Hi,
I can speak only for myself: in my data (comprising around 15 subjects
with neurological conditions) the RSN analysis gave me about 40
components explaining a total of about 80% of the total variance. Lots
of the components are "structured noise", though (motion artefacts,
csf pulsation artefacts and so on). A temporal concat ICA of a motor
paradigm yielded ca 66% explained variance of the total variance
present in the data, with ca 25% going to components relating to
primary sensorimotor areas. The "default RSN" map in both analyses
(i.e. posterior cingulum, biparietal, anterior cingulum) made up only
3% of the total variance, in my studies!
Maybe it's the amount of structured vs. gaussian noise in the data
that matters here (lots of structured noise in [neurological]
patients, less in normal subjects) - perhaps an expert can comment on
that?
Cheers,
Cornelius
On Mon, Jan 4, 2010 at 9:57 PM, Michael Scheel
<[log in to unmask]> wrote:
> Hi fsl experts,
>
> I'm currently testing fsl melodic in resting state analysis and find it to work excellent. In the sample data that I analyzed the component that
> most looked liked the default mode network had 7.16 % of explained variance and 1.21 % of total variance.
> When I'm right this means that in total there is about 17% of variance explained with all the 14 components I have.
> Is 17% explained total variance rather small - so it seems to me. Are there any recommendations of how much variance the components should explain?
>
> Thanks, Michael
--
Dr. med. Cornelius J. Werner
Department of Neurology
RWTH Aachen University
Pauwelsstr. 30
52074 Aachen
Germany
Institute of Neuroscience and Medicine
MR Physics - INM4
Research Centre Juelich
52425 Juelich
Germany
::: Please encrypt confidential data :::
|