A substance identification method based on adaptive quantitative decomposition

Abstract

The application discloses a material identification method based on adaptive quantitative decomposition, and establishes a polynomial relationship between coefficients participating in material identification and energy and effective atomic number in the interaction of rays and materials. Then, a plurality of known materials are used as calibration materials to perform scanning and reconstruction. Based on the linear attenuation coefficients of different energy regions of the calibration materials, the real effective atomic number and density of the calibration materials, and the polynomial relationship obtained previously, the effective energy of the corresponding energy region is determined. In actual operation, after scanning and reconstruction of the material to be decomposed, the effective energy is determined through calibration, the linear attenuation coefficients of different energy regions are only related to the effective atomic number and density, and through simultaneous solution of two ray and material interaction formulas containing variable parameters, the effective atomic number and density of the scanned material can be accurately obtained. The application can automatically adjust the related coefficients according to different materials, thereby improving the decomposition precision.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of material detection, and relates to a substance identification method based on adaptive quantitative decomposition, which realizes fast and accurate estimation of the effective atomic number and density of the scanned substance on the reconstructed image by adaptive quantitative analysis and decomposition of spectral CT on the image domain. BACKGROUND

[0002] Spectral CT (Spectral Computed Tomography) is an advanced technology based on X-ray imaging. By setting the energy threshold of the detector, it utilizes the attenuation characteristics of different energies of X-rays by materials to generate projection images of multiple energy regions and reconstruct them. In traditional CT imaging, the attenuation information of X-rays is recorded as a whole, and it is impossible to distinguish the attenuation of different energies. In spectral CT, the detector can segment the X-rays passing through the object by energy, thereby obtaining attenuation data of different energy regions. These data represent the response of different materials in the object to different energy X-rays, and can reconstruct CT images of multiple energy regions. This technology enables spectral CT to provide more abundant image information than traditional CT, thereby more accurately distinguishing and quantifying different materials in the object.

[0003] In order to accurately extract the quantitative information of the scanned substance, especially its inherent physical properties, the material decomposition algorithm plays an important role in spectral CT. The attenuation of a material to X-rays is mainly determined by its effective atomic number and density, and the difference in attenuation coefficients of different materials is due to these two key factors. Therefore, in non-destructive testing, the core of substance identification lies in accurately solving its effective atomic number and density. The effective atomic number refers to the attenuation characteristics of a mixed substance at the same energy as those of a pure substance with a certain atomic number when the density is determined. Therefore, the effective atomic number of a mixed substance can be represented as the value of this specific atomic number. By accurately solving the effective atomic number and density of the substance, different substances can be better identified in non-destructive testing applications, thereby improving the accuracy and effectiveness of the detection.

[0004] In current technology, there are mainly two models for solving the effective atomic number and density of a substance. The first method is based on the basis material decomposition model, which represents the attenuation coefficient of the scanned substance as a linear combination of the attenuation characteristics of several known basis materials. By calculating the combination coefficients of these basis materials, the effective atomic number and density of the substance can be indirectly estimated. The second method is a decomposition method based on the dual-effect model, which directly represents the attenuation characteristics of the substance as a combination of photoelectric effect and Compton effect, which are closely related to the effective atomic number and density of the substance, and directly solves the effective atomic number and density.

[0005] These traditional decomposition algorithms perform well in the application of dual-energy CT, providing reliable material composition information for clinical and industrial applications. However, most of these algorithms assume that the effective atomic number and density of the sample are within a certain fixed range, and are used under relatively stable scanning conditions, thus simplifying the physical model. With the development of photon counting detector (PCD) technology, the energy threshold of the detector can be flexibly adjusted, thus selecting different energy regions for imaging. This not only effectively reduces image noise, but also improves imaging quality. Compared with traditional dual-energy CT, CT based on photon counting detector (PCDCT) has greater application potential, especially when faced with complex samples. However, the traditional fixed physical model may be inaccurate under such flexible and variable scanning conditions and in complex samples, resulting in a decrease in the accuracy of the decomposition results. SUMMARY

[0006] In view of the problems in the prior art, the purpose of the present application is to provide a material identification method based on adaptive quantitative decomposition, which can estimate the effective atomic number and density of the scanned material more accurately and conveniently with less experimental operation in actual application. The present application is designed for the diversified needs of complex samples and can dynamically optimize the physical model of the interaction between radiation and matter under different scanning conditions. Compared with the traditional fixed model, the present application can automatically adjust the key parameters in the model according to the different characteristics of the scanned sample, thus adapting to different material types and scanning conditions; this adaptive mechanism greatly improves the accuracy of material decomposition and can function in various complex imaging scenarios.

[0007] The method relates to an image field decomposition technology, and is based on a dual-energy region reconstructed image of a scanned substance to obtain effective atomic number and density through one-time calibration. First, the method establishes a polynomial relationship between a coefficient participating in material identification in the interaction of rays and a substance, energy and effective atomic number. Next, two or more known materials are used as calibration materials to perform scanning and reconstruction, and effective energy of a corresponding energy region is determined based on linear attenuation coefficients (mu) of the calibration materials in different energy regions, real effective atomic number and density of the calibration materials, and the previously obtained polynomial relationship. In actual operation, after scanning and reconstruction of a to-be-decomposed substance, since the effective energy has been determined through calibration, mu values of different energy regions are only related to effective atomic number and density in the interaction model of rays and the substance. By simultaneously solving two formulas of the interaction of rays and the substance containing variable parameters, formula (12) is obtained, and an optimization algorithm such as the Newton method or the gradient descent method is used to solve formula (12), so that effective atomic number and density of the scanned substance can be accurately obtained. In formula (12), except for atomic number Z and density to be solved, only energy is an unknown quantity, and therefore the value of energy needs to be obtained before calculation, and average energy is used to represent E of the two energy regions. Therefore, the method uses the calibration materials with known atomic number and density to solve E, and through formula (11), after mu of the calibration materials is obtained through scanning, effective atomic number and density are known, and then E can be solved.

[0008] The method no longer depends on a simple double-effect model to solve effective atomic number and density in constructing the interaction model of rays and a substance. By introducing consideration of coherent scattering, the interaction model of rays and the substance is more perfect and comprehensive, so that the real physical characteristics of a complex material are better reflected. The key parameters in the interaction model are related to effective atomic number and energy of a scanned substance through polynomial fitting, and can be seen from formula (7)-(11).

[0009] The method can dynamically adjust coefficients of photoelectric effect and Compton scattering. In the photoelectric effect part, the method no longer uses a fixed exponential coefficient, but introduces a dynamic function related to energy and effective atomic number of a substance. Similarly, in the Compton scattering part, the method also introduces the function dependent on energy and material characteristics. This method allows the model to automatically adjust the related coefficients during decomposition according to different substances, so as to improve the decomposition accuracy and avoid errors caused by using fixed coefficients.

[0010] The technical scheme of the method is as follows:

[0011] A material identification method based on adaptive quantitative decomposition, and steps of the method include:

[0012] 1) constructing an interaction model of rays and a substance mu = pm (σ′ De_Compt +σ Compt ); where ρ m is the mass density, σ′ De_Compt is the de-Compton model including coherent scattering, σ Comp t is the Compton scattering cross section between the scanning ray and the material; C(E,Z eff )=Σ ij h ij Z eff i E j ,0≤i≤4,0≤j≤4; α1 is a constant, E is the energy of the scanning ray, Z eff is the effective atomic number of the scanned material, i is C(E,Z eff )Z eff The index of E, j is the index of E, h is the index of ij is the corresponding term Z in the polynomial relation eff i E j The coefficient of σ Compt =α2D(E,Z eff )f KN (E), D(E, Z eff )=Σ st g st Z eff s E t ,0≤s≤4,0≤t≤4, where s is D(E,Z eff )Z eff The index of D(E,Z eff ) in which E is the exponent, g st is the corresponding term Z in the polynomial relation eff s E t coefficient, α2 is a constant, f KN (E) is the Klein-Ceshima coefficient;

[0013] 2) Using the relevant data of the selected calibration object, the variable parameters C(E, Z eff )、D(E,Z eff ) to fit and establish variable parameters C(E,Z eff ) and energy E, effective atomic number Z eff The specific relationship between them, and the variable parameters D(E,Z eff ) and energy E, effective atomic number Z eff The specific relationship between them;

[0014] 3) Using C(E,Zeff ), D(E, Z eff ) and combining the known effective atomic number and density information of various materials in the selected calibration object, the effective energy of each energy region is solved to obtain the effective energy E L of each material in the low energy region and the effective energy E H in the high energy region.

[0015] 4) scanning the object to be detected according to the scanning parameters of the calibration object to obtain the energy spectrum CT of the object to be detected; then according to the energy spectrum CT of the object to be detected, combining the specific relationship of C(E, Z eff ), D(E, Z eff ) and the effective energy E L of each known material in the low energy region and the effective energy E H in the high energy region obtained in step 3), the linear attenuation coefficient of the object to be detected in the low energy region and the linear attenuation coefficient in the high energy region are obtained and substituted into the ray and material interaction model to solve the effective atomic number and the material density of the object to be detected.

[0016] Further, the method for obtaining the effective energy E L of each material in the low energy region and the effective energy E H in the high energy region is as follows: first, scanning and reconstructing the selected calibration object to obtain the linear attenuation coefficient μ h of each material in the calibration object in the high energy region and the linear attenuation coefficient μ l in the low energy region; using the known physical properties of the calibration materials, then solving to obtain the effective energy E L of the corresponding material in the low energy region and the effective energy E H in the high energy region.

[0017] Further, in step 4), the method for solving the effective atomic number and the material density of the object to be detected is as follows: scanning and reconstructing the object to be detected using the energy spectrum CT to obtain the linear attenuation coefficient μ l ′ of the object to be detected in the low energy region and the linear attenuation coefficient μ ′ h in the high energy region; using the effective energy estimation of the two energy regions to estimate the effective energy E L in the low energy region and the effective energy E H in the high energy region, then using the Newton gradient method to solve the formula pixel by pixel to obtain the effective atomic number Z eff and the material density p m of the object to be detected.

[0018] The main steps of the present application include:

[0019] 1) Constructing the model of interaction between rays and matter, and fitting the variable parameters in the model of interaction between rays and matter using the attenuation coefficient data of the matter: using the known data of the linear attenuation coefficient, effective atomic number and density of the material, fitting the variable coefficients in the model of interaction between rays and matter. Through these data, a polynomial relationship between the variable coefficients and energy and effective atomic number is established. On this basis, the unknown parameters in the model of interaction between rays and matter are only energy, effective atomic number and density.

[0020] 2) Estimating the effective energy of the dual-energy region: using two or more kinds of known materials as the calibration material to estimate the effective energy of the dual-energy region. The materials of the calibration material do not have to be the same as the decomposed matter, and do not need to meet the requirement that the linear attenuation coefficient of the calibration material in the base material decomposition should contain the scanned matter. However, during the scanning process, the conditions such as voltage, current, amplification ratio should be consistent with the scanning conditions of the decomposed matter. First, using the polynomial relationship established in step (1) and combining the known effective atomic number and density information of various materials in the calibration material, the effective energy of each energy region is solved. Specifically, for a single energy region, the attenuation data of these known materials are fitted by the least squares method to obtain the effective energy of the energy region. The same operation is also applicable to the other energy region, and finally the effective energy of the two energy regions under the current scanning conditions is obtained, at this time the unknown coefficients in the model of interaction between rays and matter are only effective atomic number and density.

[0021] 3) Image domain adaptive decomposition of the scanned matter by spectral CT: CT reconstruction is performed on the scanned matter to obtain the reconstructed images under the two energy regions. Combining the polynomial relationship of the variable parameters fitted in step (1) and the effective energy of different energy regions estimated in step (2), the linear attenuation coefficient data of the scanned matter under the two energy regions is substituted into the model of interaction between rays and matter. By solving the equations, the remaining unknown variables-effective atomic number and density can be solved by using optimization algorithms such as gradient descent.

[0022] The model of interaction between rays and matter is represented as: μ = ρ m (σ ph + σ Compt + σ coherent ), where μ represents the linear attenuation coefficient, ρ m represents the mass density, σ ph is the photoelectric effect cross section between the scanning rays and the matter, σ Compt is the Compton scattering cross section between the scanning rays and the matter, and σ coherentis the coherent scattering cross section between scanning ray and matter. In the model, the Compton effect is considered, and the Compton effect is included in the model. Since the model of scattering is more complex, and in order to better analyze and calculate the variable exponent m, we will include the Compton effect σ ′ De_Compt is still expressed in the form of photoelectric effect, and the exponent m is expressed as a variable C(E, Z eff ) related to effective atomic number and energy:

[0023]

[0024] where α1 is a constant, E represents the scanning energy, and Z eff is the effective atomic number of the scanned matter. In order to make the model more convenient for solving the effective atomic number and density, we will expand C(E, Z eff ) in the following polynomial form: C(E, Z eff ) = Σ ij h ij Z eff i E j , 0≤i≤4, 0≤j≤4, where i is the exponent of Z eff in C(E, Z eff ), j is the exponent of E, and h ij is the coefficient of the corresponding term Z eff i E j in the polynomial relationship. In the description of the Compton effect, the Compton effect is proportional to the atomic number of the scanned matter, and the Klein-Nishina coefficient f KN (E) is used to represent the influence of the Compton effect. However, through in-depth study of NIST data, we found that this approximation Compton model based on mass density is not accurate enough. In order to better describe the Compton effect, we introduce an additional variable D(E, Z eff ) related to E and Z eff to enrich the model: σ Compt = α2D(E, Z eff )f KN (E), where α2 is a constant. And D(E, Z eff ) is expressed in the same form of polynomial series expansion fitting representation: D(E, Z eff ) = Σ st g st Z eff s E t , 0≤s≤4, 0≤t≤4, where s is the exponent of Z in D(E, Z eff ), t is the exponent of E, and gst is the coefficient of the corresponding term in this polynomial relationship. Thus the linear attenuation coefficient of a material can be expressed as:

[0025] Two or more known materials are selected as the calibration materials, and the calibration materials are scanned. During the scanning, the voltage, current, amplification ratio and other conditions should be consistent with the scanning conditions of the decomposed material (i.e. the detected object with unknown components). The calibration materials are reconstructed to obtain the linear attenuation coefficient values of the calibration materials in the two energy regions. The energy estimation is performed in the two energy regions by using the interaction model of the rays and the materials and the fitting formula of the variable coefficients obtained through the NIST data, and the effective energy E L of the low energy region and the effective energy E H of the high energy region are obtained by using the least square method.

[0026] The scanned detected object is reconstructed to obtain the linear attenuation coefficients of the detected object in the low and high energy regions, which are denoted as μ l ′ and μ ′ h The joint solving equation is established, and the formula obtained in the above steps is brought into the equation where μ l ' and μ h ' represent the linear attenuation coefficients of the detected object in the low and high energy regions, respectively, and Z eff is the effective atomic number of the scanned material to be solved. The Newton gradient method is used to solve the effective atomic number Z eff and the material density ρ pixel by pixel in the reconstructed image, and finally the effective atomic number Z eff and the density are solved.

[0027] The advantages of the present application are as follows:

[0028] The present method is based on the image domain decomposition method, and can more accurately and conveniently estimate the effective atomic number and the density of the scanned material in practical applications with less experimental operation. The present application avoids the inaccurate decomposition results caused by the constantization of the coefficients in the double effect, and effectively realizes the self-adaptation of the coefficients in solving the effective atomic number and the density by using the NIST real data, establishing the effective polynomial estimation and bringing it into the double effect decomposition, thereby realizing the accurate identification ability of the material by the spectral CT. The present application has a relatively simple physical process and a relatively convenient experimental operation, and has a relatively broad application prospect in the practical applications such as nondestructive testing. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1Based on NIST data, C(E,Z eff ) and D(E,Z eff ) with respect to the fitted surface of energy and effective atomic number;

[0030] (a)C(E,Z eff ) versus energy and effective atomic number fitting surface, (b) D(E,Z eff ) versus energy and effective atomic number.

[0031] Figure 2 It is a flow chart of the present invention.

[0032] Figure 3 To scan material images and reconstruct tomograms, the material components are marked in the images;

[0033] (a) is a picture display of the scanned object, and (b) is a tomogram reconstructed in the low-energy area.

[0034] Figure 4 This is a comparison chart of the decomposition result images;

[0035] (a) is a comparison chart of 40~75keV, (b) is a comparison chart of 70~150keV, (a) is a comparison chart of effective atomic number, and (b) is a comparison chart of mass density. DETAILED DESCRIPTION

[0036] The present invention will be described in further detail below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0037] The adaptive quantitative decomposition method of the present invention comprises the following steps:

[0038] Step (1) The exponential variable parameter C(E, Z) in the photoelectric effect of the interaction model between radiation and matter eff ) and D(E,Z introduced in Compton scattering eff ) Use the attenuation coefficient data of the material to fit: For example, use the known data such as the linear attenuation coefficient, effective atomic number and density of the material at different energies to fit the variable coefficient in the interaction model between the ray and the material. Through these data, the variable coefficients C(E, Z eff )、D(E,Z eff ) and the specific relationship between energy and effective atomic number, that is, fitting to obtain the coefficients h ij 、g st On this basis, the only unknown parameters in the interaction model between radiation and matter are energy, effective atomic number and density;

[0039] Step (2) estimates the effective energy of the dual-energy region: two or more known materials are used as the calibration material to estimate the effective energy of the dual-energy region. The material of the calibration material does not have to be the same as the decomposed material, but during the scanning process, the conditions such as voltage, current, amplification ratio, etc. should be consistent with the scanning conditions of the decomposed material. First, the polynomial relationship established in step (1) is used, and the known effective atomic number and density information of each material in the calibration material are combined to solve the effective energy of each energy region. Specifically, for a single energy region, the attenuation data of these known materials are fitted by the least squares method to obtain the effective energy of the energy region. The same operation is also applicable to the other energy region, and finally the effective energy of the two energy regions under the current scanning conditions is obtained, and the unknown coefficients in the interaction model between the rays and the material are only the effective atomic number and the density;

[0040] Step (3) performs image domain dual-effect decomposition on the scanned material of spectral CT: CT reconstruction is performed on the scanned material to obtain the reconstructed images under the two energy regions. Combined with the polynomial relationship of the variable parameters fitted in step (1) and the effective energy of the different energy regions estimated in step (2), the linear attenuation coefficient data of the scanned material under the two energy regions are substituted into the interaction model between the rays and the material. By solving the equations, the remaining unknown variables, the effective atomic number and the density, can be solved by using optimization algorithms such as gradient descent.

[0041] Further, step (1) includes: based on the photoelectric effect and Compton scattering between spectral X-rays and matter, Alvarez proposed an approximate fitting representation for wider practical applications, and the linear attenuation coefficient of the material at energy E can be approximated as:

[0042]

[0043] where ρ m represents the mass density of the material, ρ e represents the electron density of the material The index m in the photoelectric effect is m=m'-1, which is usually set as a constant of 3-4. σ ph ,σ Compt respectively represent the photoelectric effect cross section and the Compton scattering cross section, A is the atomic mass, Z is the atomic number, and k1,k2 are constants. f KN (E) is the Klein-Nishina coefficient, and the specific form is:

[0044]

[0045] Heismann further proposed an approximate representation, and in the case of low atomic number, A≈2Z, the interaction expression can be further approximated as:

[0046]

[0047] The Z and E of the material can be solved by using PCDCT to obtain the dual-energy region data to establish a system of simultaneous equations. Or in the decomposition of the base material, the attenuation coefficient of the scanned substance can be expressed as a linear combination of the base material:

[0048] μ = Σ p w p μ p (4)

[0049] Where w p represents the decomposition coefficient corresponding to the p-th base material. Using the above dual effect formula, the effective atomic number formula can be expressed as:

[0050]

[0051] Where ρ p can be expressed as the electron density or mass density corresponding to the p-th base material, and Z p is the atomic number corresponding to the p-th base material.

[0052] The existing effective atomic number calculation model based on the dual effect (photoelectric effect and Compton scattering) is widely used in the analysis of physical properties of material decomposition. This model can usually simplify the calculation and show high precision in the case of photoelectric effect and Compton scattering as the main interaction. However, this model often ignores the relatively small coherent scattering effect. Although the influence of coherent scattering is weak at high energy, in the medium and low energy range, the contribution of this effect cannot be ignored.

[0053] Through detailed analysis of NIST data, we found that in the commonly used energy range of 40-80keV, the contribution of coherent scattering to the total attenuation coefficient of materials with atomic number between 7-20 usually exceeds 10%, and the highest can reach 16%. This shows that at medium and low energy, coherent scattering occupies a considerable proportion, and ignoring this effect may lead to inaccurate calculation results of effective atomic number, especially in the application scenarios of medium atomic number materials.

[0054] The model of the interaction of X-rays with matter is represented as:

[0055] μ = ρ m (σ ph + σ Compt + σ coherent ) (6)

[0056] In the calculation of effective atomic number, the coefficient m plays a crucial role, so the analysis of m can also reflect the difference between the two models. Related studies show that the index m of photoelectric effect is closely related to the atomic number of the material and the energy of the ray. Our research also found that m is actually a function of energy and effective atomic number. And in the real scan, the linear attenuation coefficient obtained by reconstruction is the result of multiple effects, so in order to deal with the limitations of the simplified double-effect model and obtain more accurate material decomposition results under the corresponding scanning conditions, the index m of photoelectric effect is usually adjusted empirically. For example, in the clinical application of human examination or explosive detection in security inspection application, the spectrum conditions of X-ray are relatively fixed, and the detection objects are usually specific, so m is usually set as a constant between 3 and 4.

[0057] And in the model, after considering the coherent scattering, because the model of scattering is more complex, and in order to better analyze and calculate the variable index m, we will use the de-Compton model σ ′ De_Compt Still in the form of photoelectric effect,

[0058] And the index m is expressed as a variable C(E,Z eff ) about effective atomic number and energy:

[0059]

[0060] Where α1 is a constant, E represents the scanning energy, and Z eff is the effective atomic number of the scanned material. In order to make the model more convenient for solving effective atomic number and density, we will expand C(E,Z eff ) in the following polynomial form:

[0061] C(E,Z eff ) = Σ ij h ij Z eff i E j , 0≤i≤4, 0≤j≤4 (8)

[0062] Where i is the index of Z eff in C(E,Z eff ), j is the index of E, and h ij is expressed as the coefficient of the corresponding term Z eff i E j in the polynomial relationship. In the description of Compton effect, Compton effect is proportional to the atomic number of the scanned material, and KN coefficient f KN(E), which represents the effect of Compton scattering. However, through a deep study of NIST data, we found that this mass density based approximation Compton model is not accurate enough. To better describe the Compton effect, we introduce an additional variable D(E, Z eff ) about E and Z eff to enrich the model:

[0063] σ Compt = α2D(E, Z eff )f KN (E) (9)

[0064] where α2 is a constant. And the same polynomial form of series expansion is used to fit the expression of D(E, Z):

[0065] D(E, Z eff ) =∑ st g st Z eff s E t , 0≤s≤4, 0≤t≤4 (10)

[0066] where s is the index of Z in D(E, Z eff ), t is the index of E, g st is the coefficient of the corresponding term in this polynomial relationship. Therefore, the linear attenuation coefficient μ of the material can be expressed as:

[0067]

[0068] Through this polynomial fitting method, we can more flexibly capture the characteristics of C(E, Z) and D(E, Z) under different energy and atomic number conditions (such as Figure 1 ). This dynamic fitting method not only further improves the accuracy of the model, but also more truly reflects the complexity of the physical effect with the change of the conditions.

[0069] Further, step (2) includes: the purpose of dual-energy calibration is to obtain the effective energy E of the two energy regions. Since X-rays have the characteristics of wide energy spectrum, we use the average energy E as the representative of the effective energy of each energy region, which is equivalent to assuming that the linear attenuation coefficient of the object is obtained by the average energy of the single-energy X-ray. For this purpose, the calibration process requires a known basic material phantom.

[0070] When using known calibration materials to determine the effective energy, E, special attention must be paid to the type and quantity of the selected materials. Compounds or mixtures should be avoided as these complicate the determination of the effective atomic number. Furthermore, a single material can lead to errors in the energy estimate, so at least two or more material types are required. Ideally, the calibration material should be a single element, and materials with excessively high density or atomic number should be avoided to prevent additional errors due to X-ray hardening.

[0071] Furthermore, step (3) includes: reconstructing the scanned material to obtain the linear attenuation coefficients in two energy zones, establishing the double-effect equation in two energy zones using formula (10), and solving the effective atomic number and density:

[0072]

[0073] The effective atomic number Z is solved pixel by pixel for the reconstructed image using the Newton gradient method. eff and material density ρ, and finally achieve the ability to eff The realization coefficient C(E,Z eff ) and (E,Z eff ) Adaptively solve for effective atomic number and density.

[0074] In the accuracy evaluation of decomposition, the absolute value of the decomposition error percentage APE is used for evaluation:

[0075]

[0076] where w T represents the true value, and w represents the value obtained after decomposition.

[0077] Table 1 shows the comparison of the decomposition results.

[0078]

[0079] While specific embodiments of the present invention have been disclosed for illustrative purposes, intended to facilitate understanding and implementation of the present invention, those skilled in the art will appreciate that various substitutions, variations, and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the disclosure of the preferred embodiments, and the scope of protection claimed in the present invention shall be determined by the scope of the claims.

Claims

1. A substance identification method based on adaptive quantitative decomposition, comprising the following steps: 1) Construct a model of interaction between ray and matter μ = ρ m (σ′ De_Compt +σ Compt ); where h m is the mass density, σ′ De_Compt is the de-Compton model including coherent scattering, σ Compt is the Compton scattering cross section between the scanning ray and the material; C(E,Z eff )=Σ ij h ij Z eff i E j ,0≤i≤4,0≤j≤4; α1 is a constant, E is the energy of the scanning ray, Z eff is the effective atomic number of the scanned material, i is C(E,Z eff )Z eff The index of E, j is the index of E, h is the index of ij is the corresponding term Z in the polynomial relation eff i E j The coefficient of σ Compt = α2D(E,Z eff )f KN (E), D(E, Z eff )=Σ st g st Z eff s E t ,0≤s≤4,0≤t≤4, where s is D(E,Z eff )Z eff The index of D(E,Z eff ) in which E is the exponent, g st is the corresponding term Z in the polynomial relation eff s E t coefficient, α2 is a constant, f KN (E) is the Klein-Ceshima coefficient; 2) Using the relevant data of the selected calibration object, the variable parameters C(E, Z eff )、D(E,Z eff ) to fit and establish variable parameters C(E,Z eff ) and energy E, effective atomic number Z eff The specific relationship between them, and the variable parameters D(E,Z eff ) and energy E, effective atomic number Z eff The specific relationship between them; 3) Using C(E,Z eff )、D(E,Z eff ) and combined with the known effective atomic number and density information of various materials in the selected calibration object, the effective energy of each energy zone is solved to obtain the effective energy E of each material in the low energy zone. L and the effective energy E in the high energy region H ; 4) Scan and reconstruct the object to be detected according to the scanning parameters of the calibration object to obtain the linear attenuation coefficient of the object to be detected; then, based on the linear attenuation coefficient of the object to be detected, combined with C(E, Z eff )、D(E,Z eff ) and the effective energy E of each known material in the low energy region obtained in step 3) L , effective energy E in the high energy zone H , obtain the linear attenuation coefficient of the detected object in the low-energy region and the linear attenuation coefficient of the high-energy region and substitute them into the radiation-matter interaction model to solve the effective atomic number and material density of the detected object.

2. The method according to claim 1, characterized in that Get the effective energy E of each material in the low energy region L and the effective energy E in the high energy region H The method is as follows: first, scan and reconstruct the selected calibration object to obtain the linear attenuation coefficient μ of each material in the calibration object in the high energy region. h , linear attenuation coefficient μ in the low energy region l ; Using the known physical properties of the calibration material, and then solving Get the effective energy E of the corresponding material in the low energy zone L and the effective energy E in the high energy region H .

3. The method according to claim 2, characterized in that In step 4), the method for solving the effective atomic number and material density of the object under test is: scan and reconstruct the object under test using energy spectrum CT to obtain the linear attenuation coefficient μ of the object under test in the low energy region. l ′ , linear attenuation coefficient μ in the high energy region ′ h ; Use the effective energy of the two energy regions to estimate the effective energy E of the low energy region L and the effective energy E in the high energy region H , and then use the Newton gradient method to Solve pixel by pixel to obtain the effective atomic number Z of the object being detected eff and material density ρ m .

Images