Registering images from different sources

Hi! I’m new to Fiji and am looking for the best way to register my images. I see several plugins already included that do this but from my understanding they look like they register stacks of images from the same imaging source, eg. lining up different slices from electron microscopy. I have two images of the same object (the retina) but they are from different imaging modalities and thus create images that have many differences between one another. For example, one has the standard color fundus image as its background and the other is a zoomed in/more detailed grayscale picture of the vasculature. Both images have the same pattern of blood vessels in the background and I want to be able to register one image onto the other based on a prominent blood vessel (or multiple blood vessels). This would require scaling and transforming one image to match it with the other. In fact, I would actually like to enlarge the zoomed out image to match the more detailed one instead of mapping the smaller, zoomed in image on to the larger one - if that makes sense. The goal is to be able to make measurements in the same area between each image. I can do this manually but will have to do it for hundreds of images so I’m wondering if anyone has recommendations about which plug-ins might be able to accomplish this the best.

1 Like

Hm this sounds like a relatively advanced registration problem. You can have a look at more advanced medical image registration software. These implement the necessary algorithm for multimodal image registration:

http://elastix.isi.uu.nl/
https://simpleelastix.github.io/
http://stnava.github.io/ANTs/

2 Likes

There are to my knowledge two elastix wrappers in FIJI, one by @Christian_Tischer (https://github.com/embl-cba/fiji-plugin-elastixWrapper), and one that I did, (but which is currently not as functional as I would like https://github.com/BIOP/ijl-utilities-wrappers).

If you have some test image, that’d be nice to try.

As @schmiedc mentioned, maybe Fiji is not the best option for you, so try alternatives! (and anyway almost all what Fiji does for multimodal registration is to wrap these).

3 Likes

Hi Elev,

  1. Yes, you could try this: https://imagej.net/Elastix
  2. What we often do in such cases is to segment structures of interest in both images and then, using a combination of manual pre-alignment and elastix, find the transformation recipe on the segmentations, and then apply this transformation recipe to the raw data.
  3. You could also do it semi-manually by clicking on land-mark points in both images, using https://imagej.net/BigWarp
2 Likes