Set pos 3d fixed

[alejoe91]

Cell.set_pos() fixed:

• 3d segments are moved differentially wrt to current soma location
• start and end of segments are moved anyways (deleted if-else)

[espenhgn]

@alejoe91 I cannot reproduce the problem you're seeing. I set up these test files:

#!/usr/bin/env python
'''test Cell pt3d arg'''
import LFPy
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.collections import PolyCollection
import neuron

neuron.h('forall delete_section()')

cellParameters = {
    'morphology' : 'example_morphology.hoc',
    'pt3d' : True,    
}
cell = LFPy.Cell(**cellParameters)

xzpositions = [0, 0, 10, 20]

########################
# Figure
########################
fig, axes = plt.subplots(1, len(xzpositions), figsize=(16, 9), sharex=True, sharey=True)
for i, (ax, xzpos) in enumerate(zip(axes, xzpositions)):
    if i != 0: # is soma position in (0,0,0) after creating the cell?
        cell.set_pos(xpos=xzpos, ypos=0, zpos=xzpos)

    print cell.somapos

    zips = []
    xz = cell.get_idx_polygons()
    for x, z in xz:
        zips.append(zip(x, z))
    polycol = PolyCollection(zips,
                             linewidths=(0.5),
                             edgecolors='k',
                             facecolors='none',
                             zorder=0)
    ax.add_collection(polycol)

    if cell.pt3d:
        for x, z in cell.get_pt3d_polygons():
            zips.append(zip(x, z))
        polycol = PolyCollection(zips,
                                 linewidths=(0.5),
                                 edgecolors='r',
                                 facecolors='gray',
                                 zorder=-1)
        ax.add_collection(polycol)

    #morphology mid points
    ax.plot(cell.xmid, cell.zmid, 'o', mec='none', mfc='k',
             markersize=3, zorder=0)
    
plt.show()

and

/* ----------------------------------------------------
example_morphology.hoc

This hoc file creates a neuron of the following shape:

            \       
             \     
              \   /
               \ /
                V
                |
                |
                |
                O
                
Note the conventions:
 - soma needs to be a list (soma[0], not soma),
 - use soma for the soma compartment,
 - use a name starting with dend for the dendrites.
-----------------------------------------------------*/

create soma[1]
create dend[3]

soma[0] {
    pt3dadd(0, 0, 0, 5)
    pt3dadd(0, 0, 15, 25)
    pt3dadd(0, 0, 30, 5)
}

dend[0] {
    pt3dadd(0, 0, 30, 5)
    pt3dadd(0, 0, 150, 2)
}

dend[1] {
    pt3dadd(0, 0, 150, 2)
    pt3dadd(-50, 20, 200, 1)
}

dend[2] {
    pt3dadd(0, 0, 150, 2)
    pt3dadd(30, 0, 160, 1)
}

connect dend[0](0), soma[0](1)
connect dend[1](0), dend[0](1)
connect dend[2](0), dend[0](1)

With the current master branch the soma positions are correct ([ 0. 0. 0.], [ 0. 0. 0.], [ 10. 0. 10.], [ 30. 0. 30.]) for the different x,z-offsets, as well as pt3d and segment coordinate info:
test_pt3d_fix_master.pdf

Notice on the first iteration the cell.set_pos function is not used, but the cell should default to (0,0,0) anyway.

However on your pull request the cell's soma's center position is no longer in (0,0,0) and pt3d info and segment coordinate info do not overlap as they should:
test_pt3d_fix_alesioe91.pdf

[alejoe91]

Hi,

I was wrong in deleting the if-else, because the start, end, and midpoints are updated in _update_pt3d (which calls _collect_geometry()). So I was basically moving them twice.

I think that the other fix (passing diffx, diffy and diffz to _set_pt3d_pos()) is correct, though. In fact, if you set_pos(50,0,0) for three times in a row:

• the first time the cell is correctly moved to (50,0,0)
• second time to (100,0,0)
• third time to (150,0,0)
  and that's because in _set_pt3d_pos():

def _set_pt3d_pos(self):
for i in range(len(self.x3d)):
self.x3d[i] += self.somapos[0]
self.y3d[i] += self.somapos[1]
self.z3d[i] += self.somapos[2]
self._update_pt3d()

so if the soma position hasn't changed, the 3dpoints are moved anyways. Do you agree?

I just committed a corrected version (commit 'reinserted if-else, kept diffx, diffy, diffz')

[espenhgn]

@alejoe91 I was mistaken above: the last position was (30,0,30) on the master branch when it should have been (20,0,20). Sorry about that.

However, I reran now with your code, and the cell position is not correctly instantiated to (0, 0, 0) when pt3d==True. For the simplified geometry above the result is (0,0,-15), however if pt3d==False the result is (0,0,0). The issue with non-overlapping geometry appears fixed though.

[alejoe91]

@espenhgn
The first time the cell position is set so that the end of the soma is at (0,0,0) and I don't know why actually. It might depend on the .hoc file.

Anyway, try the new version (after the first time the soma and the segments should always in the correct position) and let me know :)

[espenhgn]

There is a call to self.set_pos() immediately after calling self._collect_geometry() in cell.py line 210. An ugly fix would be to call self.set_pos() twice but I'm sure there are better ways.

[alejoe91]

I saw that, that's why I was quite surprised to see that the first time it doesn't set the position correctly.
I'm sure there are better ways as well, but for now we could keep it like this.
If you need to move your neuron it works, otherwise the soma ends at (0,0,0), which is not such a displacement.

[espenhgn]

For these morphology files the root of the morphology (the 1st data point for the first section, typ. the soma) may be some arbitrary location in 3D space. For consistency between different cell models in LFPy we made a decision that the "default" cell soma position should be having it's mid point at (0,0,0), not one of it's end points. That should also be the case if the soma.nseg > 1.

[alejoe91]

Ok, then I will try to dig deeper and see where it gets displaced.

[alejoe91]

@espenhgn
I think I found it!
in collect_geometry first you compute the start, end, mid points, then the somapos, so the first time the somapos is not (0,0,0). In set_pos(), then since the displacement is computed relative to the soma, diffx, diffy, and diffz are not 'correct'.
Easy fix: before the initial set_pos():

self.somapos=[0,0,0]

What do you think about it?

[espenhgn]

I think that should work for now. Thanks for your help!