Hi
>
> 1. In the Melodic GUI under the Post-Stats tab, there is an option to set the threshold for IC maps. As the calculation of IC took quite a while on my PC, may I know how I can reuse previous results but with a different threshold and to generate a new report?
>
Look at the usage help (2nd option of running melodic:
melodic -i <filename> --ICs=melodic_IC --mix=melodic_mix <options>
to run Mixture Model based inference on estimated ICs
Set the other options to reflect what you want (e.g. --report --Ostats if you want a html report and all stats outputs)
> 2. I have seen in this mailing list and also the av example in the FSL course (e.g. http://fsl.fmrib.ox.ac.uk/fslcourse/lectures/practicals/melodic/index.html) where there are multiple (and small) thresholded spatial maps per row in the report. In my case the thresholded spatial maps are stacked vertically (essentially one big image per row) so I need to scroll up and down to view all of the slices. This makes it very difficult to have a global view of the behaviours. Is there a setting I need to change so my report can look the same?
>
That depends on your data size, by default melodic creates images that are approx. 950 pixels wide, uses axials and scales by a factor of 2 (-s 2 -A 950 options in slicer). You can control this using the --report-maps=" <other options>" command line option to melodic, e.g. use --report_maps=" -a " or --report_maps=" -S 5 2000 " (see slicer to learn about how to control figures)
> 3. My final question concerns the interpretation of the IC time courses. In one post in the archive someone commented about reading off the time of the stimulus from the time course curve. Do I take it the (a) only the time courses which are periodic are meaningful for on/off tasks, i.e. whereby there's a cue for subject to perform a task or to relax, and
With good data and robust effects under a blocked task yes
> (b) the period of the time course corresponds to the period of the on/off cycle? So far the time courses have been anything but obviously periodic.
>
yes, though for this you need to set the --tr setting correctly. If you don;t find anything nicely periodic then that possibly points to a problem, e.g. not enough signal, poor data, other artefacts...
hth
Christian
> Thanks very much!
>
> Cheers,
> Louis.
|