Dear Karl,
Has anybody had any success with the midline rois given by the atlas curvature.buckner40.filled.desikan_killiany that is automatically applied by autorecon3 under the newest stable release, version 3.0?
Our trouble lies in the regions labeled corpuscallosum and <anything>cingulate: this atlas is giving us a corpus callosum that is exactly the ventral cingulate sulcus, and the 4 cingulate gyrus segments, which is what we are interested in in this case, incorrectly exclude their ventral portion.
I presume this has been made to work at mgh, because Rahul Desikan's paper contains an illustration that is anatomically correctly labeled for these regions.
We are providing an orig/001.mgz image to the autorecon1 step that is talairach aligned, skull stripped, and resampled to 1.0 mm voxels, but otherwise is a signed 16-bit T1-weighted MRI scan. We have thousands of them, collected over the past 15 years. We are writing a nifti .nii file, and converting it with
mri_convert -it nii -ic 128 128 96 -oc 128 128 96 -iid -1.0 0.0 0.0 /raid1/structural/MR/9517294/0036195/10_ACPC/0036195_10_T1_brain.nii /raid1/data/methods/atlas/parcel/lobes/cloud_study/clipped_atlas_hand_masks/iowa_cortex_atlas/FreeSurfer_Subjects/IowaFifty_2005_0036195/mri/orig/001.mgz
We actually mean to numerically measure the extent -- the area and thickness, say -- of anatomical regions of interest including <anything>cingulate and apply statistical tests to these data. A line of guff about Type I and Type II errors and what is only presumable when identifying anatomical rois would be of no interest, and considered needlessly defensive. After all, the other 29 regions of Desikan's really great atlas are in working order!
Do we need to convince Rahul Desikan to release the "correct" *h.curvature.buckner40.filled.desikan_killiany.gcs files with FreeSurfer 3.0 that correspond to his paper?
Greg Harris The University of Iowa College of Medicine Psychiatry NeuroImaging Lab