search

nipy / mindboggle

Automated anatomical brain label/shape analysis software (+ website)
http://mindboggle.info
Other
145 stars 54 forks source link

Joint fusion atlas construction bug #48

Closed binarybottle closed 9 years ago

binarybottle commented 9 years ago

There was misalignment between the subcortical and cortical labels in the individual OASIS-TRT-20 brains (http://mindboggle.info/data.html). It seems that s​ome of the brain images corresponding to the subcortical label files were assigned to the wrong cortical files, leading to misalignment. This may affect the accuracy of the joint fusion atlas constructed from these combined label volumes.

binarybottle commented 9 years ago

Corrected the assignments and reconstructed the individual BrainCOLOR subcortical and DKT31 cortical OASIS-TRT-20 whole-brain label volumes. Constructed a new joint fusion atlas from these combined label volumes, but there was hardly any change. Both are called "version 2".

satra commented 9 years ago

@binarybottle - so i should generate a 2mm version of this? what was the misalignment? we have processed a few datasets using that joint fusion atlas.

binarybottle commented 9 years ago

I generated the new 2mm template on the site (is ends with "v2" for "version 2," so no worries. The cortical labels were unaffected. It turns out there was a discrepancy between the IDs assigned to the MICCAI2012 challenge / BrainCOLOR volumes and the renumbered OASIS-TRT-20 volumes, so some of the transforms were applied to the wrong subcortical labels. Fortunately, there is extremely little difference from the earlier version.

satra commented 9 years ago

thanks @binarybottle