search

YangLabHKUST / UCS

Code for "UCS: a unified approach to cell segmentation for subcellular spatial transcriptomics"
2 stars 0 forks source link

Request to update Vizgen Cell Boundaries Polygon generation #3

Open rushin682 opened 2 weeks ago

rushin682 commented 2 weeks ago

Hello,

Thank you for a great repo and paper!

We noticed that for Vizgen aspect, creating of polygons (Line 140 - 152: preprocess/vizgen.py) uses the old cell boundaries h5 format. However, we have a parquet file for the cell boundaries. Is it possible to add a feature that can take the parquet file as input to create the polygons as well?

Thanks once again for your time. Looking forward to your response.

best, Rushin

chenyhvvvv commented 2 weeks ago

Thank you for your interest! Can you show me a public dataset to let me try on it?

devianggr commented 2 weeks ago

Hi,

Maybe I can help Rushin to share the public dataset with the cell_boundaries.parquet file. You can download the access dataset through this link: https://info.vizgen.com/protein-co-detection-showcase?submissionGuid=2e012b62-62b4-4a7e-97d4-cbfbd0a7a903

Best, Devi

chenyhvvvv commented 2 weeks ago

Hi, after trying and searching for reading parquet object of Vizgen, I can not find a proper package or code to read it in python. Directly using pandas can not work.

image

And there is little result of this and still an issue in https://github.com/satijalab/seurat/issues/7080.

But instead, you can use Cellpose to segment the MIP DAPI image. The code include prepare_dapi_for_cellpose in preprocess/vizgen.py to provide MIP image of mosaic_DAPI_z[].tif and segment using script like

python preprocess/cellpose_dapi.py --dapi ./data/vizgen_mouse_brain/mosaic_DAPI_mip_resized.tif --out_dir ./data/vizgen_mouse_brain