Application of dual-energy Dr material Recognition Algorithm in CT Imaging System

Author: Tang Xiaoye, Zhang Li, Chen Zhiqiang, Gao hewei, and Zhang Guowei

【Abstract]This article introducesComputingThe dual-energy fluoroscopy (DR) Material Recognition Algorithm for tomography (CT). This algorithm is based on the single-energy CT reconstruction image, through Segmentation of CT images and extraction of geometric information of the fault, effective atomic ordinal number andElectronicsDensity Distribution. The CT imaging system combined with various scanning tracks can realize effective material identification. At the same time, compared with the traditional dual-energy CTMethodCombined with single-energy CT reconstruction images, this method improves the dual-energy Dr recognition effect, and can achieve more accurate material identification and perform security checks.ApplicationThe domain has practical significance.

【Key words]Dual-energy Dr; CT imaging; material Recognition

Abstract: A dualenergy digital radiography (DR) method specified ed by segmentation of single energy reconstruction image is proposed for material recognition in Xray CT inspection systems.TheValid atomic number and equivalent electron density distribution of the scanned objects can be reconstructed with this method. compared to the conventional Dual Energy computed tomography (CT) technique, this method markedly reduces CES The amount of dualenergy detectors in the system of CT scanning,WhichShows its significance in application in the field of security inspection.

Key words: dualenergy Dr; CT image system; material Recognition

Introduction

Since the incident, public attention to terrorist incidents has been increasing worldwide. For the sake of Public Security, the inspection of explosives and other dangerous goods or contraband in passenger luggage is becoming more and more important.CurrentlyThe security inspection system used at the entrance of the airport and railway station mainly uses dual-energy X-ray to distinguish and check the substances in the passenger's luggage [1].

In 1976, Alvarez and macovski [2] first proposed a basic reconstruction method for dual-energy X-ray tomography to reconstruct the distribution of the atomic number and density of matter by solving the nonlinear equation. A new dual-energy curve method [3, 4] has been proposed in recent years. By fitting the curve of the high and low-energy transparency obtained by the detector, a precise reconstruction of the effective atomic sequence number and mass thickness of the unknown uniform material is realized.

The computer tomography technique solves the overlapping of objects in general fluoroscopy imaging.ProblemIf the projection data is complete, a precise fault image can be obtained. So far, the CT imaging system hasDevelopmentTo the fifth generation. Hounsfield, a British engineer, designed the first-generation CTF in 1967 using parallel bundle translation-rotation. The second generation CTM adopts the fan bundle translation-rotation mode. Third-generation CTDs are commonly used slice bundle detectors and source rotation scanning methods. The scanning method of the fourth generation CTR is similar to that of the third generation, but only the X-ray source rotation is performed during the scanning process. The fifth generation of CT is the electron beam scanner, which rotates the X-ray source through the rotation of the electron beam, so as to achieve fast scanning within milliseconds [5].

Combined with the features of the above technology, a dual-energy technology-based CT imaging system was developed for inspection of luggage items such as airports, customs, and stations, which not only produced tomography images, but also recognized stacked items, it can also obtain the material information of the items to be inspected to achieve more accurate material identification.

In this paper, a new method based on CT reconstruction image is proposed, which can accurately reconstruct the effective atomic number and electron density distribution of materials. Compared with the conventional dual-energy DR algorithm, this method can solve the problem of overlapping materials. Compared with the conventional dual-energy CT algorithm based on solving nonlinear equations, This method uses image segmentation technology to reconstruct blocks and solve linear equations. Combined with the CT imaging system, a small number of dual-energy detectors can be used to achieve accurate and rapid material identification, which is suitable for high customs clearance requirements of the security inspection system.

1 dual-energy tomography Reconstruction

The traditional dual-energy CT algorithm uses dual-energy projection data to reconstruct the material characteristics of objects to be tested. By decomposing the linear absorption coefficient, it can be expressed as a linear combination of the amount related to the material characteristics. The basic decomposition functions can be divided into two types: linear Combination of Attenuation Caused by Compton scattering and Photoelectric Absorption: μ (e) = C1 μ P (e) + C2 μ c (E), (1) μ P (e) is the linear Attenuation Caused by the photoelectric effect, while μ c (e) is the linear Attenuation Caused by the Compton scattering. The constants c1 and c2 are the unknown proportional coefficients of the tested materials.

Linear Combination of attenuation coefficients of the two selected base materials: μ (e) = A1 μ A (e) + A2 μ B (E), (2) where μ A (e) and μ B (e) are the linear attenuation coefficients of the two selected base materials.

According to the beer law, for an actual X-ray imaging system, when X-ray attenuation occurs through a uniform substance, the high and low-energy transparency outputs of the detector can be expressed as: Th = ∫ ehn (E) EXP-∫ μ (e) dlepd (e) de, (3A)

TL = ∫ ELN (e) exp-∫ μ (e) dlepd (e) de, (3B) Where n (e) is an X-ray energy spectrum, Pd (E) is the energy response function of the detector, and μ is the linear attenuation coefficient.

Therefore, material properties can be reconstructed through known dual-energy projection data and basic decomposition of the linear attenuation coefficient. The postreconstruction method and the prereconstruction Method Based on the reconstructed image and the projection data sine map are two classic double energy CT reconstruction algorithms.

2 dual-energy Dr material Recognition Algorithm

In the traditional dual-energy CT reconstruction algorithm, the effective atomic ordinal numbers and density of matter need to be reconstructed individually in pixels. To obtain the distribution information of density P and atomic sequence Z, we must obtain CT projection data of high energy and low energy respectively, which requires a large number of dual-energy detectors during scanning. However, in practical applications, we often do not need to reconstruct the P and Z values one by one pixel. For example, for objects with simple fault ry and uniform physical characteristics, the product only needs to be reconstructed one by one to meet the material identification requirements due to the approximation of the P and Z values of the uniform parts. Therefore, a dual-energy Dr Method Assisted by Single-energy CT reconstruction image is proposed. The geometric structure of the extracted body fault is reconstructed by dividing single-energy CT and the P and Z values are reconstructed one by one. In line track scanning, dual-energy detectors can be used in a certain direction to complete the scanning process of single-energy CT and dual-energy Dr at the same time.

For a fixed dual-energy detector, the relative linear motion of the object to be tested and the system can be equivalent to Dr scanning of the object, as shown in 1. Assuming that the segmentation of the single-energy CT reconstruction image splits the fault into N relatively even blocks for reconstruction, and the dual-energy detector obtains M sets of Dr projection data. Then we use th (I) and TL (I) to represent the high and low-energy projection data in group I, LJ (I) it indicates the length of the shot bundle corresponding to the projection data in group I after the J Block.

Through the decomposition of linear attenuation coefficient: μ (e) = p naa (A σ EL + A σ cohs+ A σ incohs)

= P naa (A σ EL + A σ SC), (4) among them, photoelectric effect reaction section A σ El = 42zn α 4m0c2e7/20, scattering reaction section

A σ SC ≈ zkn (e ),

In formula, α = 1/137 is the fine structure constant, M0 is the electronic mass, 0 = 83 π (E2/MC2) 2, Kn (e) represents the kleinnishina section of the computation scattering.

Thus, we can define the feature coefficients a1, a2: A1 = p zna, (5A) of matter)

A2 = P za, (5B) where the index n is determined by the relationship between the photoelectric effect and the atomic number of the substance. The value range is 4 ~ In this article, n = 4.5. Then the attenuation coefficient can be simplified to μ (e) = p za (Zn-1F (E, Z) + g (E, z), where F (E, Z ), G (E, z) is weakly related to the atomic number z of matter.

Therefore, we can express the high and low-energy transparency as: Th = ∫ ehn (e) exp-∫ μ (e) dlepd (e) de

= Effecehn (e) exp-A11E3-A2fKN (E)

· EPD (e) de, (6a)

TL = canceln (e) exp-∫ μ (e) dlepd (e) de

= Canceln (e) exp-A11E3-A2fKN (E)

· EPD (e) de, (6B) A1 and A2 are the line points of coefficients A1 and A2 along the ray direction, that is

A1 = a1dl, a2 = a2dl.

For each set of high and low transparency (Th, Tl), the atomic ordinal number z and mass thickness TM of the Equivalent Uniform substance can be uniquely determined by the dual-energy curve fitting method. Therefore, the corresponding coefficient (a1, a2) can be obtained by calculating the formula a1 = tmzn/A, (7A)

A2 = TMZ/. (7B) because the coefficients A1 and A2 in each block are the same, the line integral can be expressed as the weighted sum of the coefficients of each block. A1 (I) = 1_a1dl = nj = 1A1 (I, j) = ka1klk (I), (8A)

A2 (I) = 109a2dl = nj = 1A2 (I, j) = ka2klk (I ). (8B) from this we can obtain the coefficients a1 = {A1, 1, A1, 2 ,..., A1, n} and A2 = {A2, 1, A2, 2 ,..., Linear Equations of A2, n}: L1 (1) L2 (1 )... Ln (1)

L1 (2 ).........

L1 (m )...... Ln (m) A1, 1

A1, 2

A1, n

= A1 (1)

A1 (2)

A1 (m), (9A)

L1 (1) L2 (1 )... Ln (1)

L1 (2 ).........

L1 (m )...... Ln (m) A2, 1

A2, 2

A2, n

= A2 (1)

A2 (2)

A2 (M). (9B) according to the relationship between the block coefficients (A1, K, A2, K) and the atomic number and density of the corresponding substances, A1, K = P kznkak,

A2, K = P kzkak, (10) can be obtained, A1, ka2, K = Zn-1k, k, 1, 2 ,..., N. (11) So by solving equations (9A) and (9B), we can reconstruct the effective atomic number and electronic density distribution of matter.

To ensure that the number of fault blocks does not exceed the number of dual-energy data groups obtained, that is, m> N, we use the optimization method.MethodReduce the errorImpactMake Si = nj = 1lj (I), (12) and select (N-1) Non-Correlation Equation to make Si as big as possible. If there are two equations (numbers P, Q, P, Q ,..., M) Make sp = SQ, and A1 (P)

A * 1, 2 = F2 (A1, 1 ),

A * 1, n = FN (A1, 1). (14) remaining (M-N + 1) equations:

L1 (n) L2 (n )... Ln (N)

L1 (n + 1 ).........

L1 (m )...... Ln (m) A1, 1

A1, 2

A1, n

= A1 (N)

A1 (n + 1)

A1 (m), (15)

It can also be expressed,

L1 (n) A1, 1 + L2 (n) A1, 2 +...

+ Ln (n) A1, N-A1 (n) = 0,

L1 (n + 1) A1, 1 + L2 (n + 1) A1, 2 +...

+ Ln (n + 1) A1, N-A1 (n + 1) = 0,

L1 (m) A1, 1 + L2 (m) A1, 2 +...

+ Ln (m) A1, N-A1 (m) = 0. (16)

Due to the existence of the error, we define the function Fe to depict a *, ,..., The difference caused by a * 1, N. Therefore,

Fe (n) = L1 (n) A * + L2 (n) A * +...

+ Ln (n) A * 1, N-A1 (N ),

Fe (n + 1) = L1 (n + 1) A * + L2 (n + 1) A * +...

+ Ln (n + 1) A * 1, N-A1 (n + 1), (17)

Fe (m) = L1 (m) A * + L2 (m) A * +...

+ Ln (m) A * 1, N-A1 (m ).

In combination with equation (15,

Fe (n) = L1 (n) A1, 1 + L2 (n) F2 (A1, 1) +...

+ Ln (n) FN (A1, 1)-A1 (N ),

Fe (n + 1) = L1 (n + 1) A1, 1 + L2 (n + 1) F2 (A1, 1)

+... + Ln (n + 1) FN (A1, 1)-A1 (n + 1 ),

Fe (m) = L1 (m) A1, 1 + L2 (m) F2 (A1, 1) +...

+ Ln (m) FN (A1, 1)-A1 (m), (18)

That is

Fe (n) = Gn (A1, 1 ),

Fe (n + 1) = GN + 1 (A1, 1 ),

Fe (m) = GM (A1, 1), (19)

We obtain Fe (A1, 1) = Fe (n) + Fe (n + 1)

+... + Fe (m)

= Gn (A1, 1) + GN + 1 (A1, 1)

+... + GM (A1, 1 ). (20) The objective is to obtain A1, 1, that is, A1, 1 = Arg min (Fe (A1, 1) with the smallest Fe (A1, 1 )). (21) in combination with equations (15), we obtain a *, ,..., A * 1, n and the optimal solution of equations (9) A * 1 = {A *, ,..., A * 1, n }.

To sum up, the dual-energy Dr reconstruction material recognition method can beSummaryAs follows:

The reconstruction images of Single-energy CT are separated by absorption coefficient, and the geometric amount is extracted to construct the coefficient matrix of linear equations (9A) and (9b;

Using the dual-energy curve method or polynomial fitting method, a set of dual-energy Dr projection data (Th, Tl) is used to obtain the atomic ordinal number and mass thickness (z, Tm) of the Equivalent Uniform Substance );

Use the formula (7A) and (7b)ComputingCorresponding coefficients A1 and A2 are used to establish linear equations (9A) and (9b );

Use the optimization method to solve the linear equations (9A) and (9B), and calculate the block coefficients A1 and A2. Calculate the effective atomic ordinal number Zeff andElectronicsDensity P z/.

3. Numerical Simulation Verification

A water bath model consisting of 5 kinds of uniform materials is projected forward by using X-ray Spectroscopy of 140/160/kV. The dual-energy Dr projection data is obtained. Table 1 provides parameters such as the valid atomic sequence number and density of various materials. Fig 2 water bath model (A) and used X-ray spectroscopy (B) Table 1 model material and parameter simulation calculations, the single-energy CT reconstruction image is produced by a pair of Gaussian noise (mean 0, the variance 0.01) model for parallel FBP reconstruction image simulation. The size of the reconstructed image is 256 × 256, and the number of dual-energy projection data sets is 256. The reconstruction result is shown in table 2. Table 2 simulation results and errors

The simulation results show that the relative error of the valid atomic number of each material is less than 5%, and the relative error of the electronic density is about 1%. From the reconstruction results, we can see that the dual-energy Dr material recognition algorithm has a high accuracy for the identification of simple and uniform materials. However, because the reconstruction process relies on the segmentation of the single-energy CT reconstruction image, the reconstruction quality of the single-energy CT image and the geometric complexity of the model fault are important factors that affect the accuracy of the algorithm.

4 Conclusion

This paper introduces a material identification method using dual-energy X-ray DR, which is used to reconstruct the effective atomic number and electronic density distribution of the scanned object through the aid of a Single-energy CT image. Combined with a line track CT Imaging System with quasi-parallel bundle properties, on the one hand, we can obtain the tomography to solve the overlap of objects.ProblemAnd more accurate material identification. Therefore, it can meet the needs of the security inspection system for fast customs clearance, the ability to strip overlapping objects and identify materials. Because the reconstruction of Single-energy CT images has a great impact on the reconstruction of P and Z valuesResearchThe direction of CT will focus on improving the segmentation effect of CT reconstruction images of linear trajectory.

 

【ReferenceDocument]
[1] Wang Qi, Chen Zhiqiang, Yan Xiaoping, et al.. X-ray security inspection technical review [J]. CTTheoryAndApplicationStudy, (1): 32-37

[2] Robert E. Alvarez and Albert macovski, energyselective reconstructions in Xray computerized tomography [J]. Phys. Med. Biol., 1976, 21 (5): 733-744.

[3] Guowei Zhang, Li Zhang, Zhiqiang Chen, an HL Curve Method for material discrimination of Dual Energy Xray inspection systems [A]. conf. record of 2005 IEEE nuclear science symposium [C]. 2005,326-328.

[4] Guowei Zhang, Zhiqiang Chen, Li Zhang, and Jianping Cheng, exact reconstruction for Dual Energy computed tomography using an HL curve method [A]. conf. record of 2006 IEEE medical imaging Conference [C]. 2006, m14462.

[5] Zhuang tiange, CT principle and algorithm [M]. ShanghaiTransportationUniversity Press, 1992.

[6] sidky e y, Zou y, and Pan X. Volume Image Reconstruction from a straightline source trajectory, IEEE Symposium on nssmic [c], 2005.

[7] Deborah J. walter, Eric J. tkaczyk, Xiaoye Wu. accuracy and precision of Dual Energy CT imaging for the quantification of tissue fat content [J]. proceedings of Spie [C]. 2006,614 2.

[8] ohnoa y, torikoshia M, tsunooa T, et al. dualenergy Xray CT with cadmium array and its extension [J]. nuclear Instruments and Methods in physics research, 2005, a (548): 72-77.