Hi Nick,
It appears mri_compute_seg_overlap only works with .mgz files and not
.label files, which is how I saved the ROIs. I tried renaming the
labels as .mgz files, but that did not work. Any suggestions? Thanks.
> the utility for comparing segs is 'mri_compute_seg_overlap', included
> w
the utility for comparing segs is 'mri_compute_seg_overlap', included
with freesurfer. it compares two arbitrary segs. we use it to compare
the manual seg with the auto seg to get Jaccard and Dice coefficients.
On Mon, 2009-08-17 at 12:03 -0400, Bruce Fischl wrote:
> if all you want is reliability
if all you want is reliability on the same volume then you don't need any
reconstructions. Just manually label them in something like freeview and
directly compute the overlap (Nick has some tools for this if you don't -
it's pretty straightforward)
cheers,
Bruce
On Mon, 17 Aug 2009 jen...@nmr.
Hi Bruce,
Essentially, for this ROI reliability problem we've asked you about, we
want to use the high-res volumetric data to make a refined segmentation
of the amygdala. We can't do this by using the 1mm data because of its
visual integrity. But, maybe what we can do (as per your email) is mak
Hi Bruce,
Just chiming in here given that I've worked on a similar project for
brad (hi res hippocampus segmentation). essentially we want to use the
hires volumetric data to make a refined segmentation of the amygdala.
such a segmentation is not really possible to do given the visual
inte
Hi Jon,
I would use the 1mm data to create the surfaces then map them to the
hires data and do your tracing that way. You can then use the surfaces for
various things like inter-subject registration. I'd need more details to
give better advice, but generating surfaces from highres data is hard,
Hi Bruce,
Yes, we have the 1mm data. However, the project is designed to
manually trace an ROI based on the increased visibility in the high-res
scans. We need to do reliability to test our method of manual
segmentation. Perhaps I am misunderstanding, but it seems like using
the 1mm surface
can't you get 1mm data to go with your highres data? For many reasons
the high resolution data is usually quite poor for generating surface
models.
cheers,
Bruce
On Sun, 9 Aug 2009 jen...@nmr.mgh.harvard.edu wrote:
> Hi Freesurfer team,
>I am trying to get intra-rater reliability statistic
Hi Freesurfer team,
I am trying to get intra-rater reliability statistics on an ROI
manually traced twice (blindly). The scans are anisotropic (.38 x .38
x .8 mm) and so I have not run them automatically through the
Freesurfer segmentation because they fail after autorecon1. I tried to
use th
