Hi,
nope, when using spatial ICA (as implemented in melodic) the time
courses are not restricted to be orthogonal. The spatial maps,
however, are not only orthogonal but are being optimised to be as
statistically independent as possible. This is different from a PCA
decomposition where both the time courses and the spatial maps are
orthogonal and temporal ICA, where the time courses but not the
spatial maps are orthogonal.
The fact that time courses from spatial ICA are not restricted to be
orthogonal presents a major advantage over PCA and temporal ICA as
this makes it possible to extract effects like stimulus-related maps
and separate these from temporally correlated effects like stimulus-
correlated motion.
hope this helps
christian
On 2 Mar 2007, at 19:58, Xuelin Cui wrote:
> hi :
>
> I ran Melodic to fMRI data, which gave me a bunch of independent
> components(ICs). I thought these ICs should be orthogonal to each
> other since they are the eigenvectors of the covariance matrix. But
> when I applied dot product to these ICs, I found the resaults are
> not actually 0, which means these ICs are not orthogonal to each
> other. I couldn't understand this. Are they supposeed to be
> orthogonal? How do we get these ICs?
>
> Thanks a lot
>
> Xuelin
>
> ****************************************
> Xuelin Cui
> Department of Electrical Engineering
> University of Hawaii-Manoa
> Honolulu HI 96822
>
> Tel: 1-808-349-0983
> Email: [log in to unmask]
> ****************************************
--
Christian F. Beckmann
Oxford University Centre for Functional
Magnetic Resonance Imaging of the Brain,
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
Email: [log in to unmask] - http://www.fmrib.ox.ac.uk/~beckmann/
Phone: +44(0)1865 222551 Fax: +44(0)1865 222717
|