clean_grid_ppv#

crispy.grid_ridge.clean_grid_ppv(coord, refdata, coord_in_xfirst=False, start_index=1, min_length=6, method='robust')[source]#

Process and grid CRISPy coordinates in PPV space onto a reference image, labeling and cleaning skeleton structures.

Grids CRISPy coordinates onto a position-position-velocity (PPV) reference image, labels distinct ridge structures using DBSCAN, and removes endpoints to prevent overlap between structures. Structures shorter than a specified projected length are filtered out, and vertical segments are truncated based on a velocity threshold.

Parameters:

  • coord (ndarray) – Coordinates of the ridge points, shape (n, D), where n is the number of points and D is the dimensionality (typically 3D in PPV space).

  • refdata (ndarray) – Reference PPV image array defining the grid dimensions.

  • coord_in_xfirst (bool, optional, default=False) – If True, assumes the input coordinates are ordered with x as the first axis. If False, assumes z is the first axis.

  • start_index (int, optional, default=1) – The starting index for gridding the skeleton onto the reference image.

  • min_length (int, optional, default=6) – Minimum projected length (in pixels) for structures to be retained.

  • method ({"robust", "fast"}, optional, default="robust") – Method for cleaning skeleton endpoints: - “robust”: Handles diagonal connections but is computationally intensive. - “fast”: Faster but may miss diagonally connected structures.

Returns:

skel_cube – Binary array with the same shape as refdata, where gridded skeleton structures are set to True.

Return type:

ndarray

Notes

  • DBSCAN is used to label ridge structures with higher resultion than the image grid

  • Endpoints are removed to avoid overlap between distinct structures.

  • Vertical segments in velocity are removed based on a threshold (delVelMax set to 2 pixels).

  • The current implementation supports only 3D PPV data.

Examples

Grid and clean ridge coordinates in PPV space:

>>> import numpy as np
>>> from crispy import grid_ridge
>>> coords = np.array([[0, 0, 0], [1, 1, 1], [10, 10, 10]])
>>> ref_image = np.zeros((20, 20, 20))  # Reference PPV image
>>> skel_cube = grid_ridge.clean_grid_ppv(coords, ref_image, min_length=5, method="robust")
>>> print(skel_cube.shape)
(20, 20, 20)
On this page

This Page