Hi, I want to do some visualization with FreeSurfer like this command:
freeview -f $SUBJECTS_DIR/fsaverage/surf/lh.inflated:annot=aparc.annot:annot_outline=1:overlay=lh.gender_age.glmdir/lh-Avg-thickness-age-Cor/sig.mgh:overlay_threshold=4,5 -viewport 3d
For example, I did the statistical analysis with my own data, but not with Freesurfer, to visualize my result, i want to use freeview, How can I save my p-value for vertx-wise or ROI-wise into a format .mgh and fit it into Freeview???
Or if any other softwares to visualize the result, I know PySurfer.
Any suggestion will be appreciated
Thanks in advance
If I remember correctly, the significance statistic is
sign(beta) * -log(p), where
beta is your result statistic and
p is the associated p-value. Assuming you have images with these two values, nilearn will get you what you want:
import nilearn.image as nli
nli.math_img('np.sign(img) * -np.log10(p)',
I’m assuming your inputs are NIfTI, but you can do
sig.mgh as easily.
Thanks very much for your reply, before I checked out a little for Nilearn, but I still got a little confused about this:
> where beta is your result statistic and p is the associated p-value. Assuming you have images with these two values,what is beta_fname and p_fname? these are two niffti images???
Actually, for example, I did some regional-wise or voxel-wise statistics, and I got the pvalue for each ROI/voxel, and I want to write this p-value into nifti file to visulize with good colormap.
If you have any good example, that would be great, thanks very much.
My above description will work for voxel-wise statistics, but if all you have is a p-value, you’ll have to omit the
np.sign(img) part. You could instead do:
As to ROI-based analyses, if you have a voxel-wise map of ROI ID numbers, you could do something like
roi_img = nb.load(in_fname)
roi_data = roi_img.get_data()
new_data = np.zeros(roi_data.shape, dtype=np.float32)
for roi_id, pval in ...:
new_data[roi_data == roi_id] = pval
pval_img = roi_img.__class__(new_data, roi_img.affine)
I hope this is what you’re looking for.