Hi Jared
I think you description is accurate, but I believe we compute the distances both ways then average them, making it symmetric. As to surfae area corrections, I guess that kind of depends on what you care about.
cheers Bruce On Wed, 5 Sep 2018, Zimmerman, Jared wrote:
External Email - Use Caution
Hi Freesurfers,
I’m using mris_compute_parc_overlap to compare the similarity between two parcellations and I’m curious about how the average minimum distance measurement is calculated.
The help text for the function says “… a table of mean minimum distances between corresponding labels.” My questions are 1) what set of measurements is being meaned here, and 2) what exactly is the measurement. My current interpretation is that for every vertex in annot1 the minimum distance to a vertex in annot2 with the same label is calculated such that overlapping regions have a distance of zero. The mean is then calculated over these distance measures for all vertices in a given label.
If this is true, do you recommend doing any subject level corrections? Seems like subjects with larger cortical surface areas would be biased to have larger distances.
Thanks,
Jared