calc_vesselness_measure_image#

petpal.preproc.segmentation_tools.calc_vesselness_measure_image(input_image: ants.core.ANTsImage, sigma_min: float = 2.0, sigma_max: float = 8.0, alpha: float = 0.5, beta: float = 0.5, gamma: float = 5.0, morph_open_radius: int = 1, **hessian_func_kwargs) → ants.core.ANTsImage#

Computes a vesselness measure image using Hessian-based objectness filtering.

This function calculates the vesselness measure of a given 3D image using multi-scale Hessian filtering with specified parameters. We call the ants.hessian_objectness() after max-normalizing the input image. It enhances tubular structures like vessels, making them more pronounced in the output image. Optionally, a morphological opening operation can be applied to the result to refine the output and remove pepper-like artefacts.

From the docs of ants.hessian_objectness(): ‘ Based on the paper by Westin et al., “Geometrical Diffusion Measures for MRI from Tensor Basis Analysis” and Luca Antiga’s Insight Journal paper http://hdl.handle.net/1926/576. ‘

Parameters:

  • input_image (ants.core.ANTsImage) – Input 3D image for vesselness computation.

  • sigma_min (float, optional) – Minimum scale for multi-scale Hessian filtering (default: 2.0).

  • sigma_max (float, optional) – Maximum scale for multi-scale Hessian filtering (default: 8.0).

  • alpha (float, optional) – Alpha parameter for vesselness computation (default: 0.5).

  • beta (float, optional) – Beta parameter for vesselness computation (default: 0.5).

  • gamma (float, optional) – Gamma parameter for vesselness computation (default: 5.0).

  • morph_open_radius (int, optional) – Radius for the optional morphological opening operation (default: 1). If set to 0, no morphological opening will be applied.

  • **hessian_func_kwargs – Additional keyword arguments for the Hessian objectness function.

Returns:

ants.core.ANTsImage – The vesselness-enhanced image.

Notes

  • Input image must be 3D; an assertion will fail if a non-3D image is provided.

  • The input image is normalized before processing to ensure robustness.

  • Morphological opening, if applied, uses a grayscale operation to refine the tubular structures.

  • The function has defaults for vesselness computation, but can be used to detect globular or plate-like structures as well.

Raises:

AssertionError – If the input image is not 3D.

Workflow:

  1. Normalize the input image to have values between 0 and 1.

  2. Apply Hessian-based objectness filtering using the provided parameters (sigma_min, sigma_max, alpha, beta, gamma).

  3. Perform a morphological opening operation with the specified radius (morph_open_radius), if applicable.

  4. Return the computed vesselness image.