Dear all,

I would like to capture regions where the gray matter/ white matter junction is blurred in order to detect possible focal cortical dysplasias.

As I understood the WM/GM segmentation is done by the script mri_segment.

Would it be possible detect those regions where the gradient between GM and WM regions is low? 

I imagined that running mri_segment several times, with each time a different lower GM threshold and higher WM threshold, and then calculating the diference between the results, might give such information.

I am very much interested in what you think about this approach and how to do it practically.

For info here are the optional flags of mri_segment:

-slope <s>

set the curvature slope (both n and p)

-pslope <p>

set the curvature pslope (default=1.0)

-nslope <n>

set the curvature nslope (default=1.0)

-debug_voxel <x y z>

set voxel for debugging

-auto

automatically detect class statistics (default)

-noauto

don't automatically detect class statistics

-log

log to ./segment.dat

-keep

keep wm edits. maintains all values of 0 and 255

-ghi, -gray_hi <h>

set the gray matter high limit (default=100.000)

-wlo, -wm_low <l>

set the white matter low limit (default=90.000)

-whi, -wm_hi <h>

set the white matter high limit (default=125.000)

-nseg <n>

thicken the n largest thin strands (default=20)

-thicken

toggle thickening step (default=ON)

-fillbg

toggle filling of the basal ganglia (default=OFF)

-fillv

toggle filling of the ventricles (default=OFF)

-b <s>

set blur sigma (default=0.25)

-n <i>

set # iterations of border classification (default=1)

-t <t>

set limit to thin strands in mm (default=4)

-v

verbose

-p <p>

set % threshold (default=0.80)

-x <filename>

extract options from filename

-w <w>

set wsize (default=11)

-u

usage

 

Thank you very much for all comments!

Markus