comments on readme and example run through about imcsegmentationpipeline

[votti]

Hi, thanks a lot for your helpful comments!
Any comment helps to a) show me where to improve things b) motivate me to do allocate time to do so.

To 1. Really good point! I think I did not add it at the time as I aimed to add a really well-trained classifier but then, in the end, forgot to do so... I will add it to the open issues.

2. While it is mentioned in the initial setup instructions, this is really a common problem. So I will add a comment to the section where running the CellProfiler pipeline is mentioned first.

3. I will take that comment as a suggestion to better specify describe what individual points in the ### Input folders (Needs to be adapted for use) section actually means (and give this section a more meaningful name.
   Note that in case of new zipped .mcd files, it is also crucial that the user needs to adapt the pannel.csv file, such that the right channels are selected for the ilastik stack/analysis stack.
   Would it maybe be a good idead to add a section How to adapt this pipeline to your own data?

Thanks again for your feedback and let me know you see any other issues!

[glynnelson]

Yes, I think a section on 'How to adapt this pipeline to your own data' would be very useful. I hadn't actually realised that Ilastik was using the panel.csv file for selecting the channels for training, so that would be useful to point out too. Work has got in the way of continuing my testing, but I'll put up more comments as I work my way through when I get some free time.

[glynnelson]

Hi,

Got through to Sections C and D for applying the masks from Ilastik probs. A couple of issues.

1. Minor one is that the probabilities.tiff files on my system were called .tif, not .tiff, so I had to edit the cppipe files to reflect this. (Ilastik on Mac uses .tif suffix instead of .tiff?).

2. Second one is a question/ query- I am trying to run the measure_mask_basic.cppipe but I get an error when opening it: 'Error while loading None: Unknown mat file type, version 105, 116' and it won't load. I don't have matlab installed, and am unsure why it says this.

[votti]

Hi @1.: I think you selected '.tif' instead of .tiff when configuring Ilastik to output the Probabilities- unfortunately .tiff and .tif are valid filenames so Ilastik offers both.
@2: Hm I never saw this one/nobody reported that problem - I will need to investigate where this could come from. In the end the pipeline is not using mat files and also does not require Matlab at any point, so this is a really strange error indeed :/

So Cellprofiler completely refuses to load the pipeline or just throws the error and opens it anyway?
Are you using Cellprofiler 3.1.5? Have you tried to restart CellProfiler between the pipelines?

Thanks again for your reports! They are really helpful.

[glynnelson]

Re1. Yes, think the tif choice was the error. I don't know if you can have tif* as an option for the selection pf filenames by the name probabilities.tif*, or if CP would find the files if they were named either tif or tiff if it had probabilities.tif as the selection criterion.

Re. 2: I have CP 3.1.8 and it won't open the cppipe after the error. I had tried restarting CP, as I often am unsure if ti resets everything. So could be a version error.I'll try recloning from git and see how i get on.

[votti]

Hi,
I now re-checked the measurement basic and noticed that it actually contained still things related to compensation, which I now removed. I also added the cpproj files in case the cppipe still doesnt work.
Let me know if this actually works now!

Further I tried to incorporate your suggestions for documentation. Hope it made things more clear!

[glynnelson]

HI,

I'll take a look- this is exactly what I ended up doing myself yesterday- I took the compensation cppipe and adapted it to remove the compensation steps. I had previously tried recloning from git and also opening the cppipe directly from git via import from url option, but got the same error each time. I'll have a look at the new one too though, plus the updated documentation and give you any feedback.