Compensate for lateral spillover

[ArtemSokolov]

• [x] Implement a (containerized) Python version of REDSEA
• [ ] Implement a detector of potential spillover
• [ ] Use the detector to evaluate REDSEA and identify when it works or doesn't
• [ ] Update MCquant to use REDSEA for lateral spillover compensation
• [ ] Implement a new solution for spillover cases where REDSEA is ineffective

[ciszew]

Hello Developers I was wondering if the plan for lateral spill compensation is still in plans for future release of MCQuant? Any ideas how much improvement of quantification results would it offer?

[ArtemSokolov]

@clemenshug can chime in here, but I don't think we ever arrived a method that was working consistently well across different image tiles. Unfortunately, we found that REDSEA was often overcompensating, which led to some strange cell type calls (the container is available here, if you want to play with it: https://github.com/labsyspharm/redseapy). We were also working on an in-house method, but the challenge there was evaluation. Since there is no ground truth, we had to rely on manual inspection, but that doesn't really allow to quantify the exact amount of improvement.

@ciszew I am open to other alternatives, if you know of any good methods in the wild.