Hi,
On 14 Jun 2007, at 17:23, Antonios - Constantine wrote:
> Dear Steve-fsl users
>
> Indeed the use of: ''fslview flow_all_subjects sienar_tstat1'' was
> really
> very helpful.
>
> I re run as you did the same data with -c 2.5 and i also saw with
> maxc_tstat2 these two areas of significance (around ventricles and the
> middle temporal gyrus), even though not with significance of 0.94
> as you
> did, but instead with 0.73 (maxc_tstat1 doesn't give me any info
> even with a
> a range of [0.1:1] ). Do you have any idea why does this happen?
> And with a
> significance of 0.73 (p<0.27) i can't consider these results
> important? Right?
It's correct that nothing shows up for your tstat1. I'm not sure why
tstat2 is giving a different answer to what I got. I was running our
internal development version of randomise so there's possibly some
difference - we'll check into that. The new release will be out in a
couple of months hopefully.
> You also mentioned that by changing the smoothing in siena_flow2std
> script
> the results will change.But then how can we talk about objective
> results
> since the importance of our data depends on the smoothing filtering?
Good question - and one that doesn't have a "right" answer in general
in image processing I'm afraid. The extent of data smoothing, and in
the case of cluster-based analysis, the cluster-forming threshold (-c
option) are both arbitrary and there's no "right" answer. For the
smoothing, you might set it according to the expected spatial size of
the effect you're looking for (i.e. matched filter theorem), but of
course that doesn't help much if you have no particular expectation
for the size of the effect. Also, if you have spatial variability
(across subjects in this case) of the location of a focal effect, you
might set the smoothing to blur the data enough to get some overlap
of the effect across subjects - as you correctly say below.
> By
> changing the smoothing we're trying to correct the variations of
> the groups
> (parkinsonians and controls ) right? so theoretically the “accurate
> value”
> of smoothing is the one that will correct as much as possible these
> variations..But how do we know how close to the “accurate smoothing
> value”
> is the value that we choose?
>
> Finally correct me if i'm wrong, to sumarise we have 3 steps of
> assessing
> the data after using radomise:
> a)locate the area of importance with maxc_tstat1&2
> b)detect with tstat1(using both negative and positive values in
> display
> range) which group is bigger
> c)depict tstat3&4 (using both negative and positive values in
> display range)
> and in the importance areas we check whether tstat3&4 are positive
> (which
> means growth in group A and group B respectively) or negative
> (atrophy in
> group A and group B respectively)
Indeed.
Cheers, Steve.
>
> Thanks a lot once more.
>
> Antonios-Constantine Thanellas
------------------------------------------------------------------------
---
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
------------------------------------------------------------------------
---
|