Summary
When get_sample_pts_euclidean samples points along an SWC edge, the interpolation appears to place each point at the mirrored location on that edge. In a simple 3-node chain, this changes the actual distances from the root returned by the sampler, not just their order.
I’m including:
Minimal reproducible example
- Save the following as
demo.swc. The example neuron has 3 nodes on the x-axis at x = [0, 3, 10].
# id type x y z radius parent
1 1 0 0 0 1 -1
2 3 3 0 0 1 1
3 3 10 0 0 1 2
python sampling_with_demo_swc.py
expected: [ 0. 2. 4. 6. 8. 10.]
got : [0. 1. 3. 5. 7. 9.]
Potential fix
In get_sample_pts_euclidean inside src/cajal/sample_swc.py, interpolation currently uses:
(root_triple * x) + (child_triple * (1 - x))
If sampled points are intended to be measured from the parent/root along the edge, the weights seem swapped. A possible change would be:
(root_triple * (1 - x)) + (child_triple * x)
sampling_potential_fix.py demonstrates the output before and after swapping these weights, again using the 3-node example above.
Summary
When
get_sample_pts_euclideansamples points along an SWC edge, the interpolation appears to place each point at the mirrored location on that edge. In a simple 3-node chain, this changes the actual distances from the root returned by the sampler, not just their order.I’m including:
read_swc+get_sample_pts_euclidean)Minimal reproducible example
demo.swc. The example neuron has 3 nodes on the x-axis atx = [0, 3, 10].Potential fix
In
get_sample_pts_euclideaninsidesrc/cajal/sample_swc.py, interpolation currently uses:If sampled points are intended to be measured from the parent/root along the edge, the weights seem swapped. A possible change would be:
sampling_potential_fix.pydemonstrates the output before and after swapping these weights, again using the 3-node example above.