search

MouseLand / suite2p

cell detection in calcium imaging recordings
http://www.suite2p.org
GNU General Public License v3.0
354 stars 241 forks source link

Manual Labelling issue - no deconv trace #577

Closed xwchia closed 3 years ago

xwchia commented 3 years ago

Hi,

I've been attempting to manually label some ROIs that have been wrongly labelled (the ROI is larger than the actual cell).

However, i noticed that there are some ROIs that do not display a deconv trace despite the cell being obviously present.

Do refer the the screenshot below. Note that the ROI selected is an ROI that has been manually labelled by me. Screenshot 2020-11-10 at 4 59 46 PM

Wondering if anyone has any suggestions or idea why this is happening?

Thank you!

marysethomas commented 3 years ago

Hello, I am having the same issue with manually labelled cells. I am using the dev version of Suite2p.

The deconvolution does not get performed for a small subset of the manually labelled cells, let's say 3 out of 10. When I load the fall.mat in MATLAB the spks values for those cells are all "NaN"

Thanks for opening this issue @xwchia!

carsen-stringer commented 3 years ago

Are you running sparse_mode? The shape of the ROI is determined by the frames on which it's active so it shouldn't be a too unconservative estimate.

My guess for the NaN's is that you chose a cell that was fully overlapping with another cell and ops['allow_overlap'] is set to False, so all those pixels were excluded => no pixels in cell => F is all zeros (as in plot) => NaN's in spike detection. If you want to draw overlapping ROIs you need to set ops['allow_overlap']=True.

carsen-stringer commented 3 years ago

I'm going to close this issue for now but please comment if this approach fails

xwchia commented 3 years ago

Hi Carsen,

Thank you for the reply. My ops['allow_overlap'] is indeed set to False. And as shown in the photo, you are right that the ROI that i attempted to draw was fully overlapping with another ROI. I'll update again if i encounter the same issue after changing ops['allow_overlap'] = True.

If i were to set ops['allow_overlap'] = True, would that change the result of suite2p? For example, would the F change for other cells?