CT/MRI heart isolation using a graph cut algorithm
A method and related system for automatically and efficiently isolating the heart in Computer Tomography (CT) or Magnetic Resonance Imaging cardiac scans is disclosed. The method involves segmenting a heart within a set of volumetric data. In accordance with one aspect of the present invention, the set of volumetric data is processed to determine the minimum value of an energy function having a first term, a second term and a third term. The heart is segmented based on the processing of the set of volumetric data.
1. A method for segmenting a heart within a set of volumetric data by a processor, comprising:
processing the set of volumetric data to determine a minimum value of an energy function including a blob energy term that provides a segmentation with a bias towards a convex shape; and,
segmenting the surface of the heart based on the processing of the set of volumetric data by the processor.
2. The method as claimed in claim 1 , wherein the blob energy term can be expressed as
E
blob
(
f
)
=
∑
p
,
q
∈
neighbors
B
p
,
q
(
f
(
p
)
,
f
(
q
)
;
C
)
,
wherein B p,q is a cost associated with an angle between line segments pq and pC where C is a center of a seed region and p and q are nodes in a graph.
3. The method as claimed in claim 1 , wherein the energy function is expressed as E(ƒ)=E smooth (ƒ)+E data (ƒ)+E blob (ƒ) and the first term E smooth (ƒ) measures how much a pair of neighboring pixels look alike in the set of data and the second term E data (ƒ) measures how much a value at a pixel matches an expected appearance of an object.
4. The method as claimed in claim 3 , wherein the term E blob (ƒ) favors edges that are orthogonal in any radial direction.
5. The method as claimed in claim 2 , wherein E blob (ƒ) can be expressed as:
E
blob
(
f
)
=
∑
p
,
q
∈
neighbors
B
p
,
q
(
f
(
p
)
,
f
(
q
)
;
C
)
wherein
B
p
,
q
(
f
(
p
)
,
f
(
q
)
,
C
)
=
cos
2
(
pq
,
pC
)
*
δ
(
f
(
p
)
,
f
(
q
)
)
if
cos
(
pq
,
pC
)
<
0
and
B
p
,
q
(
f
(
p
)
,
f
(
q
)
,
C
)
=
0
otherwise
.
6. The method as claimed in claim 1 , further comprising automatically placing seeds inside the heart in the set of volumetric data.
7. A computer system for segmenting a heart within volumetric data, the computer system comprising:
a processor; and
a memory in communications with the processor, the memory including program code executable by the processor to perform the following steps:
processing the set of volumetric data to determine a minimum value of an energy function including a blob energy term that provides a segmentation with a bias towards a convex shape; and,
segmenting a surface of the heart based on the processing of the set of volumetric data.
8. The system as claimed in claim 7 , wherein the blob energy term can be expressed as
E
blob
(
f
)
=
∑
p
,
q
∈
neighbors
B
p
,
q
(
f
(
p
)
,
f
(
q
)
;
C
)
,
wherein B p,q is a cost associated with an angle between line segments pq and pC where C is a center of a seed region and p and q are nodes in a graph.
9. The system as claimed in claim 7 , wherein the energy function is expressed as E(ƒ)=E smooth (ƒ)+E data (ƒ)+E blob (ƒ) and the first term E smooth (ƒ) measures how much a pair of neighboring pixels look alike in the set of data and the second term E data (ƒ) measures how much a value at a pixel matches an expected appearance of an object.
10. The system as claimed in claim 9 , wherein the term E blob (ƒ) favors edges that are orthogonal in any radial direction.
11. The system as claimed in claim 8 , wherein E blob (ƒ)can be expressed as:
E
blob
(
f
)
=
∑
p
,
q
∈
neighbors
B
p
,
q
(
f
(
p
)
,
f
(
q
)
;
C
)
wherein
B
p
,
q
(
f
(
p
)
,
f
(
q
)
,
C
)
=
cos
2
(
pq
,
pC
)
*
δ
(
f
(
p
)
,
f
(
q
)
)
if
cos
(
pq
,
pC
)
<
0
and
B
p
,
q
(
f
(
p
)
,
f
(
q
)
,
C
)
=
0
otherwise
.
12. The system as claimed in claim 7 , further comprising means for automatically placing seeds inside the heart in the set of volumetric data.
13. A method for segmenting with a processor an object within a set of volumetric data, comprising:
processing by the processor of the set of volumetric data to determine in a graph cuts segmentation a minimum value of an energy function having a first term, a second term and a third term;
wherein the third term favors edges in a graph that are orthogonal in a radial direction; and,
segmenting the surface of the object based on the processing of the set of volumetric data.
14. The method as claimed in claim 13 , wherein the first term measures how much a pair of neighboring pixels look alike in the set of data and the second term measures how much the value at a pixel resembles the expected appearance of an object.
15. The system as claimed in claim 13 , wherein the third term is:
E
blob
(
f
)
=
∑
p
,
q
∈
neighbors
B
p
,
q
(
f
(
p
)
,
f
(
q
)
;
C
)
wherein
B
p
,
q
(
f
(
p
)
,
f
(
q
)
,
C
)
=
cos
2
(
pq
,
pC
)
*
δ
(
f
(
p
)
,
f
(
q
)
)
if
cos
(
pq
,
pC
)
<
0
and
B
p
,
q
(
f
(
p
)
,
f
(
q
)
,
C
)
=
0
otherwise
.
16. The method as claimed in claim 13 , wherein the object is a heart.
17. The method as claimed in claim 16 , further comprising automatically placing seeds inside the heart in the set of volumetric data.
18. The method as claimed in claim 17 , wherein the step of automatically seeding involves expanding a virtual balloon in the heart in the volumetric data until a side of the balloon has contacted the heart.
19. The method as claimed in claim 17 , wherein the step of automatically seeding involves determining a volumetric data barycenter weighted by intensity.
20. The method as claimed in claim 6 , further comprising:
determining a volumetric data barycenter weighted by intensity; and
inflating a virtual balloon in the heart in the volumetric data until a side of the balloon has contacted a wall of the heart.