Medical imaging apparatus and method of reconstructing medical image
View Patent ↗A method of reconstructing a tomographic image in a medical imaging apparatus includes: acquiring a first image of an object and a second image of the object corresponding to an image to be reconstructed; determining, in the first image, a reference region corresponding to a unit region in the second image; and updating data in the unit region based on data contained in the reference region. Thus, a medical image having both high spatial resolution and high temporal resolution is reconstructed to have substantially reduced artifacts.
1. A method of reconstructing a tomographic image in a medical imaging apparatus, the method comprising:
acquiring a first image of an object and a second image of the object, the second image corresponding to the tomographic image to be reconstructed;
determining, in the first image, a reference region corresponding to a unit region in the second image;
updating data in the unit region in the second image based on data contained in the reference region in the first image; and
reconstructing the second image by using the updated data,
wherein the first image and the second image are acquired, respectively, at phases different from each other while the object is moving.
2. The method of claim 1 , wherein the reference region has similar features to the unit region.
3. The method of claim 1 , further comprising acquiring electrocardiography (ECG) data of the object comprising a diastolic period and a systolic period,
wherein the acquiring the first image comprises acquiring the first image based on diastolic data of the object obtained during the diastolic period, and
the acquiring the second image comprises acquiring the second image based on systolic data of the object obtained during the systolic period.
4. The method of claim 1 , wherein the updating the data in the unit region comprises:
updating the second image by minimizing a cost function based on a patch-based low rank regularization.
5. The method of claim 4 , wherein the reconstructing the second image comprises:
generating a matrix v p based on data included in the unit region and the reference region;
determining eigenvalues of the generated matrix v p ; and
updating the data in the unit region based on the determined eigenvalues.
6. The method of claim 5 , wherein the patch-based low rank regularization is calculated as:
Ψ
(
x
,
R
)
=
∑
p
=
1
N
p
λ
p
v
p
h
μ
,
v
,
v
p
h
μ
,
v
∑
k
=
1
Rank
(
v
p
)
h
μ
,
v
(
σ
k
(
v
p
)
)
=
min
w
p
{
1
μ
v
p
-
w
p
F
2
+
w
p
g
μ
,
v
}
where
w
p
g
μ
,
v
=
∑
k
=
1
Rank
(
v
p
)
g
μ
,
v
(
σ
k
(
v
p
)
)
,
R represents a similar patch extraction operator,
Np is the number of patch-based low rank executions, which may be equal to the number of pixels,
Ψo(x) is the patch-based low rank regularization term,
g μ,ν is a value which makes equation
h
μ
,
ν
(
t
)
=
min
s
{
s
-
t
2
/
μ
+
g
μ
,
ν
(
s
)
}
and
shrink ν ( t ,μ)=max{0,| t|−μ|t| ν−1 }t/|t| equal
p is a number of pixels or voxels,
v p is the generated matrix,
σ k (v p ) is a k-th largest singular value in the generated matrix v p ,
h μ,ν is a Huber function, and
μ, ν are parameters of the Huber function.
7. The method of claim 6 , wherein the minimizing the cost function comprises:
minimizing the cost function with respect to a matrix w p ; and
minimizing the cost function with respect to an image x based on a result obtained by minimizing the cost function with respect to the matrix w p .
8. The method of claim 7 , wherein the minimizing the cost function with respect to the matrix w p and the minimizing of the cost function with respect to the image x are iteratively performed.
9. The method of claim 4 , wherein, in the updating the data in the unit region, the cost function is minimized based on a shrinkage operator defined as:
shrink ν ( t ,μ)=max{0,| t|−μ|t| ν−1 }t/|t|.
10. A non-transitory computer-readable recording medium having recorded thereon a program which, when executed by a computer, causes the computer to execute the method of claim 1 .
11. A medical imaging apparatus comprising:
an image processor configured to acquire a first image of an object and a second image of the object, the second image corresponding to the tomographic image to be reconstructed, determine a reference region, in the first image, corresponding to a unit region in the second image, update data in the unit region in the second image based on data included in the reference region in the first image, and reconstruct the second image by using the updated data; and
a display configured to display the second image which has been reconstructed according to a result obtained by updating the data in the unit region,
wherein the first image and the second image are acquired, respectively, at phases different from each other while the object is moving.
12. The apparatus of claim 11 , wherein the reference region has similar features to the unit region.
13. The apparatus of claim 11 , further comprising an electrocardiography (ECG) data obtainer configured to acquire ECG data of the object comprising a diastolic period and a systolic period,
wherein the image processor is configured to acquire the first image based on diastolic data of the object obtained during the diastolic period, and acquire the second image based on systolic data of the object obtained during the systolic period.
14. The apparatus of claim 11 , wherein the image processor is configured to update the second image by minimizing a cost function based on a patch-based low rank regularization.
15. The apparatus of claim 14 , wherein the image processor is configured to generate a matrix v p based on data included in the unit region and the reference region, determine eigenvalues of the generated matrix v p , and update the data in the unit region based on the determined eigenvalues.
16. The apparatus of claim 15 , wherein the image processor is configured to calculate the patch-based low rank regularization as:
Ψ
(
x
,
R
)
=
∑
p
=
1
N
p
λ
p
v
p
h
μ
,
v
,
v
p
h
μ
,
v
∑
k
=
1
Rank
(
v
p
)
h
μ
,
v
(
σ
k
(
v
p
)
)
=
min
w
p
{
1
μ
v
p
-
w
p
F
2
+
w
p
g
μ
,
v
}
where
w
p
g
μ
,
v
=
∑
k
=
1
Rank
(
v
p
)
g
μ
,
v
(
σ
k
(
v
p
)
)
,
R represents a similar patch extraction operator,
Np is the number of patch-based low rank executions, which may be equal to the number Of pixels,
Ψo(x) is the patch-based low rank regularization term,
g μ,ν is a value which makes equation
h
μ
,
ν
(
t
)
=
min
s
{
s
-
t
2
/
μ
+
g
μ
,
ν
(
s
)
}
and
shrink ν ( t ,μ)=max{0,| t|−μ|t| ν−1 }t/|t| equal,
p is a number of pixels or voxels,
v p is the generated matrix,
σ k (v p ) is a k-th largest singular value of the generated matrix v p ,
h μ,ν is a Huber function, and
μ, ν are parameters of the Huber function.
17. The apparatus of claim 11 , wherein the image processor is configured to minimize the cost function based on a shrinkage operator defined as:
shrink ν ( t ,μ)=max{0,| t|−μ|t| ν−1 }t/|t|.