search

STOmics / CellBin

Tools for generating single-cell gene expression data
MIT License
26 stars 2 forks source link

Sub-optimal nuclei segmentation #7

Open paulacauhy opened 5 months ago

paulacauhy commented 5 months ago

Hi,

I was able to run the entire Stereocell pipeline, but when taking a look at the nuclei segmentation and the transcripts assigned to each cell, I noticed that many times, a few nuclei are being taken together as one cell. In the attached images you can see transcripts assigned to different cells in different colors. On the second image, we can see the nuclei in white and how for some of the cells (e.g. yellow on the right) more than one nucleus was taken into account.

Is it possible to change the parameters of the nuclei segmentation and/or filtering to try to improve this?

Thank you!!

Screenshot 2024-01-22 at 10 32 29 Screenshot 2024-01-22 at 10 33 40
niovl commented 5 months ago

Hi，

As a rule of thumb, the density of the cells and the exposure at which the image was taken will increase the occurrence of this condition. It can be alleviated by using the watershed algorithm. Of course, you can also use better third-party cell segmentation tools, such as cellpose and qupath. stereocell will upgrade the cell segmentation module in the future to alleviate this situation, but it is still difficult to completely solve it.