I would like to ask for some input regarding the registration of images acquired on consecutive physical sections of muscle tissue. In brief, I have a high-res single field of view image from section A that I would like to map back to a low-res overview image of the consecutive section B.
- Images are consecutive muscle tissue sections of 10 um thickness
- On both sections, the membranes have been labeled with an antibody against laminin.
- One of the sections has additionally been treated with a protease, leading to “deformed membranes”
- Find the matching location of the single field of view imaged at high-res image on the image of the entire section acquired at low resolution (a ROI would be enough)
- Only the location is needed, but NOT necessarily to warp/deform the images to match their shapes precisely
Only the membrane channel of:
the high-res image taken on section A (acquired at 100x, downsampled and flipped to match size and orientation of sample B)
15w_WT_cons_NMJ_EDL2_2.1_Tbx21-561_Chrne-647_Lam-488_07_processed.tif (206.9 KB)
a crop from section B (acquired at 10x) showing the corresponding area of the high-res image taken on section A:
low_mag_crop.tif (1.1 MB)
for completeness sake, the full image of the entire section B:
low_mag.tif (11.2 MB)
The structures in the consecutive sections are not identical, but are changed in size and shape, and thus only similar. The registration needs therefore to either account for the deformations or just be “relaxed” enough.
so far I have tried
- Descriptor based registration (2d/3d)
- Elastix in Fiji (thank you @Christian_Tischer for the awesome wrapper)
I tested various settings and models in both tools, but did not find the proper settings in either of them.
Any input you might have on settings or maybe another tool would be much appreciated!
Thank you for your help