# mritotal.default.cfg # # This is the default protocol file for mritotal, which specifies all # the data-specific preprocessing steps to take. Note that you don't # have any control over the order of the preprocessing steps -- you # can only specify what is done, and how it is done. The order of # steps taken is as follows: # # * reduce data (subsample and crop) # * zero-pad data (necessary to ensure that blurring is correct) # * blur data # # Default protocol by Greg Ward 95/08/23 - this instructs mritotal to # use some simple heuristics to subsample and crop the data, and # should work with T1-weighted MRI data of *normal* human brains where # the top of the head is pretty close to the top of the scanning # volume. You can also explicitly set the subsampling and cropping # parameters if you know that your scanning protocol is quite # consistent and would prefer not to have mritotal take a guess every # time it runs; see mritotal.icbm.cfg for an example of how this is # done for the ICBM protocol. # The data is first subsampled, mainly to save memory and time. There # are four possible actions here: # * -nosubsample: mritotal will not subsample your data # * -subsample xstep ystep zstep: you specify separate step sizes # (sampling frequencies) for each of the three spatial dimensions # * -isosubsample step: you specify a single step for all three # dimensions # * -guess_subsample: mritotal will compute the step sizes based # on your data. The step sizes for the three dimensions # are calculated independently; if your input step is < 1.5mm, # then the subsampling step will be twice your input step; # otherwise, that dimension is not subsampled. -guess_subsample # At the same time that the data is subsampled, it is cropped. This # is more important than subsampling, as the fit may fail if there is # too much data (eg. if the scanning volume extends down into the # neck, as is often the case with coronal or sagittal acquisitions). # You have the same four options for cropping as for subsampling # (-nocrop, -crop, -isocrop, and -guess_crop), except that "-crop" and # "-isocrop" take rather more complicated arguments. # # In particular, they expect pairs of numbers (-crop expects three, # -isocrop only one) that specify the amount to chop off each end of # an axis. (Actually, these numbers specify how much to *extend* the # axis by, so you must give negative numbers to chop data off.) The # first number specifies how much to chop off the low end of the axis, # and the second specifies how much to chop off the high end. Note # that these are "low" and "high" in the sense implied by the MINC # standard, ie. the low end of the x axis is the patient's left, low y # is patient posterior, and low z is patient inferior -- independent # of the order or orientation of your data. Finally, the numbers can # be specified in voxels, millimetres, or as a percentage of the # original dimension extent. # # In this default protocol, we use mritotal's simple heuristic for # cropping, which is: leave the x and y dimensions alone, but chop off # anything more than 190mm below the top of the volume. As long as # the top of the subject's head is pretty close to the top of the # volume, this does a reasonable job of chopping out excess data # (ie. below the bottom of the cerebellum). # # As a slightly fancier example, the crop specification for the ICBM # protocol is "-crop 0,0 0,0 -25%,0". This illustrates how to chop off # 25% of the data at the bottom of the scan (low z) while leaving the # x and y axes alone. (Since ICBM scans have a z extent of 256 mm, # this has the effect of removing all but the top 192mm of the scan, # which is almost identical to mritotal's cropping heuristic.) You # can also specify crop amounts as mm (millimetres) or v (voxels); # eg. to remove 10 voxels at both ends of the x axis and 20mm at the # top of the head (high z), you would use "-crop -10v,-10v 0,0 # 0,-20mm". You can mix and match the three units at will; for # instance, you could chop 25% at the bottom, and 10 voxels at the # top, of the z axis with "-crop 0,0 0,0 -25%,-10v". If no unit is # specified, mm is assumed. -guess_crop # Specify the objective function used for fitting. You can use # any of the objective functions supported by minctracc (do # "minctracc -help" for an up-to-date list); currently, these # are: # # xcorr cross-correlation # mi mutual information [Collignon] # vr variance of ratios [Woods] # zscore normalized difference # ssc stochasic sign change # # Only xcorr has been thoroughly tested and routinely used for # fitting MRI data to Talairach space. -objective xcorr