Hello,
I'm using a home-grown GCA constructed from five acallosal
subjects to try to improve the automatic labeling and segmentation of
volumes from this population. However, the labeling was quite bad when
performed on a representative subject, and the
ventricles were only partially filled and crossed the IHF in areas:
Left_Lateral_Ventricle (4): linear fit = 1.00x + 0.0 (713 voxels,
overlap=0.480)
Left_Lateral_Ventricle (4): linear fit = 1.00x + 0.0 (713 voxels, peak
= 38), gca=37.8
Right_Lateral_Ventricle (43): linear fit = 1.00x + 0.0 (370 voxels,
overlap=0.292)
Right_Lateral_Ventricle (43): linear fit = 1.00x + 0.0 (370 voxels,
peak=38)
I also noticed that the 152-subject MNI atlas that came with FS
had ventricles that were much "darker" (when converted to NIFTI
and displayed in fslview) than my own atlas: the mean ventricle
intensities were about 15 and 30, respectively. We suffer from a lack
of data, too.
Is it possible to improve the GCA and labeling with such limited
data?
Jason
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
