search

gerlichlab / looptrace

Fork (from EMBL Gitlab) of the looptrace project: https://git.embl.de/grp-ellenberg/looptrace
MIT License
2 stars 1 forks source link

Auxiliary tool for using drift correction to write shifted images #288

Open vreuter opened 7 months ago

vreuter commented 7 months ago

A non-pipeline, standalone version of #151

vreuter commented 7 months ago

https://git.embl.de/grp-ellenberg/looptrace/-/blob/master/looptrace/Drifter.py#L287-358

vreuter commented 7 months ago

Requested by @josephneos1010

vreuter commented 7 months ago

@josephneos1010 since these would be very large image files (effectively doubling the input size, regarding the "input" as post-zarr-conversion), can you say how you'd use these? It's very preferable to avoid this slow write process and huge data storage, and no part of the processing or analysis operates on entire drift-corrected images, so it would be ideal to address your question via another route.

josephneos1010 commented 7 months ago

For a FOV, I would like to align certain hybridization steps so that I could visualize different signals on top of each other. I guess it doesn't have to align all hybridizations in a FOV. What would be nice is for me to select images to align and there is someway to align them with a pipeline instead of manually aligning them one by one.

vreuter commented 7 months ago

Thanks @josephneos1010

For a FOV, I would like to align certain hybridization steps...I guess it doesn't have to align all hybridizations in a FOV.

OK, so a subset could be specified...do you know anything about the structure of the timepoints you'd like to compare? (e.g., RNA t0 vs. DNA t0, RNA t1 vs. DNA t1, etc.)

so that I could visualize different signals on top of each other.

Once aligned, is only the FISH channel of interest? That would save us the space of the beads channel.

If there's any subsetting in xy (i.e., certain regions) or in z (certain slices), please also not here.

vreuter commented 7 months ago

After discussion w/ @josephneos1010 it makes most sense to restrict FOVs but leave timepoints (hybridisation rounds) unrestricted, so write out full stacks in TZYX but only for one or a few FOVs.