3dMatch percentage match less than 50% acceptable?

AFNI Message Board
1

Hi there,

I was performing 3dMatch, but wanted to look at the folder that is outputted which reports the percentage match between the standard functional template (brain networks) against your data (independent components)

The REF folder looked something like this:

0 5 0.504 0.187
1 8 0.460 0.200
2 3 0.397 0.164
3 4 0.390 0.141
4 17 0.428 0.107
5 10 0.404 0.182
6 0 0.512 0.266

the first column lists all the 7 networks, The second column contains the Indpendent components from MELODIC (GICA) that matched the most with each of the 7 networks. And the column next to that is the percentage match.

But I was wondering… would it look bad to report a match percentage less than 50%? Or should we argue that, that’s what the data is like and that it’s beyond our control?

For instance, the DMN doesn’t strictly have DMN regions activated but it also contains the occipital, and other components, therefore making it only match 51% (network 6)

I’m not sure what to do. Or is this okay?

2

Hi, Sondos-

3dMatch does give quantities of “matching-ness”, it is true. However, these are still mostly for relative matching, not really anything absolute. It is a useful way to help find closest matches among data, but everything is still pretty relative. Volumes can match because they are relatively the best match, but they can still be overall poor fits. A second pass with eyes is quite useful for discerning what are most likely to be “real” matches, I think. I am not sure what would be a “reasonable” cutoff value, a priori.

–pt