Method and system for intervention planning for transcatheter aortic valve implantation from 3D computed tomography data

1. A method for intervention planning for a transcatheter aortic valve implantation, comprising:

detecting a patient-specific aortic valve model in a computed tomography (CT) volume of a patient by:

detecting a global location of the patient-specific aortic valve model in the CT volume,

detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model, and

fitting an aortic root surface model to the CT volume; and

automatically determining at least one intervention parameter based on the patient-specific aortic valve model,

wherein detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model comprises:

detecting the plurality of aortic valve landmarks using a corresponding plurality of trained landmark detectors constrained by the detected global location of the patient-specific aortic valve model.

2. The method of claim 1 , wherein the patient-specific aortic valve model comprises a global affine model, a landmark model including a three aortic valve commissure landmarks and three aortic valve hinge landmarks, and an aortic root surface model.

3. The method of claim 1 , wherein detecting a global location of the patient-specific aortic valve model in the CT volume comprises:

detecting an affine transformation defining the global location of the patient-specific aortic valve model in the CT volume using Marginal Space Learning (MSL).

4. The method of claim 1 , wherein detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model comprises:

detecting, in a search space constrained by the detected global location, each of three aortic valve hinges and three aortic valve commissures using a respective trained landmark detector.

5. The method of claim 1 , wherein the global location of the patient-specific aortic valve model in the CT volume is detected at a first resolution and the plurality of aortic valve landmarks are detected at a second resolution that is finer than the first resolution.

6. The method of claim 1 , wherein fitting an aortic root surface model to the CT volume comprises:

generating a plurality of aortic valve shapes from a learned statistical shape model;

projecting each of the plurality of aortic valve shapes to an image space of the CT image using the detected global location of the patient-specific aortic valve model, resulting in a plurality of surface model hypotheses; and

selecting one of the surface model hypotheses as the aortic root surface model using a trained classifier.

7. The method of claim 1 , wherein detecting a patient-specific aortic valve model in a computed tomography (CT) volume of a patient further comprises:

refining the aortic root surface model using a trained boundary detector.

8. The method of claim 1 , wherein automatically determining at least one intervention parameter based on the patient-specific aortic valve model comprises:

automatically determining angulation parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images.

9. The method of claim 8 , wherein automatically determining angulations parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images comprises:

defining an annulus plane based on hinge landmark points in the patient-specific aortic valve model; and

determining the angulation parameters such that a viewing orientation of the C-arm imaging system is orthogonal to the annulus plane.

10. The method of claim 1 , wherein automatically determining angulations parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images comprises:

determining an annulus measurement from the patient-specific aortic valve model.

11. An apparatus for intervention planning for a transcatheter aortic valve implantations, comprising:

means for detecting a patient-specific aortic valve model in a computed tomography (CT) volume of a patient comprising:

means for detecting a global location of the patient-specific aortic valve model in the CT volume,

means for detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model, and

means for fitting an aortic root surface model to the CT volume; and

means for automatically determining at least one intervention parameter based on the patient-specific aortic valve model,

wherein the means for detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model comprises:

means for detecting, in a search space constrained by the detected global location, each of three aortic valve hinges and three aortic valve commissures.

12. The apparatus method of claim 11 , wherein the patient-specific aortic valve model comprises a global affine model, a landmark model including a three aortic valve commissure landmarks and three aortic valve hinge landmarks, and an aortic root surface model.

13. The apparatus of claim 11 , wherein the means for fitting an aortic root surface model to the CT volume comprises:

means for generating a plurality of aortic valve shapes from a learned statistical shape model;

means for projecting each of the plurality of aortic valve shapes to an image space of the CT image using the detected global location of the patient-specific aortic valve model, resulting in a plurality of surface model hypotheses; and

means for selecting one of the surface model hypotheses as the aortic root surface model using a trained classifier.

14. The apparatus of claim 11 , wherein the means for automatically determining at least one intervention parameter based on the patient-specific aortic valve model comprises:

means for automatically determining angulation parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images.

15. The apparatus of claim 11 , wherein the means for automatically determining angulations parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images comprises:

means for determining an annulus measurement from the patient-specific aortic valve model.

16. A non-transitory computer readable medium storing computer program instructions for intervention planning for a transcatheter aortic valve implantation, the computer program instructions when executed on a processor cause the processor to perform operations comprising:

detecting a patient-specific aortic valve model in a computed tomography (CT) volume of a patient by:

detecting a global location of the patient-specific aortic valve model in the CT volume,

detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model, and

fitting an aortic root surface model to the CT volume; and

automatically determining at least one intervention parameter based on the patient-specific aortic valve model,

wherein detecting a plurality of aortic valve landmarks in the CT volume based on the detected global location of the patient-specific aortic valve model comprises:

detecting, in a search space constrained by the detected global location, each of three aortic valve hinges and three aortic valve commissures using a respective trained landmark detector.

17. The non-transitory computer readable medium of claim 16 , wherein the patient-specific aortic valve model comprises a global affine model, a landmark model including a three aortic valve commissure landmarks and three aortic valve hinge landmarks, and an aortic root surface model.

18. The non-transitory computer readable medium of claim 16 , wherein fitting an aortic root surface model to the CT volume comprises:

generating a plurality of aortic valve shapes from a learned statistical shape model;

projecting each of the plurality of aortic valve shapes to an image space of the CT image using the detected global location of the patient-specific aortic valve model, resulting in a plurality of surface model hypotheses; and

selecting one of the surface model hypotheses as the aortic root surface model using a trained classifier.

19. The non-transitory computer readable medium of claim 16 , wherein detecting a patient-specific aortic valve model in a computed tomography (CT) volume of a patient further comprises:

refining the aortic root surface model using a trained boundary detector.

20. The non-transitory computer readable medium of claim 16 , wherein automatically determining at least one intervention parameter based on the patient-specific aortic valve model comprises:

automatically determining angulation parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images.

21. The non-transitory computer readable medium of claim 20 , wherein automatically determining angulations parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images comprises:

defining an annulus plane based on hinge landmark points in the patient-specific aortic valve model; and

determining the angulation parameters such that a viewing orientation of the C-arm imaging system is orthogonal to the annulus plane.

22. The non-transitory computer readable medium of claim 16 , wherein automatically determining angulations parameters for a C-arm imaging system for obtaining intra-operative fluoroscopic images comprises:

determining an annulus measurement from the patient-specific aortic valve model.