Apparatuses and methods for navigation in and Local segmentation extension of anatomical treelike structures

1. A local extension method for segmentation of anatomical treelike structures, the method comprising:

receiving an initial segmentation of 3D image data including an initial treelike structure;

defining a target point in the 3D image data;

extracting a region of interest based on the target point to create a sub-image;

detecting highly tubular voxels in the sub-image;

performing a spillage-constrained region growing using the highly tubular voxels as seed points;

extracting connected components from the region growing step;

pruning the extracted components to discard components not likely to be connected to the initial treelike structure and keep only candidate components likely to be a valid sub-tree of the initial treelike structure; and

connecting the candidate components to the initial treelike structure, the candidate components determined to have boundary voxels that are 4-connected to the initial segmentation and filling in voxels to fully connect the candidate component to the initial segmentation.

2. The method of claim 1 , wherein pruning comprises discarding components based on whether they have genus zero and are relatively convex.

3. The method of claim 2 , wherein discarding components comprises determining whether a ratio of a volume of the extracted component before morphological closing to a volume of the extracted component after morphological closing is greater than a threshold.

4. The method of claim 1 , wherein pruning comprises discarding or modifying components based on the components curving more than a threshold amount.

5. The method of claim 4 , wherein discarding components comprises determining whether a dot product of a direction of the extracted component and a direction of a segment of the extracted component is below a threshold.

6. The method of claim 1 , wherein pruning comprises discarding components based on a shape analysis of the shape of the extracted component.

7. The method of claim 6 , wherein discarding components comprises comparing shape analysis metrics including at least one of a minor axis length, an elongation, and an eccentricity of the extracted component to target metrics.

8. The method of claim 1 , wherein pruning comprises discarding components based on a cylindrical score of the components.

9. The method of claim 8 , wherein discarding components comprises determining a circle equivalent diameter of the extracted component.

10. The method of claim 1 , wherein pruning comprises iteratively evaluating image processing and geometric characteristics of each extracted component to determine whether the extracted component is a candidate component.

11. The method of claim 1 , wherein connecting comprises determining whether the radius of a candidate component is compatible with a terminal segment of the initial segmentation, and filling in voxels to connect the candidate component to the terminal segment if they are compatible.

12. The method of claim 1 , wherein connecting comprises determining if the candidate component includes a path of seed points to the initial segmentation.

13. The method of claim 12 , wherein if the candidate component includes the path of seed points, connecting further comprises identifying potential new growth based on a gradient ball method.

14. The method of claim 1 , further comprising creating a skeletonization of the initial segmentation and mask-labeling voxels in the initial segmentation according to the skeletonization.

15. The method of claim 1 , further comprising creating a skeletonization of the extracted components and mask-labeling voxels in the extracted components according to the skeletonization.

16. The method of claim 1 , further comprising smoothing the sub-image and enhancing the contrast in the sub-image.

17. A method comprising:

receiving 3D image data;

segmenting an initial treelike structure in the 3D image data to create an initial segmentation;

creating a skeletonization of the initial segmentation;

mask-labeling voxels in the initial segmentation according to the skeletonization;

defining a target point in the initial segmentation;

extracting a region of interest based on the target point to create a sub-image;

smoothing the sub-image;

enhancing contrast in the sub-image;

detecting highly tubular voxels in the sub-image;

performing a spillage-constrained region growing using the highly tubular voxels as seed points;

extracting connected components from the region growing results;

creating a skeletonization of the extracted components;

mask-labeling voxels in the extracted components according to the skeletonization;

pruning the extracted components to discard components not likely to be connected to the initial treelike structure and keep only candidate components likely to be a valid sub-tree of the initial treelike structure; and

connecting the candidate components to the initial treelike structure, wherein connecting comprises:

determining whether the candidate component has boundary voxels that are 4-connected to the initial segmentation and if so filling in voxels to fully connect the candidate component to the initial segmentation;

determining whether the radius of a candidate component is compatible with a terminal segment of the initial segmentation, and filling in voxels to connect the candidate component to the terminal segment if they are compatible;

determining if the candidate component includes a path of seed points to the initial segmentation, and if so identifying potential new growth based on a gradient ball method.

18. The method of claim 17 , wherein pruning comprises:

discarding components based on whether they have genus zero and are relatively convex, including determining whether a ratio of a volume of the extracted component before morphological closing to a volume of the extracted component after morphological closing is greater than a threshold;

discarding or modifying components based on the components curving more than a threshold amount, including determining whether a dot product of a direction of the extracted component and a direction of a segment of the extracted component is below a threshold;

discarding components based on a shape analysis of the shape of the extracted component, including comparing shape analysis metrics including at least one of a minor axis length, an elongation, and an eccentricity of the extracted component to target metrics; and

discarding components based on a cylindrical score of the components, including determining a circle equivalent diameter of the extracted component.