Baysor segmentaion stain

[lkmartin90]

Thanks for the great tool.
I have a question I can't find the answer to in the documentation. How do we know which stain Baysor is using to segment, and can we change this in the config file?
Sorry if I've missed this somewhere.

[quentinblampey]

Hello @lkmartin90, Baysor is not using any image for segmentation, it only uses the transcripts. Yet, Baysor can use a prior, e.g. we can run Cellpose on DAPI (or any other stain), and provide it to Baysor as a prior.

Does this help? Which technology do you use?
Let me know if you have additional questions

[lkmartin90]

Ah - sorry, my misunderstanding; I read that Baysor also used the stain for segmentation and didn't realise it was in the prior! I'm using CosMx, and despite a fairly good prior segmentation, I am struggling to get the Baysor output to look nice, mainly due to overlapping cells, even when I force it to work in 2D. However, this may just be our data.
Thanks for the help!

[quentinblampey]

Which prior_segmentation_confidence do you use? This Baysor parameter can be very important, and if you like your prior segmentation I recommend setting a high value, e.g. prior_segmentation_confidence=0.9.

Regarding the cells overlap, this happens mostly when using Baysor. I will soon improve this (once shapely==2.1.0 is released). You can either wait for this new feature, or you can also try other segmentation tools inside Sopa such as ComSeg.

[lkmartin90]

I'm using a prior confidence of 0.9. Thanks for the help, I'll also try ComSeg!