Iterative reconstruction for X-ray computed tomography using prior-image induced nonlocal regularization
Disclosed is a method for performing X-ray Computed Tomography scanning, the method including acquiring a plurality of images of an object, obtaining an initial image from the plurality of images, calculating NonLocal weight of the initial image, utilizing a current image estimation and registered prior image, performing a successive over-relaxation optimization to yield a new image estimation with an intensity of the new image estimation equal or greater than zero, performing a cycle update, generating an image of the object utilizing the new image estimation obtained from the optimization, and outputting a resultant image.
1. A method for operating an X-ray Computed Tomography (CT) scanning device, the method comprising:
acquiring, by the CT scanning device, a plurality of images of an object;
obtaining an initial image from the plurality of images;
performing a calculation of NonLocal (NL) weight utilizing a current image estimation and a registered prior image;
performing a Successive Over-Relaxation (SOR) optimization to yield a new image estimation, with an intensity of the new image estimation being equal or greater than zero;
performing a cycle update that repeats the obtaining of the initial image, the performing of the NL weight calculation, and the SOR optimization;
generating an image of the object utilizing the new image estimation obtained from the SOR optimization; and
outputting a resultant image when a stop criterion is met.
2. The method of claim 1 , further comprising performing a prior-image induced NL regularization utilizing a potential function of:
R
PINL
(
μ
)
=
∑
j
∑
k
∈
N
j
w
(
k
,
j
)
(
μ
j
-
μ
prior
,
k
reg
)
2
,
where μ prior reg denotes the registered prior image between a prior image μ prior and a current image, and N j denotes a search-window.
3. The method of claim 1 , wherein the SOR optimization yields a new image estimation according to:
μ
~
j
m
+
1
=
H
j
T
Σ
-
1
r
m
+
S
j
μ
j
m
+
β
∑
k
∈
N
j
w
(
k
,
j
)
μ
prior
,
k
reg
S
j
+
β
∑
k
∈
N
j
w
(
k
,
j
)
,
μ
j
m
+
1
=
max
{
0
,
(
1
-
ω
)
μ
j
m
+
ω
μ
~
j
m
+
1
}
with r m =y−Hμ m being a residual between obtained sinogram data and an estimated one in an m-th iteration, S j =H j T Σ −1 H j , and a relax factor ω satisfies 0≦ω≦1.
4. The method of claim 1 , wherein the output resultant image is utilized for one of lung nodule surveillance, image-guided biopsy needle, and volumetric Cone Beam CT imaging performed during a radiation therapy course.