Methods and systems for segmentation in echocardiography
Methods and systems for segmentation in echocardiography are provided. One method includes obtaining echocardiographic images and defining a search space within the echocardiographic images using a pair of one-dimensional (1D) profiles. The method also includes using an energy based function constrained by non-local temporal priors within the defined search space to automatically segment a contour of a cardiac structure with the 1D profiles.
1. A method for automatic segmentation of a cardiac structure, comprising:
obtaining echocardiographic images;
defining a search space within the echocardiographic images using a pair of one-dimensional (1D) profiles; and
using an energy based function constrained by non-local temporal priors within the defined search space to automatically segment a contour of the cardiac structure with the 1D profiles, wherein using the energy based function comprises penalizing a spatial smoothness of the acceleration of the 1D profiles.
2. The method of claim 1 , wherein using the energy based function comprises using regional statistics.
3. The method of claim 1 , wherein the cardiac structure is an interventricular septum (IVS) and further comprising using a shape prior on IVS width.
4. The method of claim 1 , wherein the non-local temporal priors comprise non-local velocity interactions.
5. The method of claim 1 , wherein using the energy based function comprises using current and previous frames of echocardiographic images to update the 1D profiles.
6. The method of claim 1 , wherein the non-local temporal priors comprise motion parameters from a previous segmentation iteration.
7. The method of claim 1 , wherein using the energy based function comprises using non-local temporal cues.
8. The method of claim 1 , wherein defining the search space comprises using a non-local means based de-noising filtering.
9. The method of claim 1 , wherein defining the search space comprises using Markov Random Fields based clustering.
10. The method of claim 1 , wherein the echocardiographic images comprise a plurality of B-mode image sequences of a heart.
11. The method of claim 1 , wherein using the energy based function comprises minimizing the following function over the defined search space:
E
frm
(
f
,
g
,
μ
up
,
μ
s
,
μ
dn
)
=
∑
i
=
1
K
[
∫
R
i
s
(
I
-
μ
i
s
)
2
ⅆ
y
ⅆ
x
+
∫
R
i
up
(
I
-
μ
i
up
)
2
ⅆ
y
ⅆ
x
+
∫
R
i
dn
(
I
-
μ
i
dn
)
ⅆ
y
ⅆ
x
]
+
λ
width
∫
a
b
(
f
+
w
-
g
)
2
ⅆ
x
+
λ
smooth
∫
a
b
(
1
+
(
f
′
)
2
+
1
+
(
g
′
)
2
)
ⅆ
x
where I is the echocardiographic images, f and g are the boundaries of the cardiac structure, w is the width of the cardiac structure and μ is a mean statistic.
12. The method of claim 11 , wherein the function includes a temporal term and is minimized to:
E temp n =E frm +λ T ∫ a b ∫ a b w ( x,y )(( A n ( x )− A n ( y )) 2 +( B n ( x )− B n ( y )) 2 ) dxdy
wherein an acceleration is defined as An=fn−2−2fn−1 and Bn=gn−2−2gn−1+gn, and w is the width of the cardiac structure.
13. A non-transitory computer readable storage medium for segmentation of a cardiac structure using a processor, the non-transitory computer readable storage medium including instructions to command the processor to:
obtain echocardiographic images;
define a search space within the echocardiographic images using a pair of one-dimensional (1D) profiles; and
use an energy based function constrained by non-local temporal priors within the defined search space to automatically segment a contour of the cardiac structure with the 1D profiles, wherein the instructions command the processor to penalize a spatial smoothness of the acceleration of the 1D profiles in the energy based function.
14. The non-transitory computer readable storage medium of claim 13 , wherein the cardiac structure is an interventricular septum (IVS) and the instructions command the processor to include at least one of regional statics, a shape prior on IVS width or non-local temporal cues in the energy based function.
15. The non-transitory computer readable storage medium of claim 13 , wherein the instructions command the processor to use at least one of a non-local means based de-noising filtering or Markov Random Fields based clustering when defining the search space.
16. An ultrasound system comprising:
an ultrasound probe configured to acquire echocardiographic images;
a processor having a segmentation module configured to define a search space within the echocardiographic images using a pair of one-dimensional (1D) profiles and use an energy based function constrained by non-local temporal priors within the defined search space to automatically segment a contour of a cardiac structure with the 1D profiles, wherein the processor is configured to penalize a spatial smoothness of the acceleration of the 1D profiles in the energy based function; and
a display configured to display the segmented contour of the cardiac structure.
17. The ultrasound system of claim 16 , wherein the cardiac structure is an interventricular septum (IVS) and the processor is further configured to include at least one of regional statics, a shape prior on IVS width or non-local temporal cues in the energy based function.
18. The ultrasound system of claim 16 , wherein the processor is configured to use at least one of a non-local means based de-noising filtering or Markov Random Fields based clustering when defining the search space.