rlittman16
commented
10 months ago I would just randomly add 0 or 1 for the z for each spot. If you don't have a reference, you can still use the watershed segmentation.
JSTA doesn't do nuclei segmentation, we'd recommend Stardist or cell pose for that. If you have nuclei, doing watershed to get a counts matrix and then annotating and treating that as your reference may work, but I haven't tested that. There are many publicly available scRNAseq reference datasets you can use too.
Hello!
Thank you for the tool. I am interested in trying this tool on my own data. I have 2D image data and Stereo-seq data, but no cell type reference file. The image size is (1200,1200).
My question is how to use the only files I have to do JSTA cell segmentation. I looked at the example and found that the Stereo-seq data (['gene', 'x', 'y', 'MIDCounts']) was similar to spots, do I need to first apply the 2D image to the watershed algorithm to obtain the mask and get the nuclei(['id', 'x', 'y']) in the example. However, neither of the two files has a z value, and there is no whether it has an impact. If so, how should I determine the value of z? How to proceed after getting spots and nuclei to obtain segmentation results.
Can you give me some help?