The input format of CODEX Image to process MCMICRO

[Yuchen588]

Hi, everyone,

I have a couple of TIF formatted data generated from CODEX toolkit, these images were stitched without further registration, so I think we can just start with the data segmentation step from the MCMICRO program. However, I wonder if the input data is necessary with a single ome.tiff data (with the metadata in it). I have a few stitched tiff data annotated their cycle and channel and I'm not sure if these data were suitable for MCMICRO segmentation and quantification process.
In addition, I have already provided the detailed marker annotation (marker.csv) and processed the segmentation (unmicst) step and it seems can not be finished successfully to get quantification data. Could you please give me some advice on the input image formats and the parameter tuition, thanks a lot!

Best,
Yuchen

#>>>>>>>>>>>>>>>>>>>>>>>>command line<<<<<<<<<<<<<<<<<<<<<<<
root@radonc-LUAD.test.run# nextflow run labsyspharm/mcmicro --in /home/lyc/LUAD/rawdata/LUAD.test.run -params-file /home/lyc/LUAD/rawdata/LUAD.test.run/params.yml
N E X T F L O W ~ version 23.04.4
Launching https://github.com/labsyspharm/mcmicro [happy_picasso] DSL2 - revision: 01c11e5 [master]
executor > local (22)
[- ] process > illumination -
[- ] process > registration:ashlar -
[- ] process > background:backsub -
[- ] process > dearray:coreograph -
[- ] process > dearray:roadie:runTask -
[- ] process > segmentation:roadie:runTask -
[df/5ae9d5] process > segmentation:worker (unmicst-2) [ 0%] 0 of 22
[- ] process > segmentation:s3seg -
[- ] process > quantification:mcquant -
[- ] process > downstream:worker -
[- ] process > viz:autominerva -

#>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>image format <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
reg001_cyc005_ch003_CD66b.tif
reg001_cyc001_ch001_DAPI-01.tif
.
.
.

#>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>The MCMICRO params.yml file<<<<<<<<<<<<<<<<<<<<<<<<<<<<
workflow:
start-at: segmentation
stop-at: downstream
qc-files: copy
tma: false
viz: false
background: false
background-method: backsub
multi-formats: '{.xdce,.nd,.scan,.htd}'
single-formats: .tif
segmentation: unmicst
segmentation-recyze: false
downstream: scimap
segmentation-channel: 1
options:
ashlar: -m 30
unmicst: --tool unmicst-solo
cypository: --model zeisscyto
ilastik: --num_channels 1
mcquant: --masks cell*.tif _cp_masks.tif
naivestates: -p png
imagej-rolling-ball: 100 -n=4 -j="-Xmx4g"
modules:
watershed:
version: 1.5.5-large

[ArtemSokolov]

Hi @Yuchen588,

Ideally, you want to register individual cycles into a single image pyramid prior to segmentation. Otherwise, you will need to figure out a way to align segmentation masks before passing them to quantification.

I am checking with @Yu-AnChen, but I think this is a great use case for his tool palom: https://github.com/labsyspharm/palom

The tool will output an image pyramid that you can place in the registration/ subfolder, and then start-at: segmentation should work as expected.

[Yuchen588]

Thanks Artem, that's may helpful for me. Our CODEX images for each marker were already registrated and want to merge into a single ome.tif format image, which can be used as a input file in the registration folder. I download the example CODEX image file (PilotTonsil_5_z08.ome.tif) and can be successfully processed segmentation and quantification by my custom code. However, we are still having difficulties with the image and I intend to use Fiji (ImageJ) software to solve the problem.

[Yuchen588]

The ImageJ software can handle tif image combinations aforementioned, I just followed the instructions at https://chat.openai.com/c/ccf595cc-8a85-475b-b219-bb7c3ff6ba36, but there seems to be another problem in the s3seg section, namely "IndexError. Index out of range". I'm not sure if it's a matrix handling problem like the matrix is too large.

[ArtemSokolov]

Responded to your question here: https://forum.image.sc/t/the-error-in-processing-s3seg/87313/4