Sholl Analysis on mouse hippocampal Astrocytes,technical help needed

Dear all,

I hope someone can help me with a few questions I have about using the Sholl analysis plugin.

I am trying to run the Sholl on 3D reconstructed images of GFAP stained sections. As you see in the image, there is many cells per image, and sometimes they overlap.

  1. Can I run the analysis on images with many cells (can the software distinguish the end of the arbors even if it was overlapping with another cell’s arbors)? Or do I have to use the ROI for each cell and run it separately?

  2. What does the “number of samples” in the settings refer to? Does it refer to the number of cells inside radius?

  3. Can I run multiple cells at the same time using the ROI?

I need help and would really appreciate it!

Thanks in advance,


I will start first with 2): That question is answered on a dedicated section of the documentation.

The challenge is always how to resolve overlapping processes and how to segment cells without disconnecting the very thin processes. You mention 3D reconstructions: If you already reconstructed the cells, then yes. It is surely possible, but you will have to provide details on exactly how you reconstructed the images.

Otherwise, if you ask about running the plugin on multiple cells on the same segmented image: yes it is (in principle possible) by means of a macro or script, but always with the constrains of the single-cell resolution allowed by the staining. In other words, it may be very hard to resolve all the cells, but it should be possible to at least resolve some cells in the image, and perform Sholl on them.

I will not give you a one-button solution, but rather propose a possible workflow:

  1. Discard the very thin, and disconnected processes, since you can’t reliably assign them to a particular cell
  2. Retrieve the centroid of the larger, remaining cells
  3. Label each binarized cell obtained in 2. with a specific gray value
  4. Loop through all the labeled cells in 3. and call the plugin:
    1. Using each centroid as the focal point of the analysis.
    2. Thresholding the image using each label value

For 1. and 2. you could probably use IJ’s particle analyzer (Analyze>Analyze Particles). For 3. MorphoLibJ, for 4., any of ImageJ’s Scripting languages

1 Like

Thanks a lot for the quick response!

I used “stacks/ Zprojection / high intensity” to make the binarized images I’m working on. It was actually a mistake to say 3D reconstruction, sorry about that!

I have been trying the steps you suggested but since I’m relatively new to Imagej it seems it will take me some time to learn how to apply them, but I’m very glad to hear it can be done.

Thank you again!