Dear FSL People,
Echoing my previous emails and recent update of atlases shipped with FSL, I
have cooked up a simple derived atlas: Harvard-Oxford Cortical
(Lateralized) which I am considering to upload as a part of
fsl-harvard-oxford-atlases binary package in NeuroDebian.
Its difference from the stock Harvard-Oxford Cortical is simple: all areas are
now present twice, separately for left and right hemisphere.
Probabilities are computed based on liberalization probability within HO
subcortical atlas shipped with FSL.
Due to int8 rounding, probabilities in Lateralized atlas for some voxels
might not add up exactly to the probability in non-Lateralized version
and be a percent less in the sum.
Representative voxel for each area is the one with maximal probability and
closest to the geometrical center of the area (if there are multiple such
Areas are sorted by the area and not by hemisphere.
Such derived atlas would be always regenerated if updates to the
Harvard-Oxford atlases would follow, and there are few basic assertions
in the code to verify that it remains "correct". Here is the code I
cooked up for this purpose:
And you can investigate the generated volumes/atlas (xml):
Please let me know if you have any objections, e.g. you would advise to ship
generated atlas in a separate binary package, e.g.
fsl-harvard-oxford-atlases-derived to disambiguate on the origins of the atlas
etc. Or if you spot some mistakes.
Thank you in advance
Yaroslav O. Halchenko, Ph.D.
http://neuro.debian.net http://www.pymvpa.org http://www.fail2ban.org
Senior Research Associate, Psychological and Brain Sciences Dept.
Dartmouth College, 419 Moore Hall, Hinman Box 6207, Hanover, NH 03755
Phone: +1 (603) 646-9834 Fax: +1 (603) 646-1419