itasc.contact_analysis.quantifiers.shape_relational

Relational shape quantifier — nucleus-vs-cell paired morphology.

Wraps itasc.contact_analysis.shape.compute_relational_table(), which pairs each nucleus with its cell on the shared (frame, id) key and emits relational quantities (nuclear:cell area ratio, centroid offset, orientation delta, …). Needs both label stacks plus a pixel size. A pooled quantity: compute_object_table() builds the tidy table in memory (no per-position artifact). The object-key column is cell_id, so the relational table pools through the same path as the per-source shape tables.

Classes

ShapeRelationalQuantifier()

Quantifies per-(frame, cell) relational nucleus-vs-cell shape.

class itasc.contact_analysis.quantifiers.shape_relational.ShapeRelationalQuantifier

Bases: Quantifier

Quantifies per-(frame, cell) relational nucleus-vs-cell shape.

quantity_id: ClassVar[str] = 'shape_relational'

Stable key; an empty value marks an intermediate (non-registered) base.

display_name: ClassVar[str] = 'Nucleus–cell shape'

Human-readable label shown wherever a quantity is selected.

requires: ClassVar[tuple[str, ...]] = ('cell_labels_path', 'nucleus_labels_path')

PositionInputs field names this quantifier needs to build.

required_build_params: ClassVar[dict[str, str]] = {'pixel_size_um': 'pixel size (µm/px)'}

Shared build-param keys that must be present and positive for this quantifier to build, mapped to the human label shown in the UI (e.g. cell density needs {"fov_area_mm2": "FOV area (mm²)"}). The studio greys the metric out — rather than letting the build fail — until they are supplied. Empty for a quantifier that needs no shared param.

table_keys: ClassVar[tuple[str, ...]] = ('frame', 'cell_id')

The index columns of this quantifier’s object_table — its natural grain (e.g. ("frame", "cell_id")). A non-empty value marks the quantity as aggregated: itasc.contact_analysis.shape_tables pools it across positions into a table named by quantity_id, keyed on these columns. Empty ⇒ not aggregated into an index-keyed table (e.g. contacts). Value columns are namespaced by quantity_id so a later joined view across tables never has colliding names.

compute_object_table(inputs, *, params=None)

The pooled tidy table for one position, computed directly from inputs.

A pooled quantifier (one that declares table_keys) implements this: it returns the same column-major table that object_table used to return after a disk round-trip, but never touches disk. None when this position yields no rows. Producers (contacts) are not pooled and do not implement it. params carries the shared build knobs (e.g. fov_area_mm2) for the quantifiers that opt in; per-position values (pixel size, frame interval) arrive via inputs.