External Email - Use Caution
Hello,
The ex-vivo BA label and annotation files named ?h.BA??_exvivo.thresh.label and BA_exvivo.thresh.annot seem to be thresholded versions of the corresponding probabilistic BA labels. Was a simple threshold on the probability used and if yes, what was the value?
In addition the annotation files, whether thresholded or not, only have one label per vertex. Are the labels based on the maximum probability across all BA labels?
Thanks
Julien
External Email - Use Caution
Hi Julien
In researching this a few years ago I asked the question
What exactly has been
done to change the .label (e.g. lh BA44.label) to .thresh.label (e.g. lh BA44.thresh.label?
And was told
The .thresh labels have been
thresholded with a p-value threshold that is "best" on our training set of manually/architectonically defined labels.
I then looked into this a bit further and found the following explanation:
On 14-09-08 12:39 PM, Anastasia Yendiki wrote:
On the difference between .thresh and "not .thresh" labels: The BA labels come from averaging histologically derived labels from a set of post mortem brains. The non-thresholded version is just the sum of the corresponding labels from the different brains in fsaverage space. So vertices are included even if only one of the brains had their BA in that vertex, which means that these labels are going to be rather large. In the thresholded version, a threshold has been applied to the average BA label, to make it have an area as close as possible to the average area of the individual BA labels. I hope this makes sense.
I hope this helps!
Trisanna Sprung-Much Research Associate McGill University MD Anderson Cancer Centre
freesurfer@nmr.mgh.harvard.edu