I would probably use --ocn with mri_volcluster to create a single volume in which each voxel value is the cluster number it belongs to. Then split them out like:

mri_binarize --i ocn.mgh --match 1 --o cluster-0001.mgh

doug

Gregory Dierksen wrote:

Hi,

I'm trying to use the mri_volcluster command to make label files from a single volume of binary voxel clusters. Once I have split the volume into individual label files I want to convter the label files back into individual cluster volumes (I am trying to use mri_label2vol to do this). The commands I've been using are:

mri_volcluster --in microbleeds-reslice.mgz --allowdiag --thmin .5 --labelbase ../scripts/clusters/testing/cluster

mri_label2vol --label ../scripts/clusters/testing/cluster-0001.label --temp orig-flip-reslice.mgz --o cluster-0001.mgz

(NOTE: the -reslice is because these volumes have been resliced from 0.4492, 0.4492, 6.0000 sized voxels to conform to 1 mm^3 sized voxels; the dimensions of the image were altered as well, from 384 x 512 x 23 voxels to 173 x 230 x 138 voxels)

When I view cluster-0001.mgz with tkmedit the result is a 3D checkerboard like pattern with no two voxels having common edges (despite the original cluster in microbleeds-reslice.mgz being a solid cluster of 90 mm^3). What I wanted to see was that each cluster could be reconstructed in a single volume (i.e. cluster-0001.mgz, cluster-0002.mgz, etc with each volume containing only one cluster).

Also, when I use mri_cor2label and then mri_label2vol to switch between label files and volumes files, the clusters (despite not being split up like I want) are solid and do not display this checkerboard effect...leading me to think that its an issue with mri_volcluster.

Any help you can offer would be great. Thanks.

Greg