Hi Gang,

First, you need some method to generate 5000 iterations of your data,
randomized by some method like the ones in Nichols & Holmes primer paper:

http://www.fil.ion.ucl.ac.uk/spm/doc/papers/NicholsHolmes.pdf
The "Primer Paper",
TE Nichols and APHolmes.
Nonparametric Permutation Tests for Functional Neuroimaging: A
Primer with Examples.
Human Brain Mapping, 15:1-25, 2002.

FSL's randomise does this for volumes, and caret_stats supports some
simple models on the surface:

ANOVA One-Way                                        
-inferential-anova-one-way
ANOVA One-Way Coordinate Difference                  
-inferential-anova-one-way-coordinate-difference
Data Smoothing                                        -data-smoothing
Data Transform - Threshold Free Cluster Enhancement   -data-transform-tfce
Descriptive Statistics                                -descriptive
Development - Metric Monotonic Test                  
-development-monotonic-test
Development coordinate metric combination            
-development-metric-coord
File Information                                      -file-information
Interhemispheric                                     
-inferential-interhemispheric
Signficance - Cluster-based Thresholding             
-significance-cluster-threshold
Signficance - Threshold Free                         
-significance-threshold-free
T-Test One-Sample                                    
-inferential-t-test-one-sample
T-Test Paired                                        
-inferential-t-test-paired
T-Test Two-Sample                                    
-inferential-t-test-two-sample

Then you need to smooth the resulting real and random t-/f-maps a bit,
before feeding it to caret_stats -significance-threshold-free, which
builds a distribution of max tfce values and thresholds it at 0.025, or
whatever alpha you specify.  It generates a report and label file -- nice.

We never released caret_stats formally, though I've provided to a few
people off-list.

The hard part is usually getting the 5000 or so random iterations of your
data, permuted/flipped in the manner described in the primer paper -- what
FSL's randomise does for volumes.

FSL is working on supporting both surface and volume data, but it will be
a while before randomise can easily read surfaces, I think.

If Freesurfer can generate these randomised statistics, then these could
be input into caret_stats -significance-threshold-free pretty easily.

Donna

> Hi,
>
> I have some cortical surfaces aligned in the spherical space in FreeSurfer
> format. I would like to know how can I use the TFCE method in Caret to do
> statistical analysis of the attributes on the cortical surface such as the
> sulcal depth, surface area or 3D vertex coordinate. Thanks a lot.
>
> Gang
> _______________________________________________
> caret-users mailing list
> [email protected]
> http://brainvis.wustl.edu/mailman/listinfo/caret-users
>

_______________________________________________
caret-users mailing list
[email protected]
http://brainvis.wustl.edu/mailman/listinfo/caret-users

Reply via email to