Fiber Populations and Trajectories

Fiber Populations & Trajectories

golgi traces curved, fascicle-following fiber trajectories through the real 3-D nerve — including through bifurcations — and seeds them with biology-realistic diameter populations. This is what enables branch-selective analysis that straight-fiber, cross-section-only tools cannot do.

Source: golgi/compute/solve_fiber_paths_nerve.py (trajectory solver), golgi/pipeline/fibers.py (driver), golgi/state_defaults/pop_presets.py (populations), golgi/figures/population.py (previews).

---

How trajectories are generated

Fibers follow the natural current-flow topology of the nerve, obtained from a Laplace solve on the endoneurium volume (a nerve-only tetrahedral mesh, built independently of the cuff so the result is the same wherever the cuff sits):

1. Laplace solve. Solve ∇²φ = 0 on the endoneurium with insulating lateral walls, injecting current at the trunk cap and draining it equally across the branch caps (a pinned interior vertex fixes the constant).
2. Cap detection + branch classification. Caps (trunk and branch ends) are found in the nerve's PCA frame: keep facets whose normal is nearly axial, restrict to a z-band near each extremum, then cluster by cross-sectional proximity and drop clusters that are too small. Each branch is a cluster.
3. Streamline integration. Seed points across the trunk cross-section are integrated along the field −∇φ from trunk toward the branches, producing smooth trajectories that respect the nerve's geometry and split at bifurcations.

The result is nerve_paths_fibers.npz (per-fiber 3-D polylines) plus nerve_paths_caps.json and a branch summary. A simpler straight, axial mode is available as an alternative. The two are selected by fiber_method ("streamlines" for curved, physics-based paths, or "axial" for straight ones).

Seeding parameters

Set in the Import wizard's Fibers step (GUI) or set_fiber_seed:

Parameter	Meaning
n_fibers	number of seed points
fiber_method	streamlines (curved) or axial (straight)
fiber_max_steps	max integration steps per streamline
fiber_seed_end	which cap to seed from
fiber_cluster_eps_mm	cap-clustering radius (branch separation)
fiber_cap_band_pct	z-band near each extremum to search for caps
fiber_min_rel_size_pct	drop cap clusters below this relative size
fiber_axial_normal_thresh	how "axial" a facet normal must be to count as a cap
fiber_auto_detect_branches	auto-find branches vs use a fixed count

See the Configuration Reference for defaults.

---

Fiber populations

A fiber isn't just a path — it has a diameter and a fiber type, which set its model and its activation behaviour. golgi ships biology-realistic population presets so you don't have to guess ("what diameter for B-fibers again?"), each carrying a literature citation and per-branch composition:

• Per-type rows (A-α, A-β, A-δ, B, C) with mean/SD diameter, distribution (gaussian / lognormal), and fraction.
• Presets such as cervical vagus (human / pig / rat) and recurrent laryngeal, plus generic myelinated-only and unmyelinated-only sets.
• A live KDE preview of the diameter distribution before you commit, and a citation shown beneath the preset.

Diameter maps to a fiber model: myelinated types use MRG (via NEURON/PyFibers or the AxonML surrogate); small unmyelinated C fibers use unmyelinated models (e.g. Tigerholm/Sundt). If a preset includes C fibers but the selected model is myelinated-only, golgi flags them rather than silently dropping them. The mechanics of turning a population into spikes and thresholds are on the Fiber Models & Activation page.

Branch selectivity. Because every fiber is labelled with its branch, recruitment and selectivity can be computed per branch — e.g. "recruit the cardiac branch while sparing the rest" — see Recruitment Sweeps & Selectivity.

---

Topographic seeding (roadmap)

A topographic-atlas seeding mode (drawing seed locations from per-fascicle functional-group distributions, e.g. Hammer/Settell vagus atlases) is described in the feature roadmap; the current release seeds across the cross-section (uniform / cap distribution) with the population presets above.

---

References

The population presets carry the literature attributions shown in the UI (pop_presets.py):

• Soltanpour N, Santer RM (1996). J Anat. and Verlinden et al. (2016). Auton Neurosci. — human cervical vagus composition.
• Settell ME, et al. (2020). J Neural Eng. and Nicolai EN, et al. (2020). J Neural Eng. — pig cervical vagus fascicular organisation.
• Pelot NA, et al. (2020). Quantified morphology of the cervical and subdiaphragmatic vagus nerves of human, pig, and rat. Front Neurosci 14:1148. — rat cervical vagus + cross-species morphometry.
• Mu L, Sanders I (2009). Anat Rec. — recurrent laryngeal nerve.
• Thio BJ, Titus ND, Pelot NA, Grill WM (2024). Reverse-engineered models reveal differential membrane properties of autonomic and cutaneous unmyelinated fibers. PLOS Comput Biol. — unmyelinated autonomic fiber properties.

---

See also

Fiber Models & Activation · Meshing · Recruitment Sweeps & Selectivity · Validation · Configuration Reference