create a squidpy or R object using baysor output

[shrutikhare-git]

Hi, I want to use baysor segmentation on my CosMx data and then visualize it in R/Python. CosMx usually outputs 5 files which are used by Seurat/Squidpy to create the object - metadata, tx, polygons, fov positions, exprMat. Can you please guide - how to reformat/create these 5 files to incorporate baysor segmentation?
Thanks!

[seigfried]

I tried running a Baysor and made my own files. However I still get this error

Error in data.frame(x = md$CenterX_global_px, y = md$CenterY_global_px, :
arguments imply differing number of rows: 0, 85911

I compared all columns and realised I did not have CenterX_local_px, CenterY_local_px, CenterX_global_px, CenterY_global_px in my metadata
I tried to manually calculate this by taking the information from a polygons.csv of a different experiement and then calculated the centroid i.e. x1+x2+...+xn/n
The results were close but not exactly the same as the CenterX_local_px, CenterY_local_px, CenterX_global_px, CenterY_global_px given in the metadata

So yes, my questions is also how I can convert Baysor output to say, a SquidPy or Seurat object

[VPetukhov]

The names are different between Baysor output and CosmX. segmentation_cell_stats.csv contains cell features, particularly x, y and z columns contain cell centroid coordinates. So, I'd suggest to simply reading segmentation_cell_stats.csv and setting it as Seurat metadata.

[shrutikhare-git]

I reformatted all baysor output to match CosMx output formats which involved calculating the centroids as well. But after loading the object in R, the cells look weird. Any ideas about what may be causing this? Thanks!