Replies: 2 comments 3 replies
-
This is a great question. Four is not very high. I tend to think this will happen when you set it really high, like 50. I wonder what this plot will look like if you go from 1-50 (in like 5-10 step increments). I'd be really curious what the background spatial components look like in those cases. To my knowledge this hasn't been systematically explored (though I'm sure Eftychios and Andrea looked at it in detail, it's just not published to my knowledge: @j-friedrich may have looked at this too). I personally have not played with the background number parameter ( It's tricky to evaluate plots like yours because you'd ideally want to make sure you were looking at the same neural component across comparisons, and as You are looking at gross statistical changes, and it might be natural to think that the loss of signal should just show up as decreased SNR, but it may be more subtle than that. For instance, if you pick one component and plot it for different Edit: I'm speculating about things that might happen. As I said I haven't explored this, so would be very curious to see. 🚀 |
Beta Was this translation helpful? Give feedback.
-
Looking at the reconstructed background movie, (You should really be looking at the reconstructed movie and residuals as well for quality inspection of results) |
Beta Was this translation helpful? Give feedback.
-
In four different iterations of CaImAn analysis of the same t-series, I varied the number of background components from one to four. However, when I thereafter calculated the SNR from the iterations, they all had roughly the same SNR. This doesn't make sense to me since it was written in the documentation that a too high number of background components will absorb signal and decrease SNR. Anyone have any idea why my result looks like this?
Beta Was this translation helpful? Give feedback.
All reactions