search

apsk14 / rdmpy

Official implementation of Ring Deconvolution Microscopy from https://arxiv.org/abs/2206.08928
MIT License
29 stars 6 forks source link

3D deconvolution? #3

Open dpshepherd opened 6 months ago

dpshepherd commented 6 months ago

This is a very nice approach,thanks for the clear paper and code. We have some 3D widefield fluorescence data that it would be great to try your approach on.

Do you have any plans to extend the model to 3D? Happy to provide some bead z stacks if it would be helpful.

apsk14 commented 6 months ago

Hey Doug!, Thank for reaching out, and for your interest! We are absolutely interested in extending to 3D. Currently, it is possible to do 3D by calibrating and deblurring each depth plane separately. The main inefficiency of this is the need to render out different PSFs at each depth. However, we feel that there is a more efficient way to do it!

I would love to collaborate on this. If you send me your z-stacks I can try to see if I can develop something quicker than just the naive slice-by-slice method. If you are interested, it is maybe best to send the data to my email apkohli@berkeley.edu (or whatever method you prefer).

If you wanted to try something right away, you can essentially just apply the code slice by slice, recalibrating and doing 2D ring deconvolution at each slice. If that's too slow you can replace ring deconvolution with DeepRD, although we only have models for 512x512 and 1024x1024 images. Let me know if you are interested in a particular image size, we can train a model for you!

dpshepherd commented 6 months ago

Sorry for being slow on this. One of the group members is collecting some better bead data and then we will a link over to you.