fixed mcflirt mean image output

[juhuntenburg]

No description provided.

[mwaskom]

Can you explain a little bit more about what the problem was here?

[juhuntenburg]

I found the mean image that is output by -stats to be different from the one output by -meanvol. When you specify both options FSL actually outputs two files ('mean_reg' and 'meanvol'), where meanvol is missing top and bottom slices. It would be an alternative to adopt this but I found it even more confusing to output two slightly different mean images, one of which is corrupted anyhow.

[coveralls]

Changes Unknown when pulling 6b23666 on juhuntenburg:fix/mcflirt_mean into * on nipy:master*.

[juhuntenburg]

Maybe I should add that my original problem was that nipype, as opposed to FSL, at the moment outputs no mean image at all when only the -mean_vol option is specified but not the -stats_imgs option. I'm still in two minds whether it wouldn't be better to just go the FSL way and output the two mean images independently for -stats and -meanvol. One reason against it is that the description of the -stats doesn't even mention a mean image.

[satra]

@juhuntenburg - how does this relate to matthew webster's comment on the fsl mailing list. there seems to be some fix that happened? did this get fixed in fsl?

[juhuntenburg]

Hej Satra, took me a while to disentangle but I believe it's two different problems. The motion parameters substantially differ between 4.1.9 and 5.0.6, but the mean images seem to be the same and the issue described above remains. But I will bring this up on the FSL mailing list, too.