Unable to load swc file from simple neurite tracer about navis

[schlegelp]

It looks like the file is a mix of tab and space separators. See this:

>>> # With default separator
>>> pd.read_csv('VFB_00023647_snt_skel-000.swc', header=None, skiprows=6)
                                                      0
0     1\t0\t231.908032 219.526534 88.631724 0.000000...
1     2\t0\t232.426842 219.007559 88.631724 0.000000\t1
2     3\t0\t232.945652 219.007559 88.631724 0.000000\t2
3     4\t0\t233.464462 219.007559 88.631724 0.000000\t3
4     5\t0\t233.983271 219.007559 88.631724 0.000000\t4
>>> # With tab separator
>>> pd.read_csv('VFB_00023647_snt_skel-000.swc', header=None, skiprows=6, sep='\t')
         0  1                                         2     3
0        1  0  231.908032 219.526534 88.631724 0.000000    -1
1        2  0  232.426842 219.007559 88.631724 0.000000     1
2        3  0  232.945652 219.007559 88.631724 0.000000     2
3        4  0  233.464462 219.007559 88.631724 0.000000     3
4        5  0  233.983271 219.007559 88.631724 0.000000     4

Bottom line: that's an issue with pandas (or the underlying Python csv parser) not being able to disentangle the mix of separators. Is there a way to change the output of the simple neurite tracer? Alternatively, you could write a little script to clean up the file(s). navis.read_swc has a delimiter parameter which defaults to ' ' whitespace and is passed directly to pandas.read_csc.

[schlegelp]

Ah: a workaround is using a regex separator.

pd.read_csv('VFB_00023647_snt_skel-000.swc', header=None, skiprows=6, sep='\s+')
n = navis.read_swc('VFB_00023647_snt_skel-000.swc', delimiter='\s+')

[schlegelp]

By-the-by: it struck me that the neuron has one one root and one end node - no branch points. Did you realize that this is one continuous line? I.e. when a neurite "ends" the skeleton just turns around, forms a loop and tracks back.

[SridharJagannathan]

Thanks, I just realised that the file seems to have arbitary number of spaces as delims and your solution to set the seperator as \s+, seems to have done the trick.
The neuron (tracing) actually looks fine when i plot it see below:

Perhaps I need to reroot it, somehow to produce the branch points?
I was just trying to do some covisualisation of gal4line, hemibrain, flywire neuron ( ilPN) as an example here:
https://github.com/SridharJagannathan/pyroglancer/blob/master/pyroglancer/examples/gal4line2flywire.ipynb

[schlegelp]

It will plot just fine but anything that requires correct topology will fail or produce nonsensical results.

[SridharJagannathan]

I tried to cut the neuron and then stitch them back together, that seems to have produced branches, however i can see that there are loops inside the neuron, do you know how to eliminate them, basically i'm thinking of an approach to use the pts to construct a graph, and then use the graph to construct the neuron.

vfb_neuron = navis.read_swc(swc_files, delimiter='\s+')
fragments = navis.cut_neuron(vfb_neuron, 500)
stitched = navis.stitch_neurons(fragments, method='ALL')

[schlegelp]

With 84cfe9f I just pushed a first pass implementation for a function that resolves this (albeit not perfectly):

>>> n = navis.read_swc('VFB_00023647_snt_skel-000.swc', delimiter='\s+')
>>> fx = navis.graph.clinic.merge_duplicate_nodes(n)
>>> fx
type                     navis.TreeNeuron
name            VFB_00023647_snt_skel-000
n_nodes                              1422
n_connectors                         None
n_branches                             59
n_leafs                                61
cable_length                  1106.133057
soma                                 None
units                        1 micrometer

You can see it now has proper leafs and branches. However when co-visualizing it, you will notice that some edges got removed (red) to break up cycles after node collapse. It's better than before but not perfect.

[schlegelp]

I'll close this issue. Re-open if you feel this is unresolved.