[caret-users] Problem segmenting macaque brain
zinke at brain.uni-bremen.de
Tue Apr 6 04:47:44 CDT 2010
I had the same problem that in genral the segementation was doing fine,
but in part, especially in occipital regions, the results haven't been
convincing. I manually added missing brains voxel with the Segmentation
Volume Editor ('Turn Voxels on') in the *GraphErroCorrected file.
Another Different step in my procedure was a proper brain extraction
prior segmentation. Here, automatic methods such as FSL's Bet might
help, but again do not give satisfying results on macaque data. So The
extraction was done again manually, making sure to remove voxels
belonging to the dura and surface blood vessels (especially the central
In the end I got results I could live with, but it was a time consuming
I hope this helps,
On 04/05/2010 11:28 PM, Jochen Ditterich wrote:
> Thank you, Donna!
> I have uploaded a bias-corrected version named
> "freckles_bias_corrected.zip" (using BrainSuite's BFC algorithm). After
> adjusting the peaks, the segmentation runs through, but the
> posteriormost parts of the cortex are always missing. I have tried a
> manual correction of the segmentation volume, but, while I am able to
> remove unwanted parts from the fiducial surface, I am apparently unable
> to add missing parts. Is this possible? If so, is there a special step
> involved other than selecting "Segmentation Volume Post Processing" and
> the manually corrected segmentation volume?
> In case it is not possible to add missing parts using a manual
> correction of the segmentation volume, is there anything else I can do?
> Shifting the grey matter peak in the appropriate direction for including
> the missing voxels makes the segmentation process fail.
> On 4/5/2010 8:12 AM, Donna Dierker wrote:
>> Hi Jochen,
>> I think your biggest problem was that your peaks were set too low.
>> The MRI quality is not the best, and perhaps some nonuniformity
>> correction (e.g., FSL's fast) might improve your result further. But
>> your GM peak was set way too low. You histogram is somewhat misleading
>> (capture attached). If you set it at the peaks, the way you did, the
>> GMpeak will be way too low. See
>> http://brainvis.wustl.edu/help/peak_tweaking for more help on peaks.
>> Try GMpeak=190, WMpeak=230 and see how you do. My results looked
>> reasonable (reasonable.jpg). In my case, though, the graph error
>> correction method cut off a lateral slab. Turning it off should restore
>> sanity (if not a few pesky handles).
>> There were two issues with stereotactical_coord_AC_R.hdr:
>> * The name suggests right hem, but the hem appeared to be a left hem
>> (left-right flipped?).
>> * It was not centered near the AC, so the origin would need to be re-set
>> before you could hope to get any decent segmentation.
>> On 04/02/2010 01:20 PM, Jochen Ditterich wrote:
>>> Hi Donna and John,
>>> I am trying to segment a macaque brain and to flatten the cortex, but
>>> I haven't been able to make it through the segmentation process
>>> without error messages. I am not sure whether this is due to quality
>>> issues with the MRI or whether it has something to do with the
>>> implanted recording chambers. I have uploaded a file named
>>> "freckles.zip" that contains both the uncropped volume as well as the
>>> cropped volumes for the two hemispheres. Any feedback would be
>>> appreciated. Does anybody have experience segmenting MRIs taken from
>>> animals with implants?
>>> Thanks a lot,
>>> Jochen Ditterich, Ph.D.
>>> Assistant Professor
>>> Center for Neuroscience
>>> University of California
>>> 1544 Newton Court
>>> Davis, CA 95618
>>> office: +1 (530) 754-5084
>>> lab: +1 (530) 754-6987
>>> fax: +1 (530) 757-8827
>>> caret-users mailing list
>>> caret-users at brainvis.wustl.edu
>> caret-users mailing list
>> caret-users at brainvis.wustl.edu
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 51314 bytes
Desc: not available
More information about the caret-users