#!/bin/tcsh -f # In preparation for doing mini- and full-probabilistic tractography, # the uncertainty of relevant tracking parameters (GFA and directions) # need to be estimated. In FATCAT, this is done using jackknife # resampling over a number of iterations. # *HOWEVER*, since there is currently no HARDI modeler in AFNI, # there's also no 'official' uncertainty estimator program. I can # imagine that in models with ODFs, one could do good approximations # of voxelwise uncertainty per direction, at least. # For the time being, the current method described here is a # patch/*very* rough estimate. For now, it's a constant cone of # uncertainty for each possible directional vector, and a constant # uncertainty for GFA. Perhaps a better model could be based on the # DTI uncertainties, but for the moment, this is one simple idea. # Other models calculated in other programs are certainly welcome; I # probably will try to do something with ODFs at some # point. Discussion (or better, contribution!) is welcome. # Not much else to be said about this. Just the script. # We need a four-brik uncertainty set: [0] for GFA, and [1..3] for the # three directional vectors of this HARDI set. # logic being: in DTI, FA is thresholded at 0.2 and has minimum uncert # of 0.015, so going for similar ratio here based on GQI-GFA threshold # being used. (0.04/0.2)*0.015~0.003. 3dcalc -a mask_DWI+orig -expr 'a*(0.04/0.2)*0.015' \ -prefix TEMP_GQI_01.nii.gz \ -overwrite # and then a bit of increasing uncertainty for each of the directions: # 3deg, 4deg, 5deg. 3dcalc -a mask_DWI+orig -expr 'a*3*(3.14159/180)' \ -prefix TEMP_GQI_02.nii.gz \ -overwrite 3dcalc -a mask_DWI+orig -expr 'a*4*(3.14159/180)' \ -prefix TEMP_GQI_03.nii.gz \ -overwrite 3dcalc -a mask_DWI+orig -expr 'a*5*(3.14159/180)' \ -prefix TEMP_GQI_04.nii.gz \ -overwrite 3dTcat \ -prefix HARDI/Simple_Uncert.nii.gz \ TEMP_GQI_??.nii.gz \ -overwrite # get rid of the evidence... \rm -f TEMP_GQI_??.nii.gz