Regarding number of cells per sample and cell area

[jai-mathur]

Hi @dcjones, I have been using proseg for CosMx data, and had 2 questions I was hoping you could answer.

1. Is there a reason why proseg reduces the number of cells in every Cosmx sample? I believe proseg takes the prior segmentation and then run the simulations, so what leads it to reduce the number of cells (do cells generally just merge into each other, which got fragmented, or cells which aren't cells just get removed?) I see a consistent difference of around 10-12 percent
2. I also notice when I make log1p cell area plots, there is a difference ther as well, where the cell.areas are higher after proseg. Is it still true that proseg will expand the boundaries of cell until it hits other cells? I have seen some expansion in my Xenium analysis (with 3.1), and I was wondering which cases (densely packed, sparsely packed) and what parameters control this boundary expansion?

As always, thank you for your hard work on this!

Warm regards

Jairaj Mathur