Skip to main content Skip to navigation

FSL Scripts

These are random bash shell scripts for FSL that I find useful.

    Computes mean value in each ROI, where ROI are defined by an integer valued atlas. Can flexibly specify ROI as a single or multiple integer values. Example usage, where values 3 and 4 in the ROI atlas define the left and right hemispheres: ROItemplate "3:L_Brain 4:R_Brain" ImgList.csv Result.csv
    or, can merge multiple values into a single ROI ROItemplate "3+4:Brain" ImgList.csv Result.csv
    Creates a set of dummy variables from a categorical variable; requires R.
  • | Github:
    Creates a version of tsdiffana or DVARS, the standard deviation of temporal difference images. See my notes on Standardizing DVARS; on the original DVARS, see Matthew Brett's Data Diagnostics and Power et al. (2012) NeuroImage, 59(3), 2142–54.
    Fix as of 2017/02/19 - Was computing DVARS with standard deviation, instead of root mean square by original definition; fix has at most a small impact, increasing some DVARS values. Thanks to Chris Gorgolewski for pointing this out.
    Based on the FSL script easythresh, this script takes two statistic images and does a conjunction analysis, testing the 'Conjunction Null Hypothesis. The interpretation of significant regions is that there is evidence of effects in both contrasts tested (not just either/or). There is no assumption of independence required between the two effects tested. For more see my conjunction research page.
    Based on a zstat, tstat or randomise P-value image creates an image of 1 minus FDR-corrected P-values; optionally creates a rendering (colored blobs on a specified background image). For example, in a Feat results directory, running stats/zstat1 mask stats/zstat1_fdrcorrp 
    will produce a 1-PFDR image called zstat1_fdrcorrp in the stats directory. To additionally create an rendered image, use the -rend option, like -rend example_func rendered_thresh_zstat1_fdrcorrp \
           stats/zstat1 mask stats/zstat1_fdrcorrp    
    For 1-P images from randomise, use the -1mp option, as in: -1mp results_vox_p_tstat1 mask results_vox_corrp_tstat1    
    Finally, if you've got SPM T images, and $dof is the degrees-of-freedom, you can use -Tdf $dof spmT_0001 0 spmT_0001_Pfdr 
    where setting the mask name to "0" has the effect of using an implicit mask (<>0 means in the brain).
    For 1 or more files it prints the maximum value in the image. Very useful when used with randomise, as in *corrp*nii*    
    which will show you the best 1-minus-P for each corrected P-value image.
    Allows you to use of fslstats tool with multiple files (generalization of
    Allows you to use of fslinfo tool with multiple files, creating a tabular report for the multiple files (optionally a CSV report).
  • output
    Plots the average power spectrum of a 4D volume; specifically the power spectrum is computed voxel-wise and then averaged over the brain. Has options to detrend and high-pass to examine the effects of those operations.
  • PlotFeatMFX.R output
    This pair of companion scripts produces a plot that visualizes a one-sample mixed effects result at a given voxel. Showing the effect magnitude for each subject and the contribution of intra- and -inter-subject variance, it explains how and why Feat FLAME mixed effects results can vary from OLS. Only is directly called by the user, like Nback.gfeat 23 37 57     
    or -usemm Nback.gfeat 44 -52 42    
    but of course requires R.