Feature Extraction

Hierarchical feature extraction for multi-level analysis.

Hierarchical Features

Extract features at multiple levels: voxels, nodes, branches, organelles, and images.

hierarchical

Hierarchy

Hierarchy(im_info, skip_nodes=True, viewer=None, use_gpu=True, low_memory=False, enable_motility=True, enable_adjacency=True, device=None, node_chunk_size=None, max_node_mask_elems=int(50000000.0))

Main orchestration class for hierarchical feature extraction.

Parameters:

Name	Type	Description	Default
`im_info`	`ImInfo`	Image metadata object.	required
`skip_nodes`	`bool`	If True, node-level features are skipped.	`True`
`viewer`	`optional`	Viewer object with `.status` attribute for status updates.	`None`
`use_gpu`	`bool`	If True and CuPy is available, some computations will attempt to use GPU. Ignored when device is set to "cpu" or "gpu".	`True`
`low_memory`	`bool`	If True, use low-memory (slower) aggregation strategies where possible.	`False`
`enable_motility`	`bool`	If False, skip all motion-related features (flow interpolation, velocities, etc.).	`True`
`enable_adjacency`	`bool`	If False, skip adjacency map construction.	`True`
`device`	`(auto, cpu, gpu)`	Backend selection for GPU-eligible operations. "auto" uses GPU if available.	`"auto"`
`node_chunk_size`	`int`	Target number of voxels per chunk when assigning voxels to nodes.	`None`
`max_node_mask_elems`	`int`	Upper bound on the size of the node/voxel mask (num_nodes * chunk_size).	`int(50000000.0)`

run

run()

Main execution method.

Voxels

Voxels(hierarchy)

Voxel-level features.

Nodes

Nodes(hierarchy)

Node-level features.

Branches

Branches(hierarchy)

Branch-level features.

Components

Components(hierarchy)

Component (organelle)-level features.

Image

Image(hierarchy)

Image-level aggregated features.

append_to_array

append_to_array(to_append)

Convert feature dict into list-of-arrays + headers.

create_feature_array

create_feature_array(level, labels=None)

Original non-streaming implementation kept for backwards compatibility. Not used inside Hierarchy anymore (which streams to CSV directly).

aggregate_stats_for_class

aggregate_stats_for_class(child_class, t, list_of_idxs, low_memory=False)

Aggregate mean/std/min/max/sum over groups of indices for a given class at time t.

Parameters:

Name	Type	Description	Default
`child_class`	`object with stats_to_aggregate and per-stat per-frame arrays`		required
`t`	`int`		required
`list_of_idxs`	`list of 1D index arrays (groups)`		required
`low_memory`	`bool`	If True, use a slower but memory-light implementation.	`False`

distance_check

distance_check(border_mask, check_coords, spacing)

Compute distance from points in check_coords to nearest border point.