Tracking method and apparatus

1. A method of determining a path of a tubular structure using processing circuitry, comprising:

obtaining volumetric medical imaging data that represents anatomy of a subject including the tubular structure;

performing position-determining processing comprising:

obtaining an initial position of a current point on said path of the tubular structure;

selecting a sub-region based on said initial position;

obtaining at least one value of at least one parameter representative of the tubular structure;

inputting to a trained model both data from said selected sub-region and said at least one parameter value, the trained model having been trained to determine paths of tubular structures;

outputting by the trained model a position for a next point on said path of the tubular structure; and

updating said initial position to said next point; and

repeating the position-determining procedure for updating the initial position thereby to obtain the path of the tubular structure.

2. A method according to claim 1 , wherein the at least one value for the at least one parameter is representative of a three-dimensional curvature of the path of the tubular structure.

3. A method according to claim 1 , wherein the at least one parameter comprises curvature and/or torsion.

4. A method according to claim 1 , wherein the at least one value for the at least one parameter is obtained from the current point and/or at least one preceding point on the tubular structure.

5. A method according to claim 1 , wherein the position-determining procedure is further configured to output at least one updated value for the at least one parameter.

6. A method according to claim 1 , wherein the selecting of the sub-region is based on a determination of an estimated position for the next point.

7. A method according to claim 6 , wherein the selected sub-region has a center that is between the respective current point and an estimated position of the respective next point.

8. A method according to claim 1 , wherein the trained model is configured to determine presence of a branch point where the tubular structure splits into a plurality of branches; and

if a branch point is present, the method comprises performing a new succession of position-determining procedures for each of the plurality of branches thereby to determine the paths of each of the plurality of branches.

9. A method according to claim 1 , wherein the data from said selected sub-region comprises a convolution.

10. A method according to claim 1 , wherein the method further comprises determining the data from said selected sub-region using a further trained model.

11. A method according to claim 10 , wherein the further trained model is further configured to perform a segmentation of at least part of the tubular structure.

12. A method according to claim 1 , wherein the trained model comprises at least one of a convolutional neural network (CNN), a recurrent network, a long short-term memory (LSTM).

13. A method according to claim 1 , wherein the at least one parameter comprises step size.

14. A method according to claim 1 , wherein the tubular structure comprises at least one of a blood vessel, an airway.

15. A method according to claim 1 , wherein the trained model comprises a self-learning model.

16. An apparatus comprising processing circuitry configured to perform a method for determining a path of a tubular structure, the method comprising:

obtaining volumetric medical imaging data that represents anatomy of a subject including the tubular structure;

performing a position-determining procedure that comprises:

obtaining an initial position of a current point on said path of the tubular structure;

selecting a sub-region based on said initial position;

obtaining at least one value of at least one parameter representative of the tubular structure;

inputting to a trained model both data from said selected sub-region and said at least one parameter value, the trained model having been trained to determine paths of tubular structures;

outputting by the trained model a position for a next point on said path of the tubular structure; and

updating said initial position to said next point; and

repeating the position-determining procedure for updating the initial position thereby to obtain the path of the tubular structure.

17. A method for training a model to determine paths of tubular structures using processing circuitry, the method comprising:

obtaining a plurality of training sets of volumetric medical imaging data, each representing anatomy of a respective subject including a respective tubular structure;

obtaining a respective ground truth track for each of the training sets of volumetric medical imaging data;

for each of the training sets of volumetric medical imaging data, obtaining data for a plurality of sub-regions along the respective tubular structure; and

training the model to determining paths of tubular structures using data from sub-regions and values for at least one parameter representative of the tubular structure.

18. A method according to claim 17 , further comprising scaling at least some of the training sets of volumetric medical imaging data to a standard vessel size and/or to multiple vessel sizes.

19. A method according to claim 17 , further comprising transforming at least some of the training sets of volumetric medical imaging data using domain adaptation techniques.

20. An apparatus comprising processing circuitry configured to perform a method for training a model to determine paths of tubular structures, the method comprising:

obtaining a plurality of training sets of volumetric medical imaging data, each representing anatomy of a respective subject including a respective tubular structure;

obtaining a respective ground truth track for each of the training sets of volumetric medical imaging data;

for each of the training sets of volumetric medical imaging data, obtaining data for a plurality of sub-regions along the respective tubular structure; and

training the model to determining paths of tubular structures using data from sub-regions and values for at least one parameter representative of the tubular structure.