Mrregister doesnt register well?

Hi mrtrix experts!

I am currently doing one small study with 9 subjects.

  1. First I made population template from these 9 subjects.
  2. I made registration of each subject to that template with mrregister (standard options, only provided .mif and mask file of both the picture and the template)

Afterwards I wanted to draw ROIs on population_template and hoped that these ROIs will appear at the same place on each 9 “mrregistered” pictures.
But it doesnt, I have serious missalignement.

What am I doing wrong? Or these are not even steps for what I am planing to achieve?

Hello David,

I am not sure if I understand your problem. Assuming your template looks reasonable, check the individual transformed images:

mrview input_transformed/*.mif -overlay.load template.mif -overlay.opacity 0.4

Make sure to use the transformed images, not the original input to population_template.

Are they well aligned with the template? If not, you’ll need to register them again using different parameters. I’d suggest starting with rigid registration. Also, make sure that the input images have similar intensity ranges and masks.


Not sure if it’s a problem with visualisation as Max suggested, or if there really is an issue. However, looking at your data it’s possible that the mean squared metric used by mrregster is not handling the large intensity variation (bias field), or maybe the signal from whatever medium the tissue is fixed in.

  • Have you used brain masks in population_template?
  • Did you perform intensity normalisation?
  • Might be worth computing FA maps, and seeing how these register. Because of the relatively similar brains of rodents, and limited white matter, you might be fine with a simple FA-based registration.