Place layered populations

A PlaceNeurons step is itself a single NeuronPopulation: its fields (morphology, soma_radius_um, depth_range_um, density_per_mm3, …) describe one group of cells. The common case is exactly that - one population:

from minisim import PlaceNeurons

PlaceNeurons(morphology="cytosolic", soma_radius_um=7.0, depth_range_um=(0.0, 200.0))

Combine several populations

To build layered anatomy - say a thin soma-targeted band sitting over a deeper cytosolic volume - set populations to a list of NeuronPopulation. The step samples each in turn and concatenates the cells; the step’s own population fields are then unused (setting both raises, so there is no ambiguity about which wins).

from minisim import PlaceNeurons, NeuronPopulation

PlaceNeurons(populations=[
    NeuronPopulation(                       # thin soma-targeted layer
        morphology="soma",
        depth_range_um=(50.0, 60.0),
        density_per_mm3=80000.0,
    ),
    NeuronPopulation(                       # deep cytosolic volume
        morphology="cytosolic",
        depth_range_um=(100.0, 300.0),
        density_per_mm3=25000.0,
    ),
])

Each population’s count is volumetric (density_per_mm3 × FOV area × depth thickness), and any min_distance_um spacing is enforced within a population, not across them - so adjacent layers may interpenetrate at their shared boundary, the physical case.

Place cells at exact positions

A population can also be placed at explicit centers instead of being density-sampled: set positions_um to a list of (z, y, x) µm tuples. Note the depth-first order, matching Cell.center_um rather than x, y, z. z is depth below the tissue surface (0 = surface); y, x are lateral in the optical-center frame - the optical axis is (0, 0), +y down and +x right (image convention). So (z, 0, 0) is dead-center in the FOV regardless of the motion margin, and these are the same coordinates the ground truth (GroundTruth.centers_um) reports back.

PlaceNeurons(
    morphology="cytosolic",
    soma_radius_um=8.0,
    positions_um=[
        (120.0, 0.0, 0.0),       # (z=depth, y, x) - on the optical axis (FOV center)
        (120.0, -60.0, 80.0),    # 60 µm up, 80 µm right of center
    ],
)

When positions_um is given, the distribution fields (density_per_mm3, depth_range_um, min_distance_um) are ignored; the shape fields (soma_radius_um, irregularity, morphology, dendrites) still apply to each placed cell. Because positions live on the population, you can mix an explicit-position population and a density-sampled one in the same populations list - handy for dropping a few cells at known spots inside an otherwise random field.