"Select smaller region for testing" Tool Only Accepts Square and Rectangular Selections

[austinlandgraf]

Describe the bug
I'm trying to use the "Select smaller region for testing" tool under the "Spot Detection" tab to only analyze one layer of cells surrounding a blood vessel (roughly a circular selection). When I attempt to select this region using the "Circle" tool and then click "Return selection to FISH-QUANT", I get the error message in Kaibu, "Annotation has to be a Polygon." When I then try the "Polygon" tool, I get the error message "This is not a rectangular selection". The only two tools the seem to work are the "Rectangle" and the "Square" tools. Are these really the only supported selection methods in Kaibu?

To Reproduce

1. Upload the example data
2. Spot detection: select channel and then click "Select smaller region for testing"
3. Under "Properties" in Kaibu popup, try selecting a region within the image using either the polygon tool or the circle tool (any tool that is not the rectangle or square tool).
4. Click "Return selection to FISH-QUANT" --> error message will popup

Logs
browser console log.docx
plugin log.docx

Screenshots

Versions:
iOS 11.6.2
Chrome Version 99.0.4844.51
FQ plugin: v0.0.16
FQ interface: v0.0.11
big-fish: v0.6.0

[muellerflorian]

Hey Austin,

unfortunately, there are no other region selections available at the moment. The purpose of this selction is to provide a cropped image to test the detection parameters faster, before you then analyze the entire image.

I assume that you would like to restrict your analysis to sub-region (e.g. your blood vessel). while this was possible in the old version of FISH-quant (Matlab), we took a different approach here. in the underlying Python library (big-fish), the entire image will be analyzed. In a post-processing step (as you already know), these results can then be assigned to certain regions (cells, nuclei).

is this approach not working for you?

[austinlandgraf]

Hi @muellerflorian - thank you for the clarification. that's very helpful!

It seems like the cell numbers are assigned from the top of the mask down. So if I wanted to restrict my analysis to just cells in a circular region, I might have to only include say cell 5, 8, 23, 43 etc. Is there a way to display the cell_id number on the mask so that one could easily figure out which cells to include or exclude for region-restricted analysis? Right now, it's difficult to tell which cell in the mask corresponds to say cell_id #10.

[muellerflorian]

i guess this could be done with a script, where we plot the segmentation results and on top the cell id (or nuclei id). would this work for you?

@Henley13, do you have a script that does this kind of plots? if not, I can try to put something together.

[Henley13]

I don't have a dedicated script for that, but if you show the segmentation mask in Fiji, Kaibu, whatever, I guess you can see the value used for each cell mask by moving the mouse over the cells, no ?

[muellerflorian]

Ok. I don't think an inspection with FIJI like this is very practical. I will check and see if I can find some old code. For another project, I created plots like the one shown below. would this work for you, @austinlandgraf ?

@Henley13. do you think we could/should add this to big-fish? might be useful for others as well in order spatially identify the cells/nuclei based on their index?

[austinlandgraf]

Hi @muellerflorian @Henley13 A plot like the example you showed would be very helpful - thank you for working on this!

[muellerflorian]

Hey @austinlandgraf,

I added a new plugin that should help. you can install it

with this link

this should give an image like that

i also updated the documentation, but somehow it is currently not correctly build. what you need to do in the plugin

You only have to paste the path to folder containing the segmentation results (usually a folder called segmentation-results). The plugin will then loop over all files in this folder, and select the onces fitting the other two parameters:

• File string: specify a string that the file-name has to contain. The default mask__ fits for both segmentation of cells and nuclei.
• Img extension: file-extension of the label images. The default is .png`, which is the extension used by our segmentation pipeline.

The plugin will then create a new subfolder called labels_numbered where the numbered label images
will be stored under their original file-name with an added suffix __numbered.png.

Let me know if this works for you!

best
Florian

[austinlandgraf]

Hi @muellerflorian , the plugin works perfectly! Thank you for you releasing this new feature! I think it's going to be very helpful for analyzing images.