#30 Problem with manual segmentation

SNAP_2.0
open
nobody
None
5
2012-10-31
2009-11-18
Anonymous
No

Many thanks for your efforts with this software.

My problem occurs in v1.8 through the new v2.0 running on a Vista 64 bit OS. This problem has occurred on every brain I have tried.

I am inputting analyze hdr/img T1s. I then <Load Segmentation=""> and overlay with hdr/img's that have been segmented using an automated program. Voxels in these segmentations are labelled 2 to 8 with 8 being brain, the rest non-brain. My goal is to manually edit any errors (e.g. eyeballs being incorrectly labelled 8 ... I relabel them). Save segmentation as analyze hdr/img.

Then, when I try to re-open the segmented image on top of the greyscale again in snap, it says the voxel dimensions are not the same. mricRon gives "unable to read this image format 132". mricro opens the segmented image but the values are wrong ... what should be "8" is now "4". When I try to input the segmentations into our pipeline, I get an error message regarding the voxel dimensions not matching.

When I put the snap-segmentation hdr's into Gwanalyze header, it says:

dim.datatype: 132=unknown datatype
dim.bitpix: 16
dim.glmax: 0

In contrast, the segmented images that I inputed give the following, respectively, for the above values:
2=byte (8 bit unsigned)
8
255

The rest of the hdr information is the same between the images.

Any advice would be most appreciated as I have manually edited a large amount of data in itk-snap?

thanks

Discussion

  • Kevin Ma
    Kevin Ma
    2010-05-20

    Hi,

    I have encountered the same problem, where manual segmentation results from ITK-SNAP cannot be read by MATLAB. The main problem was that "datatype = 132" is not supported.