Steffie Nelson wrote:
> Thanks Will. Your suggestions are very helpful. I have to ask though -
> is there a reason that the images can't be analyzed directly in xjview?
> I have a feeling that something is different, because often I get
> different answers thresholding the SPM.mat file in SPM
SPM will always threshold the t-image.
For height inference the MIP will only show those voxels above the
height threshold.
For cluster inference the MIP will only show (those voxels in) those
clusters containing more than X voxels (where X is the extent threshold).
vs. the t-image
> in xjview.
I'm not familiar with xjview. Does it have a capability to form clusters
of connected voxels ? And then only to show clusters containing more
than X voxels.
If so, it will use an algorithm for forming clusters, which may be
different to SPMs.
I'd be surprised however if you found differences using height
thresholds (rather than cluster thresholds).
>
> Steffie
>
>
> -----Original Message-----
> From: Will Penny [mailto:[log in to unmask]]
> Sent: Tuesday, March 02, 2010 11:47 AM
> To: Steffie Nelson
> Cc: [log in to unmask]
> Subject: Re: [SPM] 2nd Level Analysis Files
>
> Dear Steffie,
>
> SPM can do the thresholding for you.
>
> If you are doing a cluster size inference this will be on the
> t-statistic image.
>
> (so that only those voxels from the cluster inference appear in the MIP
> in SPM8 do (i) select a low height threshold eg. 0.001 uncorr, (ii) look
> at FWEc at the bottom of the resulting stats table, (iii) select the
> contrast again, select 0.001 uncorr for height again, and then FWEc for
> extent. SPM will then only display clusters containing more than FWEc
> voxels).
>
> You can then save the image of above threshold voxels (ie those voxels
> that appear in the MIP) by pressing the save button (in SPM's bottom
> left window). It will ask you for a name for this file.
>
> You can then load this into xjview.
>
> Best, Will.
>
> Steffie Nelson wrote:
>> Hello Imagers -
>>
>>
>>
>> I've just completed a 2^nd level analysis across 15 subjects using all
>
>> of the con_0001.img files for the contrast I'm targeting.
>>
>>
>>
>> In my resulting folder, I have several images:
>>
>>
>>
>> beta_0001.img
>>
>> ess_0001.img
>>
>> spmF_0001.img
>>
>> spmT_0002.img
>>
>>
>>
>> Which of these do I use for statistics? Meaning, on which image do I
>> now apply a threshold p-value and cluster size? I seem to get
> different
>> results in xjview depending on which image I use, which is why I ask.
>>
>>
>>
>> Thanks in advance,
>>
>>
>>
>> Steffie
>>
>>
>>
>>
>>
>
--
William D. Penny
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|