Hi
The posthoc GLM analysis (either in time or across subjects) really
only makes sense for activation data where you'd want to assume
consistent temporal response across subjects. In the case of resting
analysis you need to think hard if you do want to use this as a marker
of relative strength of an RSN - other makers (like subject-wise time
course variance r power in a certain band) might be more appropriate.
In any case I think you do not want to use full tensor ICA and instead
use time series concatenation and then maybe use the group derived
maps as templates.
hth
Christian
On 3 Oct 2008, at 14:01, Simon Robinson wrote:
> Hi,
>
> I have run the GICA again without design.mat and design.con files
> and don't
> find any significant differences in the output, other than what you
> would
> expect from running MELODIC twice - a minor reordering of components
> and
> that the melodic call logged in logs/gica referred to the design
> files in
> the 'with' case, as you would expect:
>
> /usr/local/fsl/bin/melodic -i .filelist -o groupmelodic.ica -v --nobet
> --bgthreshold=3 --tr=1.0 --report --guireport=../../report.html
> --bgimage=bg_image -
> -mmthresh=0.5 -a concat --Sdes=[path-to-design-mats]/design.mat
> --Scon=[path-to-design-mats]/design.con
>
> Is it correct to leave the two temporal design options blank? Do I
> need to
> select a particular output stats option? The FEAT version number is
> 3.05.
>
> The MELODIC manual promises ".. if a contrast matrix was specified,
> the
> table will also contain Z-statistics and p-values for all the
> contrasts.".
> Should these be (despite the absence of single-subject cons) the group
> comparisons that I am interested in and specified in the design
> matrices?
>
> Thanks in advance,
>
> Simon.
>
>
> On Thu, 2 Oct 2008 08:24:45 +0100, Steve Smith
> <[log in to unmask]> wrote:
>
>> Hi - that all sounds correct - having selected those design.mat
>> and .con files in the MELODIC GUI you should see extra information
>> using them appear in the web-page component reports - did you not get
>> that?
>>
>> Cheers.
>>
>>
>> On 1 Oct 2008, at 13:40, Simon Robinson wrote:
>>
>>> Hi,
>>>
>>> I would like to compare the level and distribution of activation in
>>> RSNs
>>> between patient and control groups. I have followed previous FSL
>>> archive
>>> threads on this topic, but it seems that those related to ICA
>>> performed on
>>> subject's data individually (with the single-session ICA option),
>>> mostly
>>> with pre v3 versions of MELODIC. If possible I would like to avoid
>>> the
>>> labour and risk of misidentification involved in a template-matching
>>> approach to identifying which component corresponds to which RSN in
>>> each
>>> subject (as in (Greicius, 2004) for the Default Mode), particularly
>>> as I
>>> would like to look at a number of RSNs.
>>>
>>> My question is whether it is possible to perform an intergroup
>>> comparison
>>> using one of MELODIC's group ICA approaches (either temporal
>>> concatenation
>>> or Tensor ICA). Some previous threads and manual entries suggest
>>> that it is,
>>> but I can't see how to do this in practice (or how it would work in
>>> the
>>> absence of back-projected single-subject responses).
>>>
>>> I was encourage by the MELODIC post-stats tab roll-over dialogue
>>>
>>> "If you are doing multi-session/multi-subjects ICA, you can also
>>> optionally
>>> select a subject "design.mat" model and "design.con" contrast
>>> file ... These
>>> will be used by MELODIC in ordering the ICA components, and in
>>> providing
>>> richer reporting information about the multiple sessions/subjects."
>>>
>>> and
>>>
>>> Steve Smith's answer Item #012747 "One of the new features in this
>>> version
>>> of MELODIC is the ability to put in temporal and/or subject models,
>>> which
>>> MELODIC then uses at the "post-ICA" stage to help you interpret the
>>> components. All you have to do is to create a FEAT-style model and
>>> contrast
>>> file and enter this into the MELODIC GUI, and it will use this for
>>> you."
>>>
>>> I ran a GICA for controls and patients using temporal concatenation
>>> and
>>> included design.mat and design.con files (defined using Glm) in the
>>> MELODIC
>>> post-stats GUI. My design looks like the following (as in
>>> Christian's post
>>> #011963 - reply to Peter Fried - only reduced from 15 controls and
>>> patients
>>> to 4 for brevity)
>>>
>>> group EV1 EV2
>>> controls patients Title EV1 EV2
>>> s1 1 1 0 | C1 controls mean 1 0
>>> s2 1 1 0 | C2 patients mean 0 1
>>> s3 1 1 0 | C3 C > P 1 -1
>>> s4 1 1 0 | C4 P > C -1 1
>>> s5 2 0 1 |
>>> s6 2 0 1 |
>>> s7 2 0 1 |
>>> s8 2 0 1 |
>>>
>>> MELODIC didn't seem to use them - at least nothing is registered in
>>> the log
>>> after the writing out of the groupmelodic.*. Can you tell me if it's
>>> possible to compare two groups in a GICA, and if so, how?
>>>
>>> Thanks in advance,
>>>
>>> Simon.
>>>
>>> --
>>> Simon Robinson, Ph.D.
>>> Postdoctoral Researcher
>>> Functional NeuroImaging Laboratory
>>> Center for Mind/Brain Sciences
>>> Via delle Regole 101 - 38060 Mattarello (TN)
>>> Italy
>>>
>>
>>
>> ---------------------------------------------------------------------------
>> 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
>> ---------------------------------------------------------------------------
_______________________________________________
Christian F. Beckmann, DPhil
Senior Lecturer, Clinical Neuroscience Department
Division of Neuroscience and Mental Health
Imperial College London
Hammersmith Hospital - London W12 0NN
Tel.: +44 (0) 208 383 3722 --- Fax: +44 (0) 208 383 2029
Email: [log in to unmask]
http://www.imperial.ac.uk/medicine/people/c.beckmann/
Senior Research Fellow, FMRIB Centre
University of Oxford
JR Hospital - Oxford OX3 9DU
Tel.: +44 (0) 1865 222551 --- Fax: +44 (0) 1865 222717
Email: [log in to unmask]
http://www.fmrib.ox.ac.uk/~beckmann
|