1_prepare_ilastik.cppipe pipeline not working

[akedarg]

Hi pipeline creator,
I am new to IMC data analysis so please bear my ignorance. My query related to this- https://bodenmillergroup.github.io/ImcSegmentationPipeline/ilastik.html

While trying to import 1_prepare_ilastik.cppipe pipeline into cellprofiler, i get this warning upfront- Error while loading SummarizeStack: Could not find SummarizeStack module. Do you want to stop processing? whether it is 4.2.1 or 4.2.6 version.

if i continue and go forward anyway, and analyze the images, analysis shows complete. there are not output images or any files in crops folder.

Please help. Thanks.

[Milad4849]

Hi @akedarg
I had a quick look, I could not reproduce the error on the first go. Before I ask you for further details, I want to ask you whether you have a specific reason to use imcSegmentationPipeline for segmentation, if not I would suggest using steinbock which offers three methods for segmentation, including two deep learning ones, which in general perform significantly better. There is extensive documentation and I will be able to offer more effective support in that case.

[nilseling]

To fix your issue you will need to correctly set the Cellprofiler plugins as explained under step 4 in the Usage section of the Readme.

[akedarg]

To fix your issue you will need to correctly set the Cellprofiler plugins as explained under step 4 in the Usage section of the Readme.

Dear Nils, that was it. Thanks.

[akedarg]

Hi @akedarg I had a quick look, I could not reproduce the error on the first go. Before I ask you for further details, I want to ask you whether you have a specific reason to use imcSegmentationPipeline for segmentation, if not I would suggest using steinbock which offers three methods for segmentation, including two deep learning ones, which in general perform significantly better. There is extensive documentation and I will be able to offer more effective support in that case.

Hi Milad, you are right. I am planning to use steinbock too. But i thought i can customize the segmentation training better with manual method. Isn't that the case? I am working with a muscle tissue which creates a problem because segmentation protocol cannot differentiate between a nucleus on the cells (where i need pixel expansion to create a membrane) and the nucleus on the fibre (where i dont want algorithm to automatically create a membrane). Let me prepare some images to show you my issue.

[akedarg]

Segmentation issue_KedarG.pdf

Here is my issue and the main reason for using imcSegmentationPipeline. The issue is that in a regenerating muscle tissue, there is a central nucleus in each regenerating fibre (page 1). 3 DPI, there is immune cell infiltration and segmentation works because it's just immune cell nuclei and no regenerating nuclei (page 2). Problem starts where ROI has both regenerating fibre and immune infiltration areas (Fig 3). Here, we need to skip fibre nuclei, but still segment the other nuclei. In this particular example, i segmented with DeepCell and it got confused. Because if such powerful deep learning cannot do it, i thought i have to train illastik and do it manually on all images and go the long way. Please let me know if if something is not clear.

[Milad4849]

@akedarg
I see the issue. Deepcell is not able to handle such a special situation as it is not trained with similar data. The same manual method is also available via steinbock.