Dear Dorian,
if you are interested in cluster level inference, the height threshold
is arbitrary and is just the feature-inducing threshold u. Once
specified, choose a cluster extent threshold k=0. From the results
table, you can figure out the minimal cluster size k0 needed to achieve
your desired FWE cluster level control. Go through results again by
specifying u and k0 as the height and cluster extent thresholds
respectively. That should give you the display you're looking for.
That's what Rik is doing on a MEG tutorial here:
http://imaging.mrc-cbu.cam.ac.uk/meg/SensorSpm
Note that SPM8 displays automatically the 0.05 critical threshold for
FWE cluster-level inference at the bottom left of the results table (FWEc).
You might also be interested in Tom's code:
http://www.sph.umich.edu/~nichols/JohnsGems5.html#Gem6
Best regards,
Guillaume.
Dorian P. wrote:
> Dear Christian,
>
> Thank you for your response. The VBM toolbox corrects the displayed
> p-values like the NS toolbox does, but it can not threshold for
> cluster level. That is, the clusters are present even under
> significance 0.05.
>
> To remove them I added some code to the spm_getSPM. All clusters not
> reaching significance are not present in the results any more (and the
> other clusters stay exactly the same).
>
> Hope I am not doing anything wrong by double thresholding voxels and
> remaining clusters.
>
> Dorian.
>
> 2009/5/26 Christian Gaser <[log in to unmask]>:
>> Hi Dorian,
>>
>> the first hack is implemented in the VBM5.1 toolbox. Use the option "Results with non-
>> stationary correction" and optionally skip the non-stationary option (which was thought
>> for use of VBM data):
>> http://dbm.neuro.uni-jena.de/vbm/download/
>>
>> Best regards,
>>
>> Christian
>>
>> --
>> ____________________________________________________________________________
>>
>> Christian Gaser, Ph.D.
>> Assistant Professor of Computational Neuroscience
>> Department of Psychiatry
>> Friedrich-Schiller-University of Jena
>> Jahnstrasse 3, D-07743 Jena, Germany
>> Tel: ++49-3641-934752 Fax: ++49-3641-934755
>> e-mail: [log in to unmask]
>> http://dbm.neuro.uni-jena.de
>>
>>
>> On Mon, 25 May 2009 12:49:05 +0200, Dorian P. <[log in to unmask]> wrote:
>>
>>> Hi SPM list,
>>>
>>> I need some help on two SPM hacks.
>>>
>>> 1.
>>> Does anybody have a hack to threshold both at voxel level (normal SPM)
>>> and at cluster level at the same time? (I.e. voxel threshold 0.0005 +
>>> cluster threshold 0.05)
>>>
>>> 2.
>>> And does anybody have a hack to use a ROI as mask? At present SPM can
>>> mask only with another contrast. There was a modified version of
>>> spm_getSPM by Jan Glascher, but when I looked inside I saw major
>>> differences from the actual function of SPM5.
>>> https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind0812&L=SPM&P=R43519
>>>
>>>
>>> Thank you for any help.
>>> Dorian
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|