Thanks for your fast answer. Is it possible at all, to use a parametric test and get p-values out of the TCFE-image?

 

Oliver

 

---

Von: FSL - FMRIB's Software Library [mailto:[log in to unmask]] Im Auftrag von Reza Salimi
Gesendet: Mittwoch, 30. Juni 2010 13:58

An: [log in to unmask]

Betreff: Re: [FSL] TFCE with SPM-Toolbox

 

Oliver,

at the moment, TFCE can be utilized for the second-level type analyses where you have a exchangeable design matrix (under the H0). This exchangeability makes the formation of distribution of the statistic of interest (i.e., maximum TFCE per volume, as recommended) under H0. Having this null distribution, you will be able to assess the actual TFCE-image's significance at each voxel, i.e., a nonparametric hypothesis test.

 

Now, in order to use fslmaths jointly with some SPM nonparametric tests (which I think might be SnPM), you might need to add a T->TFCE transformation (using fslmaths) whenever you form your T-stat and use the null distribution of the max statistic of these new images for converting you actual t-stat into a P-image.

 

Anyway, to summarise, it might be much easier to use FSL-randomise, however, you can use any nonparametric test on max{T-images}, *IF* you correctly transform your T-images to TFCE-images. 

 

2010/6/30 Grimm, Oliver <[log in to unmask]>

Dear FSL-Team, dear Stephen,

we have read your paper about TFCE with great interest and we would like to make use of this cool technique.

As we do net (yet...) use FSL, but have a very well established MATLAB/SPM-pipeline, we would like to compare our old results with new TFCE-statistics.
We wonder wether it would be possible to combine TFCE (via fslmaths-command) with SPM, e.g. like this ...

a)       Pre-processing and the first-level-analyses have been performed in SPM

b)       the first-level SPM-t-images will be fed into the fsl/TFCE procedure

c)       the resulting TFCE images should be fed into second-level analysis in SPM



However, we do not really know if this strategy is adequate and it would be great if you, as the expert, can help us with the following points:



1)       Would you recommend to use the SPM t-images for the TFCE procedure or should we take z-images or something else?

both T and Z should be fine 

2)    Do we have to feed the second-level-analysis with the TFCE-weighted images from the first-level or is TFCE applied afterwards on a second-level result image?

currently, the existing nonparametric TFCE scheme is ONLY applicable to the second-level type analyses. 

We know that you are getting plenty of these emails and your time is limited, however, maybe we will be lucky and get some brief helping comments.



All the best

Michael Plichta

Oliver Grimm



_____________________________________

Department of Psychiatry

Division for Imaging in Psychiatry

Central Institute of Mental Health






-----Ursprüngliche Nachricht-----
Von: FSL - FMRIB's Software Library [mailto:[log in to unmask]] Im Auftrag von Andreas Bartsch
Gesendet: Mittwoch, 30. Juni 2010 12:05
An: [log in to unmask]
Betreff: [FSL] AW: [FSL] corpus callosum waytotal

no - you need to add the waytotal, given that the fdt_path is the sum of the tracking forth and back in the symm mode.
spatial extent of the tract is NOT modified by the normalization but only by later thresholding. depending on your masks, the numbers can seem quite low.
cheers-
andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von LiuYan [[log in to unmask]]
Gesendet: Mittwoch, 30. Juni 2010 12:04
An: [log in to unmask]
Betreff: Re: [FSL] corpus callosum waytotal

Hi Stam:

Thank you for your advice! I also thought of that, to use the sum for the normalization. But the waytotal number of the corpus callosum is big, if I add these two numbers and divide the fdt_path by the sum, I will have a very low probability for the entire tract. After using the threshold, I get very few fibres. Could I use the average of these two numbers, as it is the same tract going through the two seeds twice?

Regards,

Yan
________________________________
Date: Wed, 30 Jun 2010 10:21:43 +0100
From: [log in to unmask]
Subject: Re: [FSL] corpus callosum waytotal
To: [log in to unmask]

Hi Yan,

you are right, you will get two waytotal numbers,  one per specified seed mask. I would add these numbers and use the sum for normalization.

Cheers,
Stam



On 30 Jun 2010, at 09:50, LiuYan wrote:

Hi FSL specialists:

Recently, we try to do the tractography for the corpus callosum. Two seed masks are defined either side of the midline in the sagittal images. In the tools box, we choose the multiple masks to get the tract. However, we have two waytotal numbers in the result, and we don't know which one to use for the normalization. Any one who has also worked on the corpus callosum could give us some advice? Many thanks in advance!

Regards,

Yan



________________________________
使用新一代 Windows Live Messenger 轻松交流和共享！ 立刻下载！<http://www.windowslive.cn/messenger/>


________________________________
使用新一代 Windows Live Messenger 轻松交流和共享！ 立刻下载！<http://www.windowslive.cn/messenger/>

--
Reza Salimi-Khorshidi,
DPhil Candidate, Dept. of Clinical Neurology, University of Oxford (Linacre College).
[log in to unmask]    
FMRIB Centre, John Radcliffe Hospital, Oxford OX3 9DU
Tel: +44 (0) 1865 222704  Fax: +44 (0)1865 222717