Register postmortem .25mm hemisphere into MNI space

Dear all,

I have scanned two postmortem hemispheres next to each other, but not touching, one left one right, at .25mm resolution. My images look like this [ ᗡD ]. I am wondering how to go about registering these hemispheres to MNI space. Should I split the GRE into two images, each with one hemisphere, and make separate masks? Do I need to split the MNI template? What MNI template should I use? I heard there is one at 0.4mm iso. Any information would be appreciated, thank you all.


Welcome Jackson!

I’m not a registration expert, so others are more than welcome to join in, but I’ll have a stab since nobody else is willing to stick their neck out.

If it were me, I would:

  • Splitting your acquired image down the middle, and registering each hemisphere separately to a template image that contains only the corresponding hemisphere.

  • While asymmetric templates tend to be preferable for many applications, here I would probably myself be looking at symmetric templates, because it would (hopefully) mean that splitting that image in half along the left-right axis would correspond to accurately splitting the two hemispheres. Any time registration is performed where there’s “content” in one image but not the other, things can go awry, and indeed the “correct” answer in such cases is ambiguous. So you want the template image to be cropped in a similar fashion to that of your sample, and my first instinct is that using a symmetric template will make that more likely.

  • Consider, after splitting the template image in two, actually padding each template image with a few zero-filled slices adjacent to the midline. If you want medial structures to be aligned as well as is possible, you don’t want them to be corrupted by any FoV edge effects (which are entirely implementation-specific), so giving the images some empty space between the FoV edge and the content you want aligned may be beneficial. Entirely hypothetical though.