Method and Apparatus for Detecting Intraventricular Dyssynchrony

• US 20080260230A1
• Filed: 09/18/2006
• Published: 10/23/2008
• Est. Priority Date: 09/16/2005
• Status: Active Grant

• Current Assignee: Coordenacao De Aperfeicoamento De Passoal De Nivel Superior, Coordenacao De Aperfeicoamento De Pessoal De Nivel Superior, Ohio State University
• Original Assignee: Ohio State University
• Inventors: Raman, Subha V., Gotardo, Paulo F.U., Boyer, Kim L.
• Granted Patent: US 8,280,136 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/10088 Magnetic resonance imaging [MRI]
• CPC Class Codes: G06T 2207/20116 Active contour; Active surface; Snakes
• CPC Class Codes: G06T 2207/30048 Heart; Cardiac
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/149 involving deformable models, e.g. active contour models
• CPC Class Codes: G06V 10/755 Deformable models or variational models, e.g. snakes or active contours

Abstract

A method and apparatus for detecting intraventricular dyssynchrony is provided. In one embodiment, the method includes: a) receiving information from an observer viewing a source image to define a plurality of points associated with muscle boundaries of a ventricle, b) extracting muscle boundaries for a series of source images to define a corresponding series of contour shape images using a Fourier descriptor model, and c) deforming each contour shape based at least in part on image forces, external forces, and internal forces to track movement of the muscle boundaries over time. In another embodiment, the method includes: a) characterizing the series of source images as a corresponding series of contour shape images representative of movement of muscle boundaries of the ventricle over time, b) classifying the series of contour shape images in a dyssynchronous class or a non-dyssynchronous class, and c) pre-diagnosing the ventricle as dyssynchronous or non-dyssynchronous.

Claims

1. A method of detecting intraventricular dyssynchrony, including:

a) receiving initialization information from an observer viewing a first source image of a series of source images of a ventricle of a beating heart over at least one cardiac cycle to define a plurality of points associated with muscle boundaries of the ventricle;

b) extracting muscle boundaries for each source image of the series of source images to define a corresponding series of contour shape images with contour shapes corresponding to the muscle boundaries using a Fourier descriptor model; and

c) deforming each contour shape in each contour shape image based at least in part on image forces, external forces, and internal forces to track movement of the muscle boundaries for the ventricle over time in the series of contour shape images.

2. The method of claim 1 wherein the initialization information includes at least four points.

3. The method of claim 1 wherein the first source image corresponds to the end of diastole for the cardiac cycle.

4. The method of claim 1 wherein the muscle boundaries include an endocardial wall and an epicardial wall.

5. The method of claim 1 wherein the set of source images include short-axis cine magnetic resonance images.

6. The method of claim 1 wherein the Fourier descriptor model includes:

$s<br><br>\left(t\right)=x<br><br>\left(t\right)+\mathrm{jy}<br><br>\left(t\right)=\underset{k=-N}{\overset{N}{\sum <br><br>}}<br><br>S_k<br><br>{\mathrm{<br><br>}}^{j<br><br><br><br>\mathrm{kt}},\mathrm{with}<br><br><br><br>j=\sqrt{-1},$

where each complex number S_k=a_k+jb_k, with a_k, b_kε

R, is treated as a rotating phasor with magnitude |S_k|=√

{square root over (a_k²+b_k²)} and phase angle

${\mathrm{\phi <br><br>}}_k={\tan }^{-1}<br><br>\left(\frac{b_k}{a_k}\right).$

7. The method of claim 6 wherein each pair of descriptors S_k and S₋

k describes an ellipse.

8. The method of claim 1 wherein the image forces include edge-based forces and texture-based forces.

9. The method of claim 8 wherein the edge-based forces have a limited zone of attraction.

10. The method of claim 1, the deforming in c) associated with image forces further including:

d) defining edges of the contour shapes in each contour shape image using an optimal zero crossing operator model to determine edge-strength vectors and gradient vectors for associated pixels of the contour shape images;

e) deforming edges of the contour shapes in each contour shape image using an averaging mask in conjunction with a range of pixels along the edges defined by the optimal zero crossing operator model;

f) determining a length of the contour shapes in each contour shape image using an approximation model; and

g) determining edge-based image forces for the contour shapes in each contour shape image based at least in part on the determined length for the corresponding contour shape.

11. The method of claim 10 wherein the approximation model includes:

$L^2=4<br><br>{\mathrm{\pi <br><br>}}^2<br><br>\underset{k=-N}{\overset{N}{\sum <br><br>}}<br><br>k^2<br><br>{<br><br>S_k<br><br>}^2=4<br><br>{\mathrm{\pi <br><br>}}^2<br><br>\underset{k=-N}{\overset{N}{\sum <br><br>}}<br><br>k^2<br><br>\left(a_k^2+b_k^2\right),$

where L is the length of the contour shape.

12. The method of claim 1, the deforming in c) further including:

d) sampling pixels values for the contour shapes in each contour shape image along a normal direction and at selected locations along a length of the corresponding contour shape to determine texture vectors associated with pixels of the contour shape images.

13. The method of claim 12 wherein the sampling in d) includes sampling pixel values for the contour shapes using:

$\begin{array}{c}s\perp <br><br>\left(t\right)=<br><br>\left[\begin{array}{c}u<br><br>\left(t\right)\\ v<br><br>\left(t\right)\end{array}\right]\\ =<br><br>\underset{k=-N}{\overset{N}{\sum <br><br>}}<br><br>k<br><br>\left[\begin{array}{cc}{a}_k& -b_k\\ b_k& a_k\end{array}\right]<br><br>\left[\begin{array}{c}\cos <br><br><br><br>\mathrm{kt}\\ \sin <br><br><br><br>\mathrm{kt}\end{array}\right]<br><br><br><br>\mathrm{at}<br><br><br><br>\mathrm{locations}<br><br><br><br>t_i\in <br><br>\left[0,2<br><br>\mathrm{\pi <br><br>}\right],\end{array}$

where 1≦

i≦

L₀ and L₀ represents the length of the corresponding contour shape, to determine the texture vectors.

14. The method of claim 1 wherein the extracting in b) for the first source image is based at least in part on the initialization information.

15. The method of claim 14 wherein the extracting in b) for each subsequent source image is based at least in part on the previous contour shape image.

16. The method of claim 1, the deforming in c) associated with external forces further including:

d) normalizing the scale of each Fourier descriptor associated with the Fourier descriptor model; and

e) determining external forces for each contour shape image to limit contour shape variations in consecutive contour shape images based at least in part on a predetermined value for a variance parameter.

17. The method of claim 16 wherein the scale-normalizing for each Fourier descriptor includes:

$\frac{S_k}{<br><br>S_1<br><br>}.$

18. The method of claim 1, the deforming in c) associated with internal forces further including:

d) determining a length of the contour shapes in each contour shape image using an approximation model; and

e) determining internal tension forces to minimize contour shape length and avoid self-crossings as contour shapes are deformed.

19. The method of claim 18 wherein the internal tension forces are weaker than the image forces.

20. The method of claim 18 wherein the approximation model includes:

$\begin{array}{c}{L}^2=<br><br>4<br><br>{\mathrm{\pi <br><br>}}^2<br><br>\underset{k=-N}{\overset{N}{\sum <br><br>}}<br><br>k^2<br><br>{<br><br>S_k<br><br>}^2\\ =<br><br>4<br><br>{\mathrm{\pi <br><br>}}^2<br><br>\underset{k=-N}{\overset{N}{\sum <br><br>}}<br><br>k^2<br><br>\left(a_k^2+b_k^2\right),\end{array}$

where L is the length of the contour shape.

21. A method of detecting intraventricular dyssynchrony, including:

a) characterizing a series of source images of a ventricle of a beating heart over at least one cardiac cycle as a corresponding series of contour shape images representative of movement of muscle boundaries of the ventricle over time using a Pourier descriptor model;

b) classifying the series of contour shape images in a dyssynchronous class or a non-dyssynchronous class; and

c) pre-diagnosing the ventricle as dyssynchronous or non-dyssynchronous based at least in part on the classifying.

22. The method of claim 21, the classifying in b) further including:

d) normalizing the series of contour shape images;

e) processing the series of contour shape images into a lower dimensional representation of principal and discriminative features of the series of contour shape images to form a classifier sample; and

f) classifying the classifier sample in the dyssynchronous class or the non-dyssynchronous class.

23. The method of claim 22, the classifying in f) further including:

g) projecting the classifier sample in a ventricle classification space defining the dyssynchronous class by a dyssynchronous class mean and defining the non-dyssynchronous class by a non-dyssynchronous class mean, the classification sample being represented in dimensions corresponding to dimensions for the ventricle classification space;

h) determining if the classification sample is closer to the dyssynchronous class mean or the non-dyssynchronous class mean; and

i) classifying the classifier sample in the dyssynchronous class or the non-dyssynchronous class based on which corresponding class mean is closest to the classifier sample.

24. A method of detecting intraventricular dyssynchrony, including:

a) observing a series of contour shape images representative of movement of muscle boundaries of a ventricle of a beating heart over at least one cardiac cycle, wherein the series of contour shape images is classified in a dyssynchronous class;

b) diagnosing the ventricle as dyssynchronous based at least in part on the observing;

c) planning treatment of the dyssynchronous ventricle to include epicardial implantation of a cardiac resynchronization therapy device;

d) identifying a dyssynchronous region of the ventricle based at least in part on the observing; and

e) planning lead placement for the cardiac resynchronization therapy device based at least in part on the dyssynchronous region of the ventricle.

25. The method of claim 24 wherein the dyssynchronous region of the ventricle includes at least one of an anteroseptal region, an inferoseptal region, an anterior region, an anterolateral region, an inferolateral region, and an inferior region.

26. The method of claim 24 wherein the identifying in d) is based at least in part on identifying a cardiac segment associated with the dyssynchronous region thickening late in relation to overall myocardial thickening.

27. A method of detecting intraventricular dyssynchrony, including:

a) characterizing a series of source images of a ventricle of a beating heart over at least one cardiac cycle as a corresponding series of contour shape images representative of movement of muscle boundaries of the ventricle over time;

b) classifying the series of contour shape images in at least one of a plurality of dyssynchronous subclasses or in a non-dyssynchronous class, each dyssynchronous subclass associated with a different region of the ventricle; and

e) pre-diagnosing the ventricle as having at least one dyssynchronous region or as non-dyssynchronous based at least in part on the classifying.

28. The method of claim 27, the classifying in b) further including:

d) normalizing the series of contour shape images;

e) processing the series of contour shape images into a lower dimensional representation of principal and discriminative features of the series of contour shape images to form a classifier sample; and

f) classifying the classifier sample in at least one dyssynchronous subclass or the non-dyssynchronous class.

29. The method of claim 28, the classifying in f) further including:

g) projecting the classifier sample in a ventricle classification space defining each dyssynchronous subclass by a corresponding dyssynchronous subclass mean and defining the non-dyssynchronous class by a non-dyssynchronous class mean, the classification sample being represented in dimensions corresponding to dimensions for the ventricle classification space;

h) determining if the classification sample is closer to any dyssynchronous subclass means or the non-dyssynchronous class mean;

i) classifying the classifier sample in each dyssynchronous subclass where the classification sample is closer to the corresponding dyssynchronous subclass mean than the non-dyssynchronous class mean; and

j) classifying the classifier sample in the non-dyssynchronous class where the classification sample is closer to the non-dyssynchronous class mean than any dyssynchronous subclass mean.

30. A method of detecting intraventricular dyssynchrony, including:

a) capturing movement of a ventricle of a beating heart over at least one cardiac cycle in a series of source images for a subject;

b) extracting muscle boundaries for each source image of the series of source images to define a corresponding series of subject contour shape images with subject contour shapes corresponding to the muscle boundaries;

c) deforming each subject contour shape in each subject contour shape image to track movement of the muscle boundaries for the ventricle over time in the subject series of contour shape images;

d) observing the subject series of contour shape images;

e) comparing the subject series of contour shape images to other series of contour shape images from other cases; and

f) pre-diagnosing the subject ventricle as dyssynchronous or non-dyssynchronous based at least in part on the observing and comparing.

31. The method of claim 30 wherein the process is used in conjunction with teaching diagnosis of intraventricular dyssynchrony.

32. The method of claim 30 wherein the process is used as a diagnostic aid by a practitioner having limited experience in diagnosis of intraventricular dyssynchrony.

33. The method of claim 30 wherein the process is used as a diagnostic aid by a more general practitioner to determine whether or not referral to a specialist is advisable.

34. The method of claim 30 wherein at least the extracting in b) and the deforming in c) are performed at a different location than the capturing in a), further including:

g) communicating the series of source images for the subject from a remote location to a central location via a communication network.

35. The method of claim 34 wherein the communication network includes the Internet.

36. The method of claim 34, further including:

h) communicating the subject series of contour shape images and the other series of contour shape images from the central location to the remote location via the communication network.

37. An apparatus for detecting intraventricular dyssynchrony, including:

a contour image processor extracting muscle boundaries for each source image of a series of source images of a ventricle of a beating heart over at least one cardiac cycle to define a corresponding series of contour shape images with contour shapes corresponding to muscle boundaries of the ventricle using a Fourier descriptor model and deforming each contour shape in each contour shape image based at least in part on image forces, external forces, and internal forces to track movement of the muscle boundaries for the ventricle over time in the series of contour shape images; and

an image classifier, in operative communication with the contour image processor, classifying the series of contour shape images in a dyssynchronous class or a non-dyssynchronous class.

38. The apparatus of claim 37, further including:

a source image processor, in operative communication with the contour image processor, receiving initialization information from an observer viewing a first source image of the series of source images to define a plurality of points associated with the muscle boundaries.

39. The apparatus of claim 37 wherein the contour image processor defines edges of the contour shapes in each contour shape image using an optimal zero crossing operator model to determine edge-strength vectors and gradient vectors for associated pixels of the contour shape images, deforms edges of the contour shapes in each contour shape image using an averaging mask in conjunction with a range of pixels along the edges defined by the optimal zero crossing operator model, determines a length of the contour shapes in each contour shape image using an approximation model, and determines edge-based image forces for the contour shapes in each contour shape image based at least in part on the determined length for the corresponding contour shape.

40. The apparatus of claim 37 wherein the contour image processor normalizes the scale of each Fourier descriptor associated with the Fourier descriptor model and determines external forces for each contour shape image to limit contour shape variations in consecutive contour shape images based at least in part on a predetermined value for a variance parameter.

41. The apparatus of claim 37 wherein the contour image processor determines a length of the contour shapes in each contour shape image using an approximation model and determines internal tension forces to minimize contour shape length and avoid self-crossings as contour shapes are deformed.

42. An apparatus for detecting intraventricular dyssynchrony, including:

an image contour processor characterizing a series of source images of a ventricle of a beating heart over at least one cardiac cycle as a corresponding series of contour shape images representative of movement of muscle boundaries of the ventricle over time using a Fourier descriptor model; and

an image classifier, in operative communication with the contour image processor, classifying the series of contour shape images in a dyssynchronous class or a non-dyssynchronous class.

43. The apparatus of claim 42 wherein the image classifier normalizes the series of contour shape images, processes the series of contour shape images into a lower dimensional representation of principal and discriminative features of the series of contour shape images to form a classifier sample, and classifies the classifier sample in the dyssynchronous class or the non-dyssynchronous class.

44. The apparatus of claim 43 wherein the image classifier projects the classifier sample in a ventricle classification space defining the dyssynchronous class by a dyssynchronous class mean and defining the non-dyssynchronous class by a non-dyssynchronous class mean, the classification sample being represented in dimensions corresponding to dimensions for the ventricle classification space, determines if the classification sample is closer to the dyssynchronous class mean or the non-dyssynchronous class mean, and classifies the classifier sample in the dyssynchronous class or the non-dyssynchronous class based on which corresponding class mean is closest to the classifier sample.

45. An apparatus for detecting intraventricular dyssynchrony, including:

a contour image processor characterizing a series of source images of a ventricle of a beating heart over at least one cardiac cycle as a corresponding series of contour shape images representative of movement of muscle boundaries of the ventricle over time; and

an image classifier classifying the series of contour shape images in at least one of a plurality of dyssynchronous subclasses or in a non-dyssynchronous class, each dyssynchronous subclass associated with a different region of the ventricle.

46. The apparatus of claim 45 wherein the contour image processor normalizes the series of contour shape images, processes the series of contour shape images into a lower dimensional representation of principal and discriminative features of the series of contour shape images to form a classifier sample.

47. The apparatus of claim 46 wherein the image classifier classifies the classifier sample in at least one dyssynchronous subclass or the non-dyssynchronous class.

48. The apparatus of claim 45 wherein the image classifier projects the classifier sample in a ventricle classification space defining each dyssynchronous subclass by a corresponding dyssynchronous subclass mean and defining the non-dyssynchronous class by a non-dyssynchronous class mean, the classification sample being represented in dimensions corresponding to dimensions for the ventricle classification space, determines if the classification sample is closer to any dyssynchronous subclass means or the non-dyssynchronous class mean, classifies the classifier sample in each dyssynchronous subclass where the classification sample is closer to the corresponding dyssynchronous subclass mean than the non-dyssynchronous class mean, and classifies the classifier sample in the non-dyssynchronous class where the classification sample is closer to the non-dyssynchronous class mean than any dyssynchronous subclass mean.

49. A method for monitoring an internal area of a subject, including:

a) receiving initialization information from an observer viewing a first source image of a series of source images of the internal area over a predetermined time to define a plurality of points associated with boundaries within the internal area;

b) extracting boundaries for each source image of the series of source images to define a corresponding series of contour shape images with contour shapes corresponding to the boundaries using a Fourier descriptor model; and

c) deforming each contour shape in each contour shape image based at least in part on image forces, external forces, and internal forces to track movement of the boundaries within the internal area over time in the series of contour shape images.

50. An apparatus for monitoring an internal area of a subject, including:

a contour image processor extracting boundaries for each source image of a series of source images of the internal area over a predetermined time to define a corresponding series of contour shape images with contour shapes corresponding to boundaries within the internal area using a Fourier descriptor model and deforming each contour shape in each contour shape image based at least in part on image forces, external forces, and internal forces to track movement of the boundaries within the internal area over time in the series of contour shape images; and

an image classifier, in operative communication with the contour image processor, classifying the series of contour shape images in a first class or a second class based at least in part on detection of predetermined discriminating characteristics for the first and second classes in the contour shape images.