Granted Patent B2

US 8,422,753 · App. 12/901,600 · Granted Apr 16, 2013

Abstract

Claims (87)

1. A method for extracting a personalized left atrium model from a 3D image volume, comprising:

segmenting a left atrium chamber body in the 3D image volume;

detecting at least one pulmonary venous ostia on the segmented left atrium chamber body using at least one trained detector; and

segmenting at least one pulmonary vein trunk connected to the left atrium chamber body based on the detected pulmonary venous ostia.

2. The method of claim 1 , further comprising:

generating the personalized left atrium model by fitting the segmented at least one pulmonary vein trunk to the segmented left atrium chamber body.

3. The method of claim 1 , wherein said step of segmenting a left atrium chamber body comprises:

localizing the left atrium chamber body in the 3D volume using marginal space learning (MSL); and

deforming a boundary of the left atrium chamber body using a learning based boundary detector.

4. The method of claim 3 , wherein said step of localizing the left atrium chamber body comprises:

estimating position candidates for the left atrium chamber body using a trained position detector;

estimating position-orientation candidates for the left atrium chamber body based on the estimated position candidates using a trained position-orientation detector;

estimating a full similarity transform of the left atrium chamber body based on the estimated position-orientation candidates using a trained position-orientation-scale detector; and

determining an initial shape of the left atrium chamber body using the estimated full similarity transform.

5. The method of claim 3 , wherein said step of deforming a boundary of the left atrium chamber body comprises:

deforming the boundary of the localized left atrium chamber body using the learning based boundary detector and an active shape model.

6. The method of claim 1 , wherein said step of detecting at least one pulmonary venous ostia comprises:

generating probability maps of pulmonary venous ostia in the 3D image volume at a plurality of resolutions using a plurality of trained detectors;

determining a configuration of the pulmonary venous ostia based on the probability maps; and

mapping the configuration of the pulmonary venous ostia to a set of major pulmonary drainage patterns.

7. The method of claim 1 , wherein said step of segmenting at least one pulmonary vein trunk comprises:

determining a center line of the at least one pulmonary vein trunk from the detected at least one pulmonary ostia on the left atrium chamber body;

defining a radius of the at least one pulmonary vein trunk; and

refining a boundary of the at least one pulmonary vein trunk.

8. The method of claim 7 , wherein said step of refining a boundary of the at least one pulmonary vein trunk comprises:

adjusting mesh points on the at least one pulmonary vein trunk based on a learning based boundary detector; and

smoothing the mesh points.

9. The method of claim 1 , further comprising:

automatically generating an ablation strategy based upon the extracted personalized left atrium model.

10. An apparatus for extraction of a personalized left atrium model from a 3D image volume, comprising:

means for segmenting a left atrium chamber body in the 3D image volume;

means for detecting a pulmonary venous ostia on the segmented left atrium chamber body using at least one trained detector;

means for segmenting at least one pulmonary vein trunk connected to the left atrium chamber body based on the detected pulmonary venous ostia.

11. The apparatus of claim 10 , further comprising:

means for generating the personalized left atrium model by fitting the segmented at least one pulmonary vein trunk to the segmented left atrium chamber body.

12. The apparatus of claim 10 , wherein said means for segmenting a left atrium chamber body comprises:

means for localizing the left atrium chamber body in the 3D volume using marginal space learning (MSL); and

means for deforming a boundary of the left atrium chamber body using a learning based boundary detector.

13. The apparatus of claim 12 , wherein said means for localizing the left atrium chamber body comprises:

means for estimating position candidates for the left atrium chamber body using a trained position detector;

means for estimating position-orientation candidates for the left atrium chamber body based on the estimated position candidates using a trained position-orientation detector;

means for estimating a full similarity transform of the left atrium chamber body based on the estimated position-orientation candidates using a trained position-orientation-scale detector; and

means for determining an initial shape of the left atrium chamber body using the estimated full similarity transform.

14. The apparatus of claim 12 , wherein said means for deforming a boundary of the left atrium chamber body comprises:

means for deforming the boundary of the localized left atrium chamber body using the learning based boundary detector and an active shape model.

15. The apparatus of claim 10 , wherein said means for detecting at least one pulmonary venous ostia comprises:

means for generating probability maps of pulmonary venous ostia from the 3D image volume at a plurality of resolutions using a plurality of trained detectors;

means for determining the configuration of the pulmonary venous ostia based on the probability maps; and

means for mapping the configuration of the pulmonary venous ostia to a set of major pulmonary drainage patterns.

16. The apparatus of claim 10 , wherein said means for segmenting at least one pulmonary vein trunk comprises:

means for determining a center line of the at least one pulmonary vein trunk from the detected at least one pulmonary ostia on the left atrium chamber body;

means for defining a radius of the at least one pulmonary vein trunk; and

means for refining a boundary of the at least one pulmonary vein trunk.

17. The apparatus of claim 16 , wherein said means for refining a boundary of the at least one pulmonary being trunk comprises:

means for adjusting mesh points on the at least one pulmonary vein trunk based on a learning based boundary detector; and

means for smoothing the mesh points.

18. The apparatus of claim 10 , further comprising:

means for automatically generating an ablation strategy based upon the extracted personalized left atrium model.

19. A non-transitory computer readable medium encoded with computer executable instructions for extraction of a personalized left atrium model from a 3D image volume, the computer executable instructions defining steps comprising:

segmenting a left atrium chamber body in the 3D image volume;

detecting at least one pulmonary venous ostia on the segmented left atrium chamber body using at least one trained detector; and

segmenting at least one pulmonary vein trunk connected to the left atrium chamber body based on the detected pulmonary venous ostia.

20. The non-transitory computer readable medium of claim 19 , the computer executable instructions further defining steps comprising:

generating the personalized left atrium model by fitting the segmented at least one pulmonary vein trunk to the segmented left atrium chamber body.

21. The non-transitory computer readable medium of claim 19 , wherein the computer executable instructions defining the step of segmenting a left atrium chamber body comprises:

localizing the left atrium chamber body in the 3D volume using marginal space learning (MSL); and

deforming a boundary of the left atrium chamber body using a learning based boundary detector.

22. The non-transitory computer readable medium of claim 21 , wherein the computer executable instructions defining the step of localizing the left atrium chamber body comprises:

estimating position candidates for the left atrium chamber body using a trained position detector;

estimating position-orientation candidates for the left atrium chamber body based on the estimated position candidates using a trained position-orientation detector;

estimating a full similarity transform of the left atrium chamber body based on the estimated position-orientation candidates using a trained position-orientation-scale detector; and

determining an initial shape of the left atrium chamber body using the estimated full similarity transform.

23. The non-transitory computer readable medium of claim 21 , wherein the computer executable instructions defining the step of deforming a boundary of the left atrium chamber body comprises:

deforming the boundary of the localized left atrium chamber body using the learning based boundary detector and an active shape model.

24. The non-transitory computer readable medium of claim 19 , wherein the computer executable instructions defining the step of detecting at least one pulmonary venous ostia comprises:

generating probability maps of pulmonary venous ostia in the 3D image volume at a plurality of resolutions using a plurality of trained detectors;

determining a configuration of the pulmonary venous ostia based on the probability maps; and

mapping the configuration of the pulmonary venous ostia to a set of major pulmonary drainage patterns.

25. The non-transitory computer readable medium of claim 19 , wherein the computer executable instructions defining the step of segmenting at least one pulmonary vein trunk comprises:

determining a center line of the at least one pulmonary vein trunk from the detected at least one pulmonary ostia on the left atrium chamber body;

defining a radius of the at least one pulmonary vein trunk; and

refining a boundary of the at least one pulmonary vein trunk.

26. The non-transitory computer readable medium of claim 25 , wherein the computer executable instructions defining the step of refining a boundary of the at least one pulmonary vein trunk comprises:

adjusting mesh points on the at least one pulmonary vein trunk based on a learning based boundary detector; and

smoothing the mesh points.

27. The non-transitory computer readable medium of claim 19 , further comprising computer executable instructions defining the step of:

automatically generating an ablation strategy based upon the extracted personalized left atrium model.

Assignments (6)