Register image sets before or after CellProfiler?

Sample details

  • Tissue slice imaged in xy multipoint (3% overlap between neighboring tiles), same tissue imaged multiple rounds using same XY multipoints stage positions
  • Scope field of view shifted slightly between rounds, so tiles do no match up perfectly (example of composite of two rounds from the same stage position: https://drive.google.com/file/d/1wLhHXeYNq_qj9Mjq-BwZB0DMRnjajuOp/view?usp=sharing)
  • Have working CP speckle counting pipeline that we have applied on prior samples

Analysis goals

  • Want to align rounds so parent cell identities are the same across rounds and speckles from all rounds can be attributed to these parent cells
  • Other analyses we do on these image sets involve stitching tiles and aligning using TrakEM2 in Fiji. Works reasonably well for generating large images, to our knowledge not a good solution for aligning tiles

Advice wanted

  • What aligning/tile registration tools would work for us? I have most familiarity with Fiji, intermediate Matlab and beginner Python user
  • Should alignment/registration be applied before input into CP? I am wary of losing information on borders of tiles if registering before analysis
  • Should alignment/registration be applied after CP? If so, should I add an image mask output to my pipeline or something like this? Or is it better to perform a displacement calculation using the location from CP and link parent cells across rounds this way?

Thank you for the advice!