Local bifurcation angle computation fails when a section has only one point

[ssssarah]

https://github.com/BlueBrain/NeuroM/blob/master/neurom/features/bifurcation.py#L57

For the following bifurcation point:

bif_point [Section(id=30, type=2, n_points=1)<parent: Section(id=29), nchildren: 0>, Section(id=31, type=2, n_points=1)<parent: Section(id=29), nchildren: 2>]

local_bifurcation_angle fails because there is only one point.

See .swc file from https://bbp.epfl.ch/nexus/web/bbp/mouselight/resources/https%3A%2F%2Fbbp.epfl.ch%2Fneurosciencegraph%2Fdata%2Fneuronmorphologies%2F1661d3d0-87b3-4978-89f6-99a601b1b706

[crisely09]

Apart from the code to account for sections with one single point, @lidakanari we would also like to know if having only one point in the section after bifurcation is "acceptable" in a morphology, in terms of quality of the recostruction.
Thank you.

[adrien-berchet]

Hello,
Did you run MCAR on this morphology (at least the Curate workflow) before running NeuroM on it?
As far as I know, sections with only one point are not valid since a section must contain at least one segment and a segment is defined by two points (see https://morphio.readthedocs.io/en/latest/specification.html#segment).

[crisely09]

Hi @adrien-berchet ,
Thank you for your answer, that is what I was fearing, that the morphologies may have "issues" as they come from external databases.
We will run the MCAR on them, thanks for the documentation.

[adrien-berchet]

You're welcome.
And basically you should consider that all morphologies should go through MCAR before being imported (except if you explicitly want to keep the original version with its potential errors).

I let you close this issue if you don't have any other question.

[crisely09]

@adrien-berchet maybe the function should catch this error and inform the user that the morphology is not correct?
If you think this is reasonable that would be a nice thing to have, if not, we can close the ticket.

[adrien-berchet]

I think all functions should be able to consider that the morphology is valid. Nevertheless, we may add a general try/except block to give some hints to the user when the computation fails (usually because of an invalid morphology).