Autominerva do not show my image

[tkcaccia]

I have a multi-immunofluorescent image with 7 channels: DAPI, PDL1, FOXP3, Ki67, CD4, HMW cytokeratin, autofluorescence.

The folder vis is create and populated by all files and folder. When I open the file index.html, the MINERVA is open but there is not the image of the tissue

[ArtemSokolov]

Hi @tkcaccia,

If you check the .jpg images in the viz directory, do they look correct or are they empty?

[tkcaccia]

There are not empty but their quality is not good.
I am able to open the image correctly with QuPath. Maybe this problem is related as well as why the segmentation fails

[ArtemSokolov]

Yea, that just looks like white noise to me. Does this match what you see in raw/ and registration/? At what step in the pipeline does your signal disappear?

[tkcaccia]

My image is an ome.tiff in 16 bit with 7 channels.

My params.yml

workflow:
  start-at: registration        
  stop-at: quantification         
  segmentation: unmicst  
  segmentation-channel: 1
  viz: true                    

options:
  s3seg: --nucleiRegion localMax

My original image:

In the folder raw the image is changed and looks empty. The scale of each channel is changed.

Registration image:

The registration image showed some problems

[tkcaccia]

Here below the link to the original image 2.3 GB.
Via WeTransfer: https://we.tl/t-R0AMqyayn7

[ArtemSokolov]

Yea, something weird is going off. Registration should not be changing your image this much.
@jmuhlich might be able to say more.

How many tiles/cycles are you stitching/registering?

[tkcaccia]

I do not know how to answer your question. Mcmicro seems to change the raw image too

…

On Sun, 18 May 2025, 23:02 Artem Sokolov, ***@***.***> wrote: *ArtemSokolov* left a comment (labsyspharm/mcmicro#578) <#578 (comment)> Yea, something weird is going off. Registration should not be changing your image this much. @jmuhlich <https://github.com/jmuhlich> might be able to say more. How many tiles/cycles are you stitching/registering? — Reply to this email directly, view it on GitHub <#578 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYRG5K52VGSJBBC5ODGOBD27DYPTAVCNFSM6AAAAAB427ZIS6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQOBZGIYTAMRWGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[ArtemSokolov]

MCMICRO treats the raw/ directory as read-only, so it should not be changing the input image.

The registration step assembles multiple tiles and cycles into a single image. There is more information on the website of the module that does this: https://labsyspharm.github.io/ashlar/overview/overview-land.html#how-does-it-work

If your input image is already stitched and registered, then you don't really need the registration step, and you can start-at: segmentation instead.

[tkcaccia]

I tried to start from registration but index.html still looks empty (all dark).
I opened the images in jpg and they look ok.

Here below a tile of the DAPI

[ArtemSokolov]

@tkcaccia I forwarded your issue to Minerva developers. You can find the link directly above this comment.