Adding cell-volume size factor normalisation for spatial

[wlason]

This paper suggests that normalisation using cell volumes as size factors (instead of library sizes) for data with good segmentation results and scTransform for data with poor segmentation results performs better for spatially resolved data compared to single-cell lognorm from Scanpy.

https://doi.org/10.1186/s13059-024-03303-w

This is the summary of the considerations the Authors propose:

I was hoping we can add these alternative normalisation methods to panpipes spatial preprocess workflow.
I think the scTransform is already added as norm_hvg_flavour: seurat, but the volume-based normalisation is only available in R.

[wlason]

Suggested way to do this:

X = adata.layers['raw'].copy()
Xnorm = np.divide(X.todense().T, adata.obs['cell_area'].values).T * 1000
# logtransformation
Xnorm = np.log1p(Xnorm)

# convert back to a sparse matrix.
from scipy import sparse
adata.X = sparse.csr_matrix(Xnorm.copy())

Taken from the ELIXIR Spatial Transcriptomics Course.