Back to the main page.

Bug 701 - ft_volumesegment skullstrip output is smoothed

Reported 2011-05-26 17:16:00 +0200
Modified 2011-06-01 13:42:00 +0200
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Mac OS
Importance: P1 normal
Assigned to: Jan-Mathijs Schoffelen
Depends on:
See also:

Robert Oostenveld - 2011-05-26 17:16:59 +0200

... whereas I expected it to be identical to the part of the original anatomy inside the skull. I.e. I had expected ft_plot_ortho(seg.anatomy==mri.anatomy) to result in a binary image of all voxels considered to be inside the skull.

Robert Oostenveld - 2011-05-26 17:17:23 +0200

>> seg.cfg ans = output: {'brain' 'scalp' 'skull' 'skullstrip'} trackconfig: 'report' checkconfig: 'pedantic' checksize: Inf downsample: 1 spmversion: 'spm8' write: 'no' keepintermediate: 'no' coordsys: '' units: '' inputfile: [] outputfile: [] template: '/Volumes/Data/roboos/matlab/fieldtrip/external/spm8/templates/T1.nii' name: '/private/tmp/tp64614982_f885_4170_929b_ec82133e0a28' smooth: [5 5 5 5] threshold: [0.5000 0.1000 0.5000 0.5000] version: [1x1 struct]

Jan-Mathijs Schoffelen - 2011-05-26 19:02:59 +0200

it actually does not smooth the anatomy. it applies the smoothing to create the brainmask from the tissue probability maps. Indeed it however gives misleading output to the screen. I'll look into this. Also, there is a potential typecasting issue in the multiplication with the brainmask. For the time being I will solve this by enforcing the anatomy to be double-precision.

Jan-Mathijs Schoffelen - 2011-05-26 19:24:30 +0200

I improved the comments to the screen, to actually reflect what is happening to which volume, and enabled the skullstrip to be of the same class as the input ft_plot_ortho(seg.anatomy==mri.anatomy) looks pretty OK to me.