X-ray dual-energy CT reconstruction method

Abstract

The present disclosure relates to a self-prior information based X-ray dual-energy CT reconstruction method, which can utilize information inherent in data to provide a prior model, thereby obtaining a reconstructed image with a high quality. The X-ray dual-energy CT reconstruction method according to the present disclosure comprises: (a) rating an energy spectrum and establishing a dual-energy lookup table; (b) collecting high-energy data pH and low-energy data pL of a dual-energy CT imaging system using a detector of the dual-energy CT imaging system; (c) obtaining projection images R1 and R2 of scaled images r1 and r2 according to the obtained high-energy data pH and low-energy data pL; (d) reconstructing the scaled image r2 using a first piece-wise smooth constraint condition and thereby obtaining an electron density image; and (e) reconstructing the scaled image r1 using a second piece-wise smooth constraint condition and thereby obtaining an equivalent atomic number image. In the present disclosure, the noise in the dual-energy reconstructed image can be effectively prohibited while keeping the resolution by effectively using information inherent in data.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to Chinese Patent Application No. 201310318608.4 filed Jul. 26, 2013, the disclosure of which are incorporated herein by reference in its entirety.

TECHNICAL FIELD

The present disclosure relates to a Computer Tomography (CT) reconstruction method, and in particular, to self-prior information based X-ray dual-energy CT reconstruction methods.

BACKGROUND

A CT image contrast is largely related to an X-ray source energy spectrum distribution used for scanning. The traditional CT uses a ray source with an energy spectrum distribution for imaging. Sometimes, information ambiguity may occur, which results in two different materials being completely the same on a CT image. The dual-energy CT uses two energy spectrums with different distributions for imaging an object, which can eliminate the information ambiguity in a single energy spectrum. The dual-energy X-ray CT imaging technology can utilize difference between attenuations of a material at different energy levels to obtain distribution information about multiple physical characteristic parameters of an object, for example, an electron density distribution, an equivalent atomic number distribution, and single-energy attenuation images at multiple energy levels. Thus, the dual-energy X-ray CT can be used for calibration of ray hardening of the traditional CT, acquisition of a clinical energy spectrum image with a high contrast, detection of particular and dangerous goods in industry and safety inspection and so on. Compared with the traditional X-ray CT imaging technology, breakthroughs of the dual-energy CT in its imaging function are of great significance in applications such as a medical diagnosis technology, lossless detection and safety inspection etc., and thus attract more and more attention in recent years. In addition, the dual-energy X-ray CT reconstruction method is a search hotspot currently.

Currently, there are three methods for dual-energy CT reconstruction as follows: (1) a post-processing method, in which attenuation coefficient distribution images are reconstructed from low-energy data and high-energy data respectively, then synthesis calculation is performed on the two attenuation coefficient distribution images, and thereby, a single-energy image or a distribution image of an energy independent physical quantity (for example, an atomic number, an electron density) can be obtained; (2) a pre-processing method, in which an energy dependent signal and an energy independent signal are parsed from low-energy data and high-energy data (i.e., the so-called dual-energy decomposition), wherein, the parsed signals belong to a projection region, and then, the parsed signals are reconstructed using a traditional CT reconstruction method; and (3) a synthesis iterative method, in which low-energy data and high-energy data are reconstructed directly using an iterative method. At present, the pre-processing method is widely used, because on one hand, the pre-processing method is more accurate than the post-processing method, and can better eliminate an effect of an x-ray broad spectrum; and on the other hand, the pre-processing method has a less calculation amount than the synthesis iterative method.

At present, with respect to the dual-energy decomposition, there are two decomposition methods, i.e., material based decomposition and dual-effect decomposition (for example, with reference to a non-patent document 1). However, in the dual-effect decomposition method, the equivalent atomic number reconstructed image generally has a poor signal-to-noise ratio, and in contrast, the electron density image has a better signal-to-noise ratio (for example, with reference to a non-patent document 2). In addition, in the dual-energy CT reconstruction, a MonteCarlo method or an experiential method can be used for estimating energy spectrum data of the dual-energy CT system, and moreover, a lookup table can also be established (for example, with reference to non-patent documents 2 and 3).

Prior Art Documents

• Non-patent document 1: Y. Xing, L. Zhang, X. Duan, J. Cheng, Z. Chen, “A reconstruction method for dual high-energy CT with Mev X-rays,” IEEE Trans Nucl. Sci. vol. 58, no. 2, pp 537-546, 2011;
• Non-patent document 2: Guowei Zhang, dual-energy X-ray imaging algorithm and application research [D], Beijing: Engineering Physics at Tsinghua University, 2008; and
• Non-patent document 3: G. Zhang, Z. Chen, L. Zhang, and J. Cheng, Exact Reconstruction for Dual Energy Computed Tomography Using an H-L Curve Method, 2006 IEEE Nuclear Science Symposium Conference Record, pp. M14-462, 2006.

In addition, in the dual-energy CT reconstruction, two primary physical characteristic parameters are an equivalent atomic number and an electronic density. Since there is strong unbalance in a process of the dual-energy decomposition, it results in amplification of a noise of the dual-energy CT reconstructed image, especially amplification of a noise of the equivalent atomic number distribution.

SUMMARY

The present disclosure is proposed to solve the above problem. The purpose of the present disclosure is to provide dual-energy CT reconstruction methods, which can provide a prior model using information inherent in data, so as to obtain a reconstructed image with a high quality.

The present disclosure provides an X-ray dual-energy CT reconstruction method, comprising:

(a) collecting high-energy data p_H and low-energy data p_L of a dual-energy CT imaging system using a detector of the dual-energy CT imaging system;

(b) obtaining projection images R₁ and R₂ of scaled images r₁ and r₂ according to the obtained high-energy data p_H and low-energy data p_L;

(c) reconstructing the scaled image r₂ using a first piece-wise smooth constraint condition and obtaining a decomposition coefficient a₂; and

(d) reconstructing the scaled image r₁ using a second piece-wise smooth constraint condition and obtaining a decomposition coefficient a₁.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, the scaled images r₁ and r₂ are defined in equation (4) as follows:

$\begin{array}{cc}{r}_1\equiv \mathrm{diag}\left(\frac{1}{{\omega }_1+\varepsilon }\right)a_1{r}_2\equiv \mathrm{diag}\left(\frac{1}{{\omega }_2+\varepsilon }\right)a_2& \left(4\right)\end{array}$

the projection images R₁ and R₂ are defined in equation (6) as follows:R₁≡H′₁r₁ R₂≡H′₂r₂  (6)wherein, H′₁≡Hdiag(ω₁+ε) and H′₂≡Hdiag(ω₂+ε), in which H is a projection matrix, a₁ and a₂ are decomposition coefficients, ε is a vector with small constant coefficients, and ω₁ and ω₂ are vectors which can be selected randomly,

in the step (c), r₂ is reconstructed using the following equation (9) as the first piece-wise smooth constraint condition:

$\begin{array}{cc}\underset{r_2}{\min }{\nabla r_2}_{p}s.t.R_2\equiv H_2^{\prime }{r}_2& \left(9\right)\end{array}$

and in the step (d), r₁ is reconstructed using the following equation (8) as the second piece-wise smooth constraint condition:

$\begin{array}{cc}\underset{r_1}{\min }{\nabla r_1}_{p}s.t.R_1\equiv H_1^{\prime }{r}_1.& \left(8\right)\end{array}$

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, in the step (c), an effective linear attenuation coefficient μ_H at a high energy level is reconstructed according to the high-energy data p_H, ω₂=μ_H is selected, and r₂ is reconstructed using equation (9) as the first piece-wise smooth constraint condition.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, in the above step (d), ω₁=a₂ is set and r₁ is reconstructed using equation (8) as the second piece-wise smooth constraint condition.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, the method further comprises:

obtaining an electronic density image ρ_e according to a₂=r₂×diag(ω₂+ε) and ρ_e=2a₂ using dual-effect decomposition.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, the method further comprises:

obtaining an equivalent atomic number image Z^eff according to

$Z^{\mathrm{eff}}\approx r_1^{\frac{1}{\lambda -1}},\lambda \approx 4$ using dual-effect decomposition.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, the method further comprises:

obtaining an electron density image ρ_e according to a₂=r₂×diag(ω₂+ε) and ρ_e=2a₂ using dual-effect decomposition.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, the method further comprises:

obtaining an equivalent atomic number image Z^eff according to

$Z^{\mathrm{eff}}\approx r_1^{\frac{1}{\lambda -1}},\lambda \approx 4$ using dual-effect decomposition.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, in the step (d), an effective linear attenuation coefficient μ_H at a high energy level is reconstructed according to the high-energy data p_H, ω₁=μ_H is set, and r₁ is reconstructed using the following equation (8) as the second piece-wise smooth constraint condition:

$\begin{array}{cc}\underset{r_1}{\min }{\nabla r_1}_{p}s.t.R_1\equiv H_1^{\prime }{r}_1,& \left(8\right)\end{array}$

then an equivalent atomic number image Z^eff is obtained using a₁=r₁×diag(ω₁+ε) and

$Z^{\mathrm{eff}}\approx r_1^{\frac{1}{\lambda -1}},\lambda \approx 4.$

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, in the steps (c) and (d), r₁ and r₂ are reconstructed using an ART+TV method.

In addition, in the X-ray dual-energy CT reconstruction method according to the present disclosure, in the steps (c) and (d), r₁ and r₂ are reconstructed using a split Bregman method.

Compared with the prior art, the present disclosure has the following effects: (1) the noise in the dual-energy reconstructed image can be effectively prohibited while keeping the resolution by effectively using information inherent in data (for example, ω₁ and ω₂ are selected as μ_H or μ_L, or ω₁ is selected as a₂); (2) the algorithm can conveniently be designed by establishing reconstruction by means of a prior model; and (3) the method is not limited to one scanning method, and is also suitable for different scanning methods such as a fan beam, a cone beam, a circular orbit, a spiral orbit etc., and can increase robustness of iterative reconstruction using this prior method; and (4) compared with a dual-effect decomposition method in the art, a more stable result can be obtained by directly reconstructing a ratio of the coefficients.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flowchart of a self-prior information based dual-energy CT reconstruction method with dual-effect decomposition as an example according to the present disclosure.

FIG. 2 is an image obtained by reconstruction using a self-prior information based dual-energy CT reconstruction method with dual-effect decomposition as an example according to the present disclosure, wherein, (a) is an electron density reconstructed image, and (b) is an equivalent atomic number reconstructed image.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

With respect to “self-prior”, it is a term proposed by the inventor, since a prior model used in the process of reconstructing the following scaled images r₁ and r₂ can be obtained from the data per se, for example, a linear attenuation coefficient result obtained by reconstructing single high-energy data previously using a traditional CT method or ω₁=a₂ (a₂ is also a decomposition coefficient) used when reconstructing a₁ (a decomposition coefficient), i.e., structural information of a₂ is used as prior information. In addition, in the present disclosure, the data per se is processed to obtain reconstruction of a single-energy attenuation image or reconstruction of a₂, which are placed in a prior model. Therefore, this is referred to as utilization of information inherent in the data, and thereby, a reconstructed image with a high quality can be obtained.

The embodiments of the present disclosure will be described below with reference to accompanying drawings.

First of all, an energy spectrum is rated and a dual-energy lookup table is established. Then, a detector of a dual-energy CT imaging system is used to collect high-energy data and low-energy data of the dual-energy CT imaging system. Here, assume that the high-energy data and the low-energy data of the dual-energy CT are p_H and p_L respectively, wherein, p_H and p_L are represented by equations (1) and (2) as follows:−ln ∫w_H(E)exp(−Hμ(E))dE=p_H  (1)−ln ∫w_L(E)exp(−Hμ(E))dE=p_L  (2)wherein, w_H(E) and w_L(E) in the above equations (1) and (2) are normalized high-energy and low-energy energy spectrum distributions, which can be generated in many ways, including two energy spectrums generated by a pseudo-dual-energy and fast-switch X-ray machine of a dual-layer sandwich detector or energy spectrum distributions obtained using two X-ray machines. In addition, μ(E) is a linear attenuation coefficient of an object, and H is a projection matrix. According to the traditional CT reconstruction method, estimated values {circumflex over (μ)}_H and {circumflex over (μ)}_L of effective linear attenuation coefficient distributions μ_H and μ_L at a high energy level and a low energy level can be obtained, {circumflex over (μ)}_H and {circumflex over (μ)}_L can be used as prior information.

At present, in the art, the dual-energy decomposition includes two decomposition methods, i.e., material based decomposition and dual-effect decomposition. In addition, both decompositions can be expressed as equation (3) as follows:μ(E)=a₁φ₁(E)+a₂φ₂(E)  (3)wherein, φ₁(E) and φ₂(E) in the above equation (3) are in different predetermined function forms in the material based decomposition and the dual-effect decomposition, and in addition, a₁ and a₂ are decomposition coefficients.

In addition, two new vectors are introduced in the method according to the present disclosure, i.e., scaled images r₁ and r₂, which are expressed by equation (4) as follows:

$\begin{array}{cc}{r}_1\equiv \mathrm{diag}\left(\frac{1}{{\omega }_1+\varepsilon }\right)a_1{r}_2\equiv \mathrm{diag}\left(\frac{1}{{\omega }_2+\varepsilon }\right)a_2& \left(4\right)\end{array}$

In the above equation (4), diag( ) represents a diagonal matrix, and elements on a diagonal thereof are values of the vector in the parentheses, ε is a vector with small constant coefficients, to avoid occurrence of a zero value in the denominator, and ω₁ and ω₂ are two vectors which can be selected randomly. In addition, H′₁≡Hdiag(w₁+ε) and H′₂≡Hdiag(ω₂+ε) are defined, and in combination with the above equations (1)-(4), equation (5) can be obtained as follows:−ln ∫w_H(E)exp(−H′₁r₁φ₁(E)−H′₂r₂φ₂(E))dE=p_H −ln ∫w_L(E)exp(−H′₁r₁φ₁(E)−H′₂r₂φ₂(E))dE=p_L  (5)In addition, assume thatR₁≡H′₁r₁ R₂≡H′₂r₂  (6)wherein, R₁ and R₂ are projection images of the scaled images r₁ and r₂ respectively, and H′₁ and H′₂ are corresponding projection operators respectively.

Next, equation (5) is simplified using equation (6), to obtain equation (7) as follows:−ln ∫w_H(E)exp(−R₁φ₁(E)−R₂φ₂(E))dE=p_H −ln ∫w_L(E)exp(−R₁φ₁(E)−R₂φ₂(E))dE=p_L  (7)

Here, the above equation (7) can be referred to as data under prior definition of a divisor. In addition, for each pair of collected high-energy data and low-energy data, equation (7) forms a non-linear binary equation set. (R₁, R₂) can be obtained according to (p_H, p_L) by solving this equation set, or (R₁, R₂) can be obtained according to a known data pair (p_H, p_L) by establishing a lookup table using a method similar to that described in the non-patent document 2. The remaining problem is to reconstruct (r₁, r₂) according to (R₁, R₂). It can be known from the above equation (6) that the reconstruction of (r₁, r₂) can be completed using any traditional CT reconstruction method. However, in a case that (r₁, r₂) is reconstructed using a traditional dual-energy CT reconstruction method, since there is strong unbalance in a process of dual-energy decomposition, it results in amplification of a noise of the dual-energy CT reconstructed image, especially amplification of a noise of the equivalent atomic number distribution.

In contrast, the method according to the present disclosure is characterized in that a₁ and ω₁ are enabled to have similarity and a₂ and ω₂ are enabled to have similarity by constraining piece-wise smooth of r₁ and r₂, i.e., mathematical expressions thereof can be sparse, which enables improvement of quality of the reconstructed image during reconstruction using such characteristics. Both a₁ and ω₁ are images, and similarity therebetween refers to similarity of structures between the images. For example, ω₁ is smooth in a place where a₁ is smooth, and ω₁ has an edge in a place where a₁ has an edge. In addition, due to similarity of structures between ω₁ and ω₂ and μ_H or μ_L, ω₁ and ω₂ can be selected as μ_H or μ_L (wherein, μ_H or μ_L can be reconstructed from p_H and p_L respectively according to a traditional single-energy CT) or other prior images similar to ω₁ and ω₂ (for example, in the present disclosure, a₂ can be used in the dual-effect decomposition method). In addition, in a case that both ω₁ and ω₂ are uniform constant vectors, this method is degraded to a normal dual-energy reconstruction method. In addition, in the present disclosure, in order to improve the quality of the reconstructed image, piece-wise smooth constraints on r₁ and r₂ are implemented by using the following conditions, i.e., equation (8) (a second piece-wise smooth constraint condition) and equation (9) (a first piece-wise smooth constraint condition),

$\begin{array}{cc}\underset{r_1}{\min }{\nabla r_1}_{p}s.t.R_1\equiv H_1^{\prime }{r}_1& \left(8\right)\\ \underset{r_2}{\min }{\nabla r_2}_{p}s.t.R_2\equiv H_2^{\prime }{r}_2& \left(9\right)\end{array}$

In the above equations (8) and (9), ∥∇r∥_p represents a p-order norm of a gradient of r. Here, piece-wise smooth of r is implemented by minimization of ∥∇r∥_p.

In addition, in the above equations (8) and (9), except for subscripts, the expressions are totally the same. Therefore, reconstructions of r₁ and r₂ can be implemented independently using the same method. However, on the other hand, in terms of the dual-energy CT, in order to optimize the quality of the reconstructed image, the selection of ω₁ and ω₂ can be optimized respectively from the characteristics of the dual-energy CT according to practical conditions. For example, for a dual-effect decomposition method (with reference to the non-patent document 2), the equivalent atomic number reconstructed image generally has a poor signal-to-noise ratio, and in contrast, the electron density image has a better signal-to-noise ratio. In addition, based on the dual-effect decomposition method, calculation equations of an equivalent atomic number Z^eff and an electron density ρ_e are as shown in equation (10) as follows:

$\begin{array}{cc}{Z}^{\mathrm{eff}}={\left(\mathrm{diag}\left(\frac{1}{a_2}\right)a_1\right)}^{\frac{1}{\lambda -1}},\lambda \approx 4p_e=2a_2& \left(10\right)\end{array}$ wherein, λ is a parameter indicating a change of an photoelectric effect with energy in the dual-effect decomposition, and ω₁=a₂ can be selected. Thus, the following equation (11) is obtained:

$\begin{array}{cc}{Z}^{\mathrm{eff}}\approx r_1^{\frac{1}{\lambda -1}},\lambda \approx 4& \left(11\right)\end{array}$

Thus, a noise level of the equivalent atomic number reconstructed image can be better controlled, since a value of the equivalent atomic number can be enabled to be relatively stable by constraining piece-wise smooth of r₁.

Here, in the present disclosure, specifically by taking the dual-energy CT reconstruction based on dual-effect decomposition as an example, the following specific embodiments are given using the method according to the present disclosure.

FIG. 1 is a flowchart of a self-prior information based dual-energy CT reconstruction method with dual-effect decomposition as an example. FIG. 2 is an image obtained by reconstruction using a self-prior information based dual-energy CT reconstruction method with dual-effect decomposition as an example, wherein, (a) is an electron density reconstructed image, and (b) is an equivalent atomic number reconstructed image.

In the present embodiment, as shown in FIG. 1, first of all, an energy spectrum is rated and a dual-energy lookup table is established. With respect to establishment of a lookup table, determined known materials are made to have different thicknesses, and thereby, (R₁, R₂) are known. Then, these materials are placed into the dual-energy CT system for data collection, so as to obtain (p_H, p_L). A table is generated using these data. In general, more than two materials and dozens of thicknesses are used. Published documents in the art (for example, the non-patent document 2) can be referred to for more detailed description.

Then, a detector of a dual-energy CT imaging system is used to collect high-energy data p_H and low-energy data p_L of the dual-energy CT imaging system.

Next, values of projection images R₁ and R₂ of scaled images r₁ and r₂ are obtained through a lookup table or according to the above binary equation set (7).

Next, μ_H is reconstructed from p_H using a traditional single-energy CT reconstruction method, and ω₂=μ_H is selected. That is, structure information of μ_H is used as prior information.

Next, H′₂ is obtained according to the above equation H′₂≡Hdiag(ω₂+ε). Then, r₂ is reconstructed according to the above equation (9) (the first piece-wise smooth constraint condition), i.e.,

$\underset{r_2}{\min }{\nabla r_2}_{p}s.t.R_2\equiv H_2^{\prime }{r}_2.$

In addition, there are a variety of methods in the art to be selected for implementing the above optimization problem, and the optimization is generally completed by means of iteration. In the present application, the optimization will be described by taking an ART+TV method as an example.

1) Initialize r₂=1.

n=1, and r₂^{n, 1}=r₂;

2) For n=2, . . . , N_iter:

ART iterations: for m=2, . . . , N_ART:

$r_2^{n,m}=r_2^{n,m-1}+H_i·\frac{{\left[R_2^{\prime }\right]}_i-{\left[H\right]}_i·r_2^{n,m-1}}{{{\left[H\right]}_i}_2};$ wherein, i is a ray index number, [H]_i is a vector in an i^th row of the system matrix, N_iter is a total number of iterations.

3) Apply regularization constraints to each element of the vector r₂^{n, NART}: r₂^{n, NART}=max(r₂^{n, NART}, 0), wherein, N_ART is a number of ART iterations.

4) Apply total variation minimization iteration to the vector r₂^{n, NART}:dⁿ=∥r₂^n,1−r₂^n,NART∥₂, r₂^n,1=r₂^n,NART

Total variation steepest descent method: for k=2, . . . , N_TV, α=0.2, ε=10⁻⁸:

$v^{n,k-1}=\frac{\partial {r}_{\mathrm{TV}}}{\partial r}{|}_{r=r^{n,k-1}},v^{n,k-1}=\frac{v^{n,k-1}}{v^{n,k-1}};$ $r_2^{n,k}=r_2^{n,k-1}-\alpha ·d^n·v^{n,k-1};$

5) r₂^{n+1, 1}=r₂^{n, NTV}; return to 2) to start a next iteration, wherein, N_TV is a minimized number of iterations for TV.

In addition, as described above, in the present disclosure, except for the above ART+TV method, r₂ can also be reconstructed using other methods such as a split Bregman method.

Next, a decomposition coefficient a₂ and an electron density ρ_e are obtained according to the above equations (4) and (10), i.e., a₂=r₂×diag(ω₂+ε) and p_e=2a₂ are calculated. Thereby, an electron density image can be obtained. FIG. 2(a) is an electron density reconstructed image which is obtained as described above. As shown in FIG. 2(a), the edge of the electron density image is clear.

Next, ω₁=a₂ is set, and thereby H′₁ is obtained according to the equation H′₁=Hdiag(a₂+ε). That is, structure information of a₂ is used as prior information here.

Next, r₁ is reconstructed according to the above equation (8) (the second piece-wise smooth constraint condition), i.e.,

$\underset{r_1}{\min }{\nabla r_1}_{p}s.t.R_1\equiv H_1^{\prime }{r}_1.$ In addition, steps of a specific implementation for reconstructing r₁ are the same as those for reconstructing r₂ described above.

Next, an equivalent atomic number

$Z^{\mathrm{eff}}\approx r_1^{\frac{1}{\lambda -1}}$ is obtained according to equation (11), and thereby, an equivalent atomic number image can be obtained. FIG. 2(b) is an equivalent atomic number reconstructed image which is obtained as described above. As shown in FIG. 2(b), uniformity of the noise of the equivalent atomic number image in a local region of the same material is largely improved, there is no abnormal point, and slim shaped substances can be reconstructed well.

In addition, reconstruction of r₁ may also be different from the above. ω₁=μ_H is set (i.e., taking structure information of μ_H as prior information) and r₁ is reconstructed using equation (8) as the second piece-wise smooth constraint condition, and then an equivalent atomic number image Z^eff is obtained using a₁=r₁×diag(ω₁+ε) and

$Z^{\mathrm{eff}}={\left(\mathrm{diag}\left(\frac{1}{a_2}\right)a_1\right)}^{\frac{1}{\lambda -1}},\lambda \approx 4.$ Thus, an equivalent atomic number image with a high quality can also be obtained similarly.

As described above, the present disclosure is described by taking the dual-energy CT reconstruction based on dual-effect decomposition as an example, but it is not limited thereto. The method according to the present disclosure can also be applied to material based decomposition.

As described above, in the present disclosure, the noise in the dual-energy reconstructed image can be effectively prohibited while keeping the resolution by effectively using information inherent in data, and the algorithm can conveniently be designed by establishing reconstruction by means of a prior model. In addition, the method according to the present disclosure is not limited to one scanning method, and is also suitable for different scanning methods such as a fan beam, a cone beam, a circular orbit, a spiral orbit etc., and can increase robustness of iterative reconstruction using this prior method. Moreover, compared with a dual-effect decomposition method in the art, the X-ray dual-energy CT reconstruction method according to the present disclosure obtains a more stable result by directly reconstructing a ratio of the coefficients.

Claims

1. An X-ray dual-energy CT reconstruction method, comprising: (a) collecting high-energy data pH and low-energy data pL of a dual-energy CT imaging system using a detector of the dual-energy CT imaging system;(b) obtaining projection images R1 and R2 of scaled images r1 and r2 according to the obtained high-energy data pH and low-energy data pL;(c) reconstructing the scaled image r2 using a first piece-wise smooth constraint condition and obtaining a decomposition coefficient a2; and(d) reconstructing the scaled image r1 using a second piece-wise smooth constraint condition and obtaining a decomposition coefficient a1.

2. The X-ray dual-energy CT reconstruction method according to claim 1, wherein, the scaled images r1 and r2 are defined in equation (4) as follows:

3. The X-ray dual-energy CT reconstruction method according to claim 2, wherein, in the step (c), an effective linear attenuation coefficient μH at a high energy level is reconstructed according to the high-energy data pH, ω2=μH is selected, and r2 is reconstructed using equation (9) as the first piece-wise smooth constraint condition.

4. The X-ray dual-energy CT reconstruction method according to claim 2, wherein, in the above step (d), ω1=a2 is set and r1 is reconstructed using equation (8) as the second piece-wise smooth constraint condition.

5. The X-ray dual-energy CT reconstruction method according to claim 4, further comprising: obtaining an electron density image ρe according to a2=r2×diag(ω2+ε) and ρe=2a2 using dual-effect decomposition.

6. The X-ray dual-energy CT reconstruction method according to claim 4, further comprising: obtaining an equivalent atomic number image Zeff according to

7. The X-ray dual-energy CT reconstruction method according to claim 2, further comprising: obtaining an electronic density image ρe according to a2=r2×diag(ω2+ε) and ρe=2a2 using dual-effect decomposition.

8. The X-ray dual-energy CT reconstruction method according to claim 2, further comprising: obtaining an equivalent atomic number image Zeff according to

9. The X-ray dual-energy CT reconstruction method according to claim 2, wherein, in the step (d), an effective linear attenuation coefficient μH at a high energy level is reconstructed according to the high-energy data pH, ω1=μH is set, and r1 is reconstructed using the following equation (8) as the second piece-wise smooth constraint condition:

10. The X-ray dual-energy CT reconstruction method according to claim 2, wherein, in the steps (c) and (d), r1 and r2 are reconstructed using an ART+TV method.

11. The X-ray dual-energy CT reconstruction method according to claim 2, wherein, in the steps (c) and (d), r1 and r2 are reconstructed using a split Bregman method.