3d skeleton from z-stack of binarized images

I have 10 10x10 um binarized images recorded in 0.1 um steps (so the volume is 10x10x1 um). The images are of entangled fibers. I want to generate a 3d skeletonization of these fibers (branch structure from which to calculate branch lengths, number of branches, porosity, tortuosity, etc.).

How do I do it? Thanks in advance.

[Iris]

Hi, I have a very similar problem, working with 3D TiFF files of cytoskeletal structures which I would like to classify by shape (whether they are curved or straight) and length. They are separated objects, so I won't have to worry with nodes. Still I need to skeletonize, which I have not figured out how to do. I have done a distance transformation, but would need to perform a non-maximum suppression or something similar now before performing the Hough Transformation. How did you work out the 3D skeletonization back then? Many thanks in advance!

Answer 1

[Image Analyst]

Wow. Not easily. Tracking fibers in 3 dimensions is not trivial. Of course you can skeletonize but where fibers touch there will be a connection, and you don't want one there. We've done it, but it's very complicated and not just a short demo we can upload here.

[Rory Staunton]

It seems I should first generate a 3d array of voxels, input the known values from the images in the correct locations, and then interpolate the values of those in between. From there, I could at least generate an isosurface. With an isosurface I guess I could begin to skeletonize by 'shrinking' the surface until nothing is left but nodes and branches? Let me know if you suggest a different approach.

[Image Analyst]

Not sure hwy you need an isosurface or why you need to interpolate.