Glial cell cross section representation

[eleftherioszisis]

Currently, node cross section is assumed to be a circle, which does not convey enough information in order to represent glial cells.

According to Dan it would be sufficient if we could add one more parameter in the value array for each node which will reflect the real perimeter of the cell which will be imported from the file. The area of the cross section can be represented by a circle with equivalent area.

To summarize an extended value list is required:

(x, y, z, radius, id, type, parent_id, real_perimeter)

where x,y,z is the center of the cross section, radius a value that provides the same area as in the cross section and real_perimeter the perimeter of the node's cross section which will be loaded from file. This extends the existent dataformat by one value.

Maybe we could have a glial cell class and dedicated features for glial cells in the features module. All features that apply on neurites and neurons can still be applied due to the same positional format.

[juanchopanza]

I can't really say how easy or disruptive this would be (but I suspect it would be problematic, as neurom is designed around the SWC format.) I think we can re-consider this when a data format specification and data samples become available.

[dkeller9]

Since we do not control the SWC format, it is impossible to change that standard on our own. I would be happy to update our h5 specification to reflect the desired glia cell additions, though. Would this be sufficient?

[lidakanari]

I think it would be preferable to create an h5-glia specification (if the data are provided in h5 of course) and then have an swc-glia equivalent representation. We don't need to change the current specs but based on them create an extended version that includes the appropriate modifications for glia cells. Does that sound feasible?

[dkeller9]

Okay, I'll do so. Basically, it just involves associating an extra number with each point in the morphology to represent perimeter of the cross section.

[juanchopanza]

I would imagine we would need a new spec, which could reference the current ones to avoid duplication. But better consult with @mgeplf and @jdcourcol.

[mgeplf]

Not sure there's anything I can add to this, as I don't know anything about it.

[asanin-epfl]

@eleftherioszisis please feedback here otherwise I will close the issue.

[asanin-epfl]

Closing. In NeuroM v2 the reading is on MorphIO. MorphIO yet can't read glial cells but the PR is in progress BlueBrain/MorphIO#256. Please consider open an issue/PR there for any additional glial functionality.