System and method for ground glass nodule (GGN) segmentation
View Patent ↗A system and method for ground glass nodule (GGN) segmentation is provided. The method comprises: selecting a point in a medical image, wherein the point is located in a GGN; defining a volume of interest (VOI) around the point, wherein the VOI comprises the GGN; removing a chest wall from the VOI; obtaining an initial state for a Markov random field; and segmenting the VOI, wherein the VOI is segmented using the Markov random field.
1. A method for ground glass nodule (GGN) segmentation, comprising:
selecting a point in a medical image, wherein the point is located in a GGN;
defining a volume of interest (VOI) around the point, wherein the VOI comprises the GGN;
removing a chest wall from the VOI;
obtaining an initial state for a Markov random field; and
segmenting the VOI, wherein the VOI is segmented using the Markov random field.
2. The method of claim 1 , further comprising:
acquiring the medical image.
3. The method of claim 2 , wherein the medical image is acquired using a computed tomographic (CT) imaging technique.
4. The method of claim 1 , wherein the point is automatically selected.
5. The method of claim 1 , wherein the point is manually selected.
6. The method of claim 1 , further comprising:
detecting the GGN using a computer-aided GGN detection technique.
7. The method of claim 1 , further comprising:
detecting the GGN manually.
8. The method of claim 1 , wherein the GGN is one of a pure GGN and a mixed GGN.
9. The method of claim 1 , wherein the chest wall is removed by performing a region growing.
10. The method of claim 1 , further comprising:
defining one of a shape and a size of the VOI.
11. The method of claim 1 , wherein the initial state for the Markov random field is obtained by performing a region growing on the VOI after the chest wall is removed.
12. The method of claim 1 , wherein the step of segmenting the VOI using the Markov random field comprises:
defining a posteriori probability for the VOI; and
labeling each pixel in the VOI using a maximum of the posteriori probability, wherein each pixel in the VOI is labeled as one of a GGN and a background.
13. The method of claim 12 , wherein the defined posteriori probability is computed by
P ( L|F )∝ P ( F|L ) P ( L ).
14. The method of claim 12 , wherein the step of labeling each pixel is computed by
l
-
x
(
i
)
=
arg
min
l
∈
{
g
,
b
}
{
U
(
g
,
i
-
1
)
+
1
2
σ
2
[
f
(
x
_
)
-
μ
g
]
,
U
(
b
.
i
-
1
)
+
1
2
σ
2
[
f
(
x
_
)
-
μ
b
]
}
.
15. The method of claim 12 , wherein the labeling comprises:
scanning the VOI until a convergence is reached.
16. The method of claim 1 , further comprising:
performing a shape analysis to remove blood vessels attached to or near the GGN after the VOI has been segmented using the Markov random field.
17. The method of claim 1 , further comprising:
displaying the VOI segmented using the Markov random field.
18. A system for ground glass nodule (GGN) segmentation, comprising:
a memory device for storing a program;
a processor in communication with the memory device, the processor operative with the program to:
define a volume of interest (VOI) around a GGN using data associated with a medical image of a lung;
remove a chest wall from the VOI;
obtain an initial state for a Markov random field; and
segment the VOI, wherein the VOI is segmented using the Markov random field.
19. The system of claim 18 , wherein the processor is further operative with the program code to:
acquire the medical image, wherein the medical image is acquired using a computed tomographic (CT) imaging technique.
20. The system of claim 18 , wherein the chest wall is removed by performing a region growing.
21. The system of claim 18 , wherein the initial state for the Markov random field is obtained by performing a region growing on the VOI after the chest wall is removed.
22. The system of claim 18 , wherein the processor is further operative with the program code when segmenting the VOI using the Markov random field to:
define a posterirori probability for the VOI; and
label each pixel in the VOI using a maximum of the posteriori probability, wherein each pixel in the VOI is labeled as one of a GGN and a background.
23. The system of claim 22 , wherein the defined posteriori probability is computed by
P ( L|F )∝ P ( F|L ) P ( L ).
24. The system of claim 22 , wherein the step of labeling each pixel is computed by
l
-
x
(
i
)
=
arg
min
l
∈
{
g
,
b
}
{
U
(
g
,
i
-
1
)
+
1
2
σ
2
[
f
(
x
_
)
-
μ
g
]
,
U
(
b
.
i
-
1
)
+
1
2
σ
2
[
f
(
x
_
)
-
μ
b
]
}
.
25. The system of claim 18 , wherein the processor is further operative with the program code to:
perform a shape analysis to remove blood vessels attached to the GGN in the VOI segmented using the Markov random field.
26. The system of claim 18 , wherein the processor is further operative with the program code to:
display the VOI segmented using the Markov random field, wherein the GGN is visible.
27. A computer program product comprising a computer useable medium having computer program logic recorded thereon for ground glass nodule (GGN) segmentation, the computer program logic comprising:
program code for selecting a point in a medical image, wherein the point is located in or near a GGN;
program code for defining a volume of interest (VOI) around the point, wherein the VOI comprises the GGN;
program code for removing a chest wall from the VOI;
program code for obtaining an initial state for a Markov random field; and
program code for segmenting the VOI, wherein the VOI is segmented using the Markov random field.
28. A system for ground glass nodule (GGN) segmentation, comprising:
means for selecting a point in a medical image, wherein the point is located in a GGN;
means for defining a volume of interest (VOI) around the point, wherein the VOI comprises the GGN;
means for removing a chest wall from the VOI;
means for obtaining an initial state for a Markov random field; and
means for segmenting the VOI, wherein the VOI is segmented using the Markov random field.
29. A method for ground glass nodule (GGN) segmentation in pulmonary computed tomographic (CT) volumes using a Markov random field, comprising:
selecting a GGN from data associated with a pulmonary CT volume;
defining a volume of interest (VOI) around the GGN;
removing a chest wall from the VOI by performing a region growing on the VOI;
obtaining an initial state for an iterated condition mode (ICM) procedure by segmenting the VOI after the chest wall is removed; and
segmenting the VOI using a Markov random field, wherein the segmentation comprises:
defining a posteriori probability for the VOI; and
performing the ICM procedure, wherein the ICM procedure comprises labeling each pixel in the VOI using a maximum of the posteriori probability, wherein each pixel in the VOI is labeled as one of a GGN and a background until each pixel in the VOI is labeled.
30. The method of claim 29 , wherein the defined posteriori probability is computed by
P ( L|F )∝ P ( F|L ) P ( L ).
31. The method of claim 29 , wherein the step of labeling each pixel during the ICM procedure is computed by
l
-
x
(
i
)
=
arg
min
l
∈
{
g
,
b
}
{
U
(
g
,
i
-
1
)
+
1
2
σ
2
[
f
(
x
_
)
-
μ
g
]
,
U
(
b
.
i
-
1
)
+
1
2
σ
2
[
f
(
x
_
)
-
μ
b
]
}
.
32. The method of claim 29 , wherein the ICM procedure begins from the initial state.