This workflow implements feat FSL individual analyses as performed in the experiment detailed in (Soleman et al, "Large scale fMRI parameter study on a production grid", MICCAI-G'08).

An example of inputs (in VBrowser XML dialect) can be found here. In this document, files are represented by their LFNs in vlemed's LFC.

Processor "feat":

1. generates the experiment input (design) file from the parameters

2. calls feat FSL.

Processor "convertString" tweaks the output file names so that it fits the user's experiment organization.

Below is detailed the meaning of the input parameters.

Files remaining constant among the iterations (thus iterated with cross prdct)

* feat_description: contains a URL to the command-line descriptor used to generate grid jobs. The descriptor contains an LFN-link (in vlemed's LFC) to the executable to run. The executable is a bash script tweaking input files before calling FSL feat.

* designFileTemplate: file containing constants for the analysis. Used as a template to fill in varying parameters.

Files associated to a particular patient  (thus linked together with dot prdct)

* feat_files_1: the fMRI 4D scan (EPI) in .nii.gz format.

* PAR file: parameter file from the scanner, used to grab theTR

 * fmri_custom{1,2,3}: stimuli files associated to the fMRI experiment.

 * highres_file: the T1 scan of the patient in .nii.gz format.

Parameters to sweep on (iterated with cross product)

* fmri_regstandard_dof: degrees of freedom for EPI-to-std-brain registration

* fmri_regstandard_dof: degrees of freedom for EPI-to-T1 registration

* fmri_convolve_phase: HRF delay. Note that it's used 3 times in feat, once for each stimulus. Those 3 inputs are combined with a dot.

* fmri_smooth: size of the smoothing kernel