A lot of these tools are written over 2015 - 2017 mostly when I was a part of 3Scan as an intern. Thanks to Matthew Goodman, Mike Pesavento, Jenny Folkesson, Ted Blackman, and Joanna Lipinski for their collaborations and ideas.
This repository contains programs needed to obtain a 3D skeleton and quantify the skeletonized array to statistics.
Input must be a binary array with z in its first dimension.
Skeletonization on a 3D binary array is performed by iteratively removing the boundary points until a line in the center is obtained by convolving the image with structuring elements from the paper.
This function is implemented using cython for fast execution and pyximport is used to automatically build and use the function
Using mayavi is a useful way to visualize these test stacks
Install mayavi via:
conda install -c menpo mayavi=4.5.0
This version works on python 3
Versions < 4.5 don't work on python 3.x
mayavi also requires that QT4 (pyqt) is installed
This may be in conflict with matplotlib >=1.5.3, which started to use qt5. So, use mayavi and matplotlib in a separate conda environment to avoid conflicts
View the stack contours:
import mayavi as mlab
mlab.contour3d(anynpynonbooleanarray)
mlab.options.offscreen = True
mlab.savefig("arrayName.png")
To validate two networks after conversion of skeleton stack into obj files use NetMets - Software wrote and maintained under UH STIM Lab by Dr.Mayerich, I wrote a part of it while I was pursuing my masters at UH.
Github code
Build guide
After installing NetMets, validate skeleton against groundtruth
netmets objfile1 objfile2 --sigma 3