binarybottle
commented
9 years ago Thank you very much! I have confirmed in Neurostars that Mindboggle's default is 31 cortical surface labels per hemisphere, and I have revised the http://mindboggle.info/faq/labels.html and mindboggle/mio/labels.py documentation so that they all clearly state this.
I used to have an argument where one could set the labeling protocol to either 31 or 25 labels to conduct experiments just like the one you are interested in, but found that this confused users (hence the Neurostars query). I would suggest using the relabel_volume() function in mindboggle/guts/relabel.py to consolidate the few labels necessary to convert DKT31 to DKT25: [10,23,26,27,19,20] to [2,2,2,3,18,18](+1000 for left, +2000 for right), then run thickinthehead() on the resulting files for comparison. I am very interested to hear what you find!
In the FAQ (http://www.mindboggle.info/faq/labels.html), there is some rationale outlined for combining some of the regions that are delineated in the DKT atlas, e.g., combining the three inferior frontal regions into a single region. However, in the example mindboggle output, it appears that the three regions are not combined (e.g., thickinthehead_per_freesurfer_cortex_label.csv). I can see that it is currently an open question of which is better, i.e., 31 or 25 cortical regions (https://neurostars.org/p/2680/). However, that post does indicate that mindboggle currently outputs 25 regions.
Is this an discrepancy in the current output vs. FAQ due to a change to using 31 regions instead? Perhaps a solution to this would be to default to one set of regions (31 vs. 25), but allow the user to specify if they want the other as a flag when running mindboggle? In either case, the FAQ should be updated to match the current output of mindboggle.
Part of my interest in using mindboggle was to see how cortical thickness estimates would look in the DKT-25 protocol, so I wanted to see an example output that included both sets (i.e., FreeSurfer would use the 31 regions still, mindboggle would provide the 25 regions, calculated from the same participant.