[MINC-users] NMI for minc volumes

[jon pipitone]

Hey MINC Users,

I'm wondering if anyone here knows about, or has written, some code to
compute the normalised mutual information of two minc volumes[1]?

I see that minctracc uses nmi, and has a -measure option to output various
objective functions[2], except not NMI.  I'm hoping that there is a magical
utility out there to do this already that I'm just not aware of.

Otherwise, I figure I'd dip into the minctracc code and see if I can expose
the measure either through the -measure flag, or as a separate utility (a
la xcorr_vol).  If I go this route and anyone has helpful pointers or
gotchas, let me know.

Thanks all,

Jon Pipitone
Research Imaging Centre
Centre for Addiction and Mental Health
Toronto, Ontario


[1] I've found one other person asking about this way back:
http://www.bic.mni.mcgill.ca/pipermail/minc-users/2005-August/000805.html

[2] But I'm wary of the quality of the output:
http://www.bic.mni.mcgill.ca/pipermail/minc-users/2007-September/001705.html