Skeleton to a networkx graph conversion

[venugovh]

Hi @schlegelp !

I am trying to convert a skeleton to a networkx graph upon which I can then use many graph-based algorithms such as shortest path (by assigning the distance between nodes as edge-weights), etc.

Before, I dig deep into this, I wanted to know if this is already done or is there a straightforward way within skeletor to convert a skeleton to a networkx graph? Any input will be very helpful.

I am really enjoying this library by the way!

Thanks!

[schlegelp]

Yeah. Each skeleton has a .graph property that holds the networkx graph representation of the neuron:

>>> n = navis.example_neurons()
>>> n.graph
<networkx.classes.digraph.DiGraph at 0x162849f10>

It is automatically regenerated when the neuron is changed. The distance between nodes is used as weights:

>>> n.graph.edges[(2, 1)]  
{'weight': 34.64101615137755}

There are also explicit methods to (re-)generate the graph:

>>> n.get_graph_nx()
<networkx.classes.digraph.DiGraph at 0x1474c4050>
>>> navis.neuron2nx(n)
<networkx.classes.digraph.DiGraph at 0x162a5bf50>

For the record: the same exists for igraph as .igraph property, get_igraph() method and navis.neuron2igraph function.

[venugovh]

omg, this is a lifesaver! Thank you very much for the quick response!

[venugovh]

@schlegelp Sorry I closed this issue prematurely! I am able to see the graph in myskeleton.skeleton.vertex_graph. But when I called .edges[(a, b)], I am getting an attribute entity_index: 0 in the dict. I am looking for the node distance values as you mentioned.
Just a pre-cursor, I have an stl file read in trimesh and generated the skeleton using the wavefront method.
Am I missing something here?
Thanks!

[schlegelp]

Can you give a minimal code example? I have no clue what myskeleton.skeleton.vertex_graph is supposed to be.

[venugovh]

Sure, so sorry for the ambiguity!

import trimesh
import skeletor
mesh= trimesh.load_mesh('u_bend.stl')
skeleton = skeletor.skeletonize.by_wavefront(mesh, waves=1)
skeleton.show(mesh=True)
G = skeleton.skeleton.vertex_graph
print(G.edges[(1, 0)])

The STL file can be downloaded from:

https://drive.google.com/file/d/1YzNwCBoZ4l0L7E_ht0NOMVXqsSiG7PgR/view?usp=sharing

Thank you!

[schlegelp]

skeletor is totally independent of navis. That's on purpose to keep skeletor comparatively light-weight.
Your skeleton is an instance of the skeletor.Skeleton class which is not the same as a navis.TreeNeuron. The skeleton.skeleton is actually a trimesh.Path object cached for plotting. It just so happens that it to also provides a graph representation but it apparently does not hold distances as edge weight.

Bottom line: you have to convert your skeletor.Skeleton to a navis.TreeNeuron if you want to follow the example I laid out above. Fortunately that's very simple:

n = navis.TreeNeuron(skeleton.swc)
G = n.graph

G will hold the distance between nodes as edges. That said: not having looked at your mesh, you might not have an edge (1, 0) in your skeleton.

[schlegelp]

Apologies: I just realised that I misread your earlier question/thought I was in a different repo 🤦 (in my defence: it was fairly late in the evening).

To clarify: I'm talking about navis which I typically use to further process skeletons. So what I am recommending is install navis and then do this:

import navis
n = navis.TreeNeuron(skeleton.swc)
G = n.graph

[venugovh]

Thank you for the clarification.

Apologies: I just realised that I misread your earlier question/thought I was in a different repo 🤦

To clarify: I'm talking about navis which I typically use to further process skeletons.

No problem, thank you for the help! I will be sure to check out navis!