kentman234 Wrote:
-------------------------------------------------------
> After I did ICA to the data of one subject, i got
> 25 components (ICs). Now i need to work on the
> time courses file which is of HDR format. This
> file contains the time courses o fthe 25
> components. How can afni read this file?
I assume you have not only HDR but als IMG files - the former is the header with meta-information of the data, the latter the data itself.
You can convert these with the 3dcopy (or the obsolete 3dANALYZEtoAFNI) program.
> Next, I need to choose a part of
> the ICs (15 for example) for training and to test
> the rest.
I don't understand the logic here. It seems that if you decompose your data in components that are 'independent' that means that data in one component have no (or little, if they are not truly independent) information on the other components, and thus I don't see how you would be able to classify anything.
in fMRI machine learning world it's common to use take-one-run out cross-validation (where a 'run' is a set of consecutive volumetric measuruments, typically between 5-10min). For training all but one run is used, and the remaining run is used for testing. This is repeated N times (if there are N runs), each time taking a different run.
If you combine cross-validation with other processing steps such as feature selection, please consider that it's important to make sure that you are not mixing up data used both for training and testing. For example, you might bias your results if you select informative features on the whole dataset and then run cross-validation. The correct approach is to run feature selection on the training set in each cross-validation fold. See [
miplab.unige.ch] for details.
Other than that I would suggest to start simple, maybe with a region of interest that you expect to be informative, or if you have no idea which area is informative, consider using a searchlight (information mapping).
Hope this helps.