(0)176-8012-7921
Mail: [log in to unmask]
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Steve Smith
Sent: Monday, March 02, 2009 12:10 PM
To: [log in to unmask]
Subject: Re: [FSL] TFCE in fMRI
Hi - the TFCE option in fslmaths is there in case you are interested
in the TFCE enhancement (e.g.) to a raw zstat image - to enhance
cluster-like aspects of the data. However, as Reza said, it is not
able to convert TFCE-enhanced values into p-values, because there is
not currently a simple direct conversion - currently, the only way to
convert TFCE scores into p-values is via randomise.
You can't apply randomise to FMRI first-level timeseries analysis
because of the temporal smoothness in the data - you could maybe
whiten the data and then run randomise (a la Bullmore) but we've not
tested that. You can apply randomise to higher-level analyses (see
previous posts) - in general this is usable for most if not all second-
level designs, though currently only allows OLS modelling, i.e.
doesn't apply the more advanced variance modelling that FLAME does.
Cheers.
On 2 Mar 2009, at 10:48, Auer, Tibor MD. wrote:
> Hi Reza,
>
> Thank you for your fast reply.
>
> It was my fault. On the page
(http://www.fmrib.ox.co.uk/fsl/randomize/index.html
> ) I have read that fslmaths can be used to test TFCE "on an
> existing stats image". But it seems it is true just after applying
> randomise. Is it correct? Or what else "fslmaths -tfce" is for?
>
> Another questions. According to the page mentioned above "randomise"
> is an option for simple GLM (t-tests, one-way ANOVA, simple
> correlation). In fMRI it means some cases of higher level analysis.
> What about 2-way ANOVAs? What about 1st level, where design matrix
> contains canonical HRF? And when we use nuisance effects (e.g.
> motion-correction parameters, or temporal derivative of HRF), the
> exchangeability of data is not assured. Does it means that randomise
> is not advised to use in these cases?
>
> Best,
>
> Auer, Tibor M.D.
> Biomedizinische NMR Forschungs GmbH
> am Max-Planck Institut für Biophysikalische Chemie
> Am Fassberg 11
> 37077 Göttingen
> Germany
> Phone/Work: +49-(0)551-201-1725
> Phone/Home: +49-(0)551-250-4576
> Mobile: +49-(0)176-8012-7921
> Mail: [log in to unmask]
>
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf Of Reza Salimi
> Sent: Monday, March 02, 2009 11:01 AM
> To: [log in to unmask]
> Subject: Re: [FSL] TFCE in fMRI
>
> Hi Tibur,
> applying the TFCE to a statistical map (i.e., a t- or z-stat map)
> cannot be directly used for a statistical inference (i.e.,
> thresholding to find activation blobs in the brain),
> and since there is no parametric form for the TFCE statistic, you
> need to run a nonparametric (randomise) analysis, with TFCE
> inference and threshold the output p-value volume (which is
> *tfce*corrp* volume in randomise).
> cheers
>
>
>
>
> 2009/3/2 Auer, Tibor MD. <[log in to unmask]>
> Dear FSL Users,
>
> I try to use TFCE (Threshold Free Cluster Enhancement) in fMRI. I
> have read that "fslmath" can apply TFCE on a previously calculated
> statistical map (e.g. ztats1.nii.gz).
>
> To make everything easier, I have written a script to produce
> activation map after TFCE (see attached).
> I used the pre-fixed options 2 0.5, and according to my
> understanding thresholded the resulted image with 0.95 to get p=0.05
> equivalent. The rest of the script is just to overlay and produce
> image.
>
> The attached script produces more sensitive however quite noisy image.
> I ask someone with more knowledge on TFCE to check whether my
> implementation is correct.
>
> Thanks,
>
> Auer, Tibor M.D.
>
> Biomedizinische NMR Forschungs GmbH
>
> am Max-Planck Institut für Biophysikalische Chemie
>
> Am Fassberg 11
>
> 37077 Göttingen
>
> Germany
>
> Phone/Work: +49-(0)551-201-1725
>
> Phone/Home: +49-(0)551-250-4576
>
> Mobile: +49-(0)176-8012-7921
>
> Mail: [log in to unmask]
>
>
>
>
>
>
>
>
> --
> G. Salimi-Khorshidi,
> D.Phil. Student, Dept. of Clinical Neurology, University of Oxford.
> [log in to unmask] http://www.fmrib.ox.ac.uk/~reza
> FMRIB Centre, John Radcliffe Hospital,
> Headington, Oxford, OX3 9DU
> Tel: +44 (0) 1865 222466 Fax: +44 (0)1865 222717
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|