Rebecca
Gray matter images from vbm are really probability images of a voxel being gray matter so the values range from 0 to 1. Therefore there is no way to find the perfect threshold at which you will get all and only gray matter since voxels that have a low but non-zero probability of being gray matter will be given a value of 1 if your threshold is >0 while voxels that have a high but not 100% chance of being gray matter will be excluded if your threshold is high. Usually when trying to select mostly gray matter values in the range of .15-.25 are used but this is a very rough guide.
--------------------
Darren Gitelman, MD
Sent from my mobile.
Office Ph: (312) 908-8614
Fax: (312) 908-5073
On Aug 23, 2011, at 6:23 AM, Rebecca Steketee <[log in to unmask]> wrote:
> Dear all,
>
> I am puzzled by the output of the imcalc function 'i1>...'. I am trying to create a binary mask of the grey matter image obtained by segmentation (I first normalized the grey matter image). I assumed that using a threshold of >0 would assign a value of 1 to the grey matter only (and all tissue that is not grey matter would be assigned the value 0, and therefore not be in the image). But instead, some white matter (e.g. around the ventricles) seems to be included into the mask as well, appearing as grey regions and having values of e.g. ~0.5. Could this be the result of imperfect segmentation?
> I tried to work my way around this by using different thresholds, but with a threshold of e.g. 0.4 some of the white matter still remains while at the same time I seem to lose some cortical tissue.
> What is the solution for this? Should I just play around with the threshold to find a balance, or is there a mathematical operation that I can use in imcalc that will result in an image containing grey matter tissue only?
>
> Thanks in advance for your help,
>
> Rebecca
|