On 17 Dec 2009, at 18:16, Kundu, Prantik (NIH/NIMH) [F] wrote:
Hi all
I have fMRI data at short TRs (1 sec) in which some timecourses for spatial components, ie. DMN, are 'picking up' the fluctuations due to respiration occuring at about 0.3 Hz.
I'm surprised about this as normally ICA does a very good job of separating RSNs from other signals such as non-neural physiological effects.
Have you tried increasing the ICA dimensionality? How are you judging that the DMN component is including the respiration effects?
It's not currently possible to enter temporal information _into_ the ICA process, but if you have confound regressors then you could pre-regress those out of the data after running the other pre-processing, and before running melodic. But we don't normally need to do this because of the above.
Cheers.
Can I provide my physiological traces in some way so that MELODIC will insert the data into the mixing matrix so that decomposition can be aided with this information? I distinguish this from using the traces as post-decomposition regressors for fitting.
Thanks,
Prantik