Simple neurite tracer

simpleneuritetracer

#1

Hello
I am trying to measure neurite outgrowth in neurospheres (like this)
I can do each one individually but it is very time consuming!!!
Does any one know of a macro - that will allow simple neurite tracer to automatically select all neuritis?
thanks!!!

msh%2020%208%20bit%20|nullxnull


#2

Seems like a job for machine learning. The recent #CARE approach by @fjug comes to mind.


#3

Is this data 2D or 3D? I’m not sure if a simple hack will lead to satisfactory results…
#CARE might help you get the neuritis a bit crisper in their appearance, but you still need a good filament detector or such that can tell multiple of them apart when they get dense and grow on top of each other.


#4

Welcome to the forum @smom,

The IJ2 re-write of SNT (currently in beta-status, available through the Neuroanatomy update site) implements a Scijava Service - SNTService - that allows you to programmatically access the entire SNT API during an interactive tracing session, thus allowing you to blend automated workflows with manual curation steps. It also allows you to run scripts (in any of the IJ2 supported languages) from within SNT.

If you subscribe to the Neuroanatomy update site, you will find a template script (in the Script Editor under Templates>Neuroanatomy>Scripted_Tracing_Demo) (source code on GitHub), that you can use as a starting point:

You could adapt it to perform the following:

  1. Segment and skeletonize the processes of the neurosphere. Do not worry much about the quality of the segmentation near the surface of the sphere. It is only important that the distal end-points of the skeleton reflect the end-points of the extended neurites
  2. Retrieve the coordinates of all the end-points of the extended dendrites
  3. Manually define the center of the sphere (or programmatically from the centroid of the sphere)
  4. In a loop, use SNTService to perform auto-tracing between the center of the sphere and each of the end-points. Don’t worry about how the A* search will perform within the sphere core, again, you will discard that result later on
  5. Loop through the obtained paths and delete all the nodes with coordinates contained by the sphere core.
  6. Activate the new Edit Mode of SNT and perform any proof-editing as needed.

There are some inefficiencies in the workflow above, but on modern computers they are likely negligible. Depending on how consistent are your images, all the above could be performed with minimum human interaction.

Do let us know if you need help with the above,