Hello,
I am working on comparing hand-drawn subcortical volumes to those output automatically by FreeSurfer. While reading about the segmentation process on the wiki, I came across this line:
"The final segmentation is based on both a subject-independent probabilistic atlas and subject-specific measured values. The atlas is built from a training set, i.e., a set of subjects whose brains (surfaces or volumes) have been labeled by hand. These labels are then mapped into a common space (Talairach space for volumes..." (http://surfer.nmr.mgh.harvard.edu/fswiki/FreeSurferAnalysisPipelineOverview)
Because of the way we decided to delineate our ROI, we need to compare the shapes we defined in AFNI to those defined by this training set and output by FreeSurfer. I've been looking for more information about this probabilistic atlas, but I cannot find any more information about it. The only thing I found was the last line of this thread from the mail archives:
Am 23.08.2011 um 04:03 schrieb Anthony Dick:
The subcortical segmentation identifies subcortical structures such as
thalamus, basal ganglia structures, hippocampus, amygdala, and
cerebellum. The cortical parcellation delineates different areas of the
cortex, such as the superior temporal sulcus, the central sulcus, the
pars triangularis, opercularis, and orbitalis of the inferior frontal
gyrus. A recommendation is to get a copy of the Duvernoy atlas, which
covers definitions of different parts of the cortex. The Mai atlas is
good for learning subcortical structures.
Was the Mai atlas used to create the define the structures in the training set? Could you provide me with any more information regarding the training set?
no, we created our own atlas by manually labeling 39 subjects according to the CMA conventions, then iteratively going back over them to correct for consistent inaccuracies in the manual labelings. It's still an ongoing process as we are currently working on improving the manual putamen labels.
cheers Bruce
On Wed, 19 Oct 2011, cloud.ctrl@me.com wrote:
Hello,
I am working on comparing hand-drawn subcortical volumes to those output automatically by FreeSurfer. While reading about the segmentation process on the wiki, I came across this line:
"The final segmentation is based on both a subject-independent probabilistic atlas and subject-specific measured values. The atlas is built from a training set, i.e., a set of subjects whose brains (surfaces or volumes) have been labeled by hand. These labels are then mapped into a common space (Talairach space for volumes..." (http://surfer.nmr.mgh.harvard.edu/fswiki/FreeSurferAnalysisPipelineOvervie w)
Because of the way we decided to delineate our ROI, we need to compare the shapes we defined in AFNI to those defined by this training set and output by FreeSurfer. I've been looking for more information about this probabilistic atlas, but I cannot find any more information about it. The only thing I found was the last line of this thread from the mail archives:
Am 23.08.2011 um 04:03 schrieb Anthony Dick:
The subcortical segmentation identifies subcortical structures such as
thalamus, basal ganglia structures, hippocampus, amygdala, and
cerebellum. The cortical parcellation delineates different areas of the
cortex, such as the superior temporal sulcus, the central sulcus, the
pars triangularis, opercularis, and orbitalis of the inferior frontal
gyrus. A recommendation is to get a copy of the Duvernoy atlas, which
covers definitions of different parts of the cortex. The Mai atlas is
good for learning subcortical structures.
Was the Mai atlas used to create the define the structures in the training set? Could you provide me with any more information regarding the training set?
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
cloud.ctrl＠me.com