[MINC-users] mincresample modifying data

[Andrew Wood]

Hi all,

I've been experiencing some behaviour coming out of mincresample that I
can't explain. I've been trolling through the source code and wondering
about a number of things that might contribute to my problem, but I've
boiled it down to one simple example:

Here is a volume in.mnc (a signed float volume):
https://www.dropbox.com/s/cyfujyenc4j7h8b/in.mnc

$ mincresample in.mnc rsl.mnc

Then compare with mincstats:

$ mincstats -min in.mnc
Min:               0
$ mincstats -min rsl.mnc
Min:               -5.345658291e-11

And here it is from the header:

$ mincinfo -attvalue image:valid_range in.mnc
0 26989.404296875
$ mincinfo -attvalue image:valid_range rsl.mnc
-5.3456582910227368188e-11 26989.404296875


I also tried with -keep_real_range, which I gather doesn't make sense on
float data anyway, but definitely further mucks things up:

$ mincresample in.mnc rsl_krr.mnc
$ mincstats -min rsl_krr.mnc
*** mincstats - reported min (-5.34566e-11) doesn't equal header (0)
*** mincstats - reported max (6.90125e-11) doesn't equal header (1)
Min:               -5.345658291e-11

Does anyone have any insight into what might be happening?

Thanks,
Andrew