Solved: Sholl Analysis Error: False Intersections

Dear all,

I am trying to perform the Sholl analysis of skeletonised microglia on Fiji. However, as shown below in red arrows, the plugin falsely recognised intersections at quite a few spots where no signals from microglia processes could be seen. This error always appeared at edges of the image and along a straight line from the defined centre of the soma. I have got the same issues using one of the Fiji’s sample images of rat neuron without skelenisation. Thus, this should be a common problem of this plugin.

I would appreciate it if anyone could let me how to avoid this problem or eliminate these false intersections from the marked image.

Thank you very much indeed.

Tatsuya

Hi @tmanabe, Welcome to the forum!

The plugin always considers foreground pixels to have a non-zero value in binary images. In your example, those are rendered in white (while the skeletonized cell is rendered in black). The following example should explain this better:

On the left is your image, on the right its inverted counterpart (with inverted LUT, so that both images look the same) NB: You can test this by yourself by running “Image>Invert”, and hovering the cursor over the image canvas while reading pixel intensities in the IJ status bar).

I have written about this in the past, and the documentation has a dedicated section on this with a similar illustration. Also, note that confusing the phases of a binary image is not really specific to Sholl Analysis, but impacts any other operation performed in ImageJ. It is disheartening to see it keeps being a source of confusion for Sholl, and especially frustrating to see it in published literature.

So I really ask you (and whoever reads this) to please suggest improvements to the plugin’s UI on how to avoid this in the future. One way would be to force users to threshold the image using IJ’s “Threshold Widget” so that at least the pixels to be analyzed by the plugin would appear highlighted in red (the plugin already requires this for grayscale images, but not binary ones).

1 Like

Thanks for your swift response! Indeed, once I inverted the background and foreground pixels by “Edit > Invert,” I was able to sort out this problem.

This plugin has already forced to threshold the images beforehand. I hold this issue has popped up because I thresholded my image by choosing the over/under display mode. When I selected the red display mode with the dark background ticked, I have not got this problem without inverting the pixels.