Tangram Deconvolution

[abhishekmaj08]

Hello,
I have a question about the deconvolution process. As I understand it, there are 2 steps: (1) segmentation by watershed algorithms that gives the cell_count and (2) deconvolution by alignment. I was wondering how well does Tangram perform when in step (1) the provided image is not DAPI? If the segmentation process does not perform properly, then that should affect the cell_count and in turn the deconvolution by alignment step right? So what should I use in the segmentation step in case of a non-DAPI image.

Thanks

[gaddamshreya1]

Hi @abhishekmaj08 ,

Thank you for your interest in Tangram!
Here are some other tutorials by Squidpy for segmentation that maybe useful for you:

• Segmentation of H&E images: LINK
• Cellpose Model for Fluorescence/H&E images: LINK
• Stardist Model for Fluorescence/H&E images: LINK

I would recommend trying these segmentation methods whichever fits your use case. However, you can also use any other segmentation methods, get cell count per spot and continue with deconvolution.