[Question] Does MCMICRO expect preprocessed input images compared to Cytokit?

[ShuaibingRen]

Hi there,

I'm comparing MCMICRO with Cytokit for CODEX data analysis and noticed significant differences in their preprocessing capabilities.

Key Differences Observed

Preprocessing Step	Cytokit	MCMICRO
Best Focus Selection	✅ CNN-based Z-plane selection	❌ Not included
Drift Compensation	✅ 3D subpixel cycle-to-cycle alignment	❌ Only stitching via Ashlar
3D Deconvolution	✅ Richardson-Lucy with auto PSF	❌ Not included

So I have some questions about MCMICRO:

1. Input Requirements: Does MCMICRO expect users to provide already preprocessed images (focused, drift-corrected, deconvolved) as input?

2. Preprocessing Responsibility: For CODEX data with:

   • Multi-cycle Z-stacks needing focus selection
   • Cycle-to-cycle drift requiring compensation
   • Optical artifacts needing deconvolution

   Should these be handled before running MCMICRO, or are there recommended external tools?

3. Design Philosophy: Is MCMICRO intentionally designed to start from "analysis-ready" images rather than raw microscopy data?

[ArtemSokolov]

Hi @ShuaibingRen,

The second entry in your table is off. ASHLAR does simultaneous stitching and registration: https://labsyspharm.github.io/ashlar/overview/overview-land.html#how-does-it-work

I also don't think it's quite right to say that MCMICRO expects already preprocessed images, given that it can accepts any Bioformats-compatible raw images as input. I think a more accurate statement would be to say that MCMICRO was developed to be used primarily for multiplexed 2-D image analysis and doesn't offer a lot of functionality for Z-stack processing.

[ShuaibingRen]

@ArtemSokolov Thank you for the clarifications! They’ve cleared up my misunderstandings and helped me better understand the use cases of MCMICRO.