Issues with cell identification

[lhastings19]

Hey everyone,

I'm having trouble getting initial spatial estimates with a low number of false negatives. While I've been able to find parameters that allow me to identify three or four neurons, I think that the true count is closer to 15-20. I'm also getting multiple ROIs per neuron that aren't consistently merged in later steps. I've tried to solve this problem with a wide range of parameter values (I did a grid search with more than 750 outputs), but what's shown here is the best that I could do:

I'm thinking that the problem has to do with the pixels in some of the neurons having a pretty low correlation. There are some cells that don't show up at all on the correlation summary image:

Looking at the raw data in ImageJ, a lot of these unrepresented cells have nice calcium transients, which makes me think that their temporal correlation should be high enough to be picked up. For example, here is the average pixel intensity over time in one of those cells:

These neurons are big and relatively synchronized, and I don't have that many of them in my field of view. Could any of those factors be making it harder to get accurate spatial footprints?

I would really appreciate any suggestions on how to move forward! Thanks in advance.