Iterative branching structure segmentation method and system

1. A method, comprising:

receiving an image representing a branching structure;

determining a starting feature of the branching structure;

selecting a subregion of the image based on the starting feature;

segmenting the branching structure in the subregion to generate a fully segmented branching structure output for the subregion;

generating a set of next features based on the fully segmented branching structure output, wherein:

the set of next features indicates where the branching structure in the image continues beyond the subregion; and

generating the set of next features comprises analyzing the fully segmented branching structure output with a neural network to generate the set of next features; and

for each of the next features, repeating the selecting of the subregion based on the next feature, the segmenting of the branching structure, and the generating of the set of next features.

2. The method of claim 1 , further comprising:

repeating the selecting of the subregion, the segmenting of the branching structure, and the generating of the set of next features until each next feature has been used to select a subregion, segment the branching structure in the subregion, and generate a set of next features based on the fully segmented branching structure output.

3. The method of claim 2 , further comprising:

combining the segmented branching structure based on the starting feature and segmented branching structures based on each of the next features.

4. The method of claim 3 , further comprising:

combining the combined segmented branching structure with a prior combined segmented branching structure.

5. The method of claim 4 , wherein:

the image includes a discontinuity dividing the branching structure into a first branching structure and a second branching structure;

the prior combined segmented branching structure was segmented from the first branching structure; and

the starting feature is a feature of the second branching structure.

6. The method of claim 1 , wherein:

segmenting the branching structure in the subregion comprises segmenting the branching structure in the subregion based on a previously segmented branching structure in a previous subregion.

7. The method of claim 6 , wherein:

the subregion overlaps the previous subregion.

8. The method of claim 1 , wherein:

segmenting the branching structure in the subregion comprises generating the segmented branching structure using a first convolutional neural network.

9. The method of claim 8 , wherein:

the first convolutional neural network is a U-net architecture.

10. The method of claim 8 , wherein:

generating the set of next features comprises generating the set of next features using a second convolutional neural network.

11. The method of claim 1 , further comprising:

analyzing the image representing the branching structure to determine the starting feature of the branching structure.

12. The method of claim 1 , wherein generating the set of next features comprises generating multiple features for the set of next features.

13. The method of claim 1 , wherein the branching structure is a vasculature of a living entity.

14. The method of claim 1 , wherein each subregion has a dimension of at least two image elements.

15. The method of claim 1 , wherein:

generating the set of next features further comprises generating each next feature of the set of next features on a perimeter of the subregion.

16. A system, comprising:

a communication interface configured to receive an image representing a branching structure; and

a processor configured to:

determine a starting feature of the branching structure;

select a subregion of the image based on the starting feature;

segment the branching structure in the subregion to generate a fully segmented branching structure output for the subregion;

generate a set of next features based on the fully segmented branching structure output, wherein:

the set of next features indicates where the branching structure in the image continues beyond the subregion; and

the processor is further configured to generate the set of next features by analyzing the fully segmented branching structure output with a neural network to generate the set of next features; and

for each of the next features, repeat the selecting of the subregion based on the next feature, the segmenting of the branching structure, and the generating of the set of next features.

17. The system of claim 16 , wherein the processor is further configured to:

repeat the selecting of the subregion, the segmenting of the branching structure, and the generating of the set of next features until each next feature has been used to select a subregion, segment the branching structure in the subregion, and generate a set of next features based on the fully segmented branching structure output.

18. The system of claim 16 , wherein the processor is further configured to:

combine the segmented branching structure based on the starting feature and segmented branching structures based on each of the next features; and

combine the combined segmented branching structure with a prior combined segmented branching structure.

19. The system of claim 16 , wherein the processor is further configured to:

segment the branching structure in the subregion by segmenting the branching structure in the subregion based on a previously segmented branching structure in a previous subregion.

20. A system, comprising:

means for receiving an image representing a branching structure;

means for determining a starting feature of the branching structure;

means for selecting a subregion of the image based on the starting feature;

means for segmenting the branching structure in the subregion to generate a fully segmented branching structure output for the subregion;

means for generating a set of next features based on the fully segmented branching structure output, wherein:

the set of next features indicates where the branching structure in the image continues beyond the subregion; and

the means for generating the set of next features comprises means for analyzing the fully segmented branching structure output with a neural network to generate the set of next features; and

means for, for each of the next features, repeating the selecting of the subregion based on the next feature, the segmenting of the branching structure, and the generating of the set of next features.

21. The system of claim 20 , further comprising:

means for segmenting the branching structure in the subregion comprises segmenting the branching structure in the subregion based on a previously segmented branching structure in a previous subregion.