Optimal decomposition method of basis material for dual-energy CT images

By using all possible triple-based substance combinations and projection methods, the need for four basic substances decomposition in dual-energy CT images is solved, improving the accuracy and objectivity of the decomposition, and reducing the influence of human factors.

CN114376592BActive Publication Date: 2024-11-01LIAONING CAMPO IMAGING MEDICAL CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210028258.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-11
Publication Date
2024-11-01
Estimated Expiration
2042-01-11

AI Technical Summary

Technical Problem

The existing dual-energy CT image multiple basic substance decomposition technology cannot meet the quantitative requirements of four basic substances: liver tissue, blood, fat and contrast agent in the pixels of liver image, and the solutions of mathematical models in traditional methods are affected by human factors. Accuracy and objectivity.

Method used

All possible combinations of triple-based substances are used to decompose the substances, and by calculating the volume fraction of each base substance at the corresponding position of the pixel point, using the formula μM(Ej) = α1μ1(Ej) + α2μ2(Ej) + α3μ3(Ej), Where α1 + α2 + α3 = 1, α1 ≥ 0, α2 ≥ 0, α3 ≥ 0, α3 ≥ 0, traverse all triple-based substance combinations, select the solution under the smallest ternary-based substance combination, and use the projection method when there is no solution Obtain an approximate solution.

Benefits of technology

It improves the accuracy and objectivity of basic material decomposition, weakens the influence of human factors, and can select the optimal solution from a variety of possible solutions, which is suitable for the analysis of basic material components at different pixel points.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114376592B_ABST
    Figure CN114376592B_ABST
Patent Text Reader

Abstract

The present invention relates to a method for optimizing the decomposition of base materials of dual-energy CT images. The base material components of each pixel point are determined by calculating the volume fraction of each base material in the pixel point. The volume fraction of each base material in the pixel point is calculated using a triple base material list containing all triple base material combinations, and then the final base material component is determined based on the #imgabs0# (#imgabs1#) norm of the decomposed base material component; when there is no solution for the volume fraction of the base material, the linear attenuation coefficients of the pixel point and each base material at two different energies are defined as multiple points in the same plane, and then the points corresponding to the pixel point are projected onto a convex hull composed of the points corresponding to each base material, and the projection points are used instead of the pixel point to obtain an approximate solution for the volume fraction of each base material in the pixel point. The present invention does not need to artificially define the number and sequence of triple base material elements in the triple base material list before material decomposition, thereby improving and ensuring the accuracy and objectivity of the decomposition result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method for optimizing the decomposition of base materials of dual-energy CT images, and belongs to the technical field of medical imaging. Background Art

[0002] With the increasing application of dual-energy CT imaging technology in clinical practice, the clinical requirements for the decomposition of dual-energy CT images under multiple base substances are becoming higher and higher. For example, in the quantitative analysis of fatty liver, the clinic needs to analyze the distribution of four base substances in each pixel of the liver dual-energy CT image: liver tissue, blood, fat, and contrast enhancer (such as iohexol). However, since dual-energy CT technology only uses two different energies for imaging, it is essentially unable to meet the requirements for the decomposition of multiple base substances (>2).

[0003] The current development of dual-energy CT image multiple-substance decomposition technology can be briefly summarized as follows: by assuming mass conservation and / or volume conservation, a mathematical model is established, and by solving this mathematical model, dual-energy CT images can be decomposed into multiple basis materials. The specific solution method is to perform three specified basis material decompositions on each pixel in the image according to the spatial position of its corresponding linear attenuation coefficient. [2] For different pixels, different three basic substances are specified according to the different spatial positions of the corresponding linear attenuation coefficients to solve, so as to achieve the purpose of decomposing the pixels of the dual-energy CT image into multiple basic substances. For example, in the quantification of fatty liver, the three basic substances of liver tissue, blood and fat can be decomposed for pixel one, and the three basic substances of liver tissue, blood and contrast agent can be decomposed for pixel two, so as to achieve the quantification requirements of the four basic substances of liver tissue, blood, fat and contrast agent in the pixels of the liver image. This basic material decomposition method uses a triple list method to solve the mathematical model (equation), and rigidly stipulates that each pixel in the dual-energy CT image can have at most three basic material components, and different pixels can have different three basic material components. A triple list is given (the elements in the list are triple basic materials). When the mathematical model obtains a feasible solution based on the first triple element in the triple list, it will no longer continue to solve based on the last triple element in the triple list. Therefore, the triple list given in advance and the arrangement order of the basic material triplets in the list are very important, which determines the existence and uniqueness of the solution of the mathematical model. So which triplets are selected to enter the list and how to determine the order between triplets? The traditional solution is to use prior knowledge to make artificial regulations before doing basic material decomposition. Human factors play a dominant role in it, which will inevitably affect the accuracy and objectivity of the feasible solution of the mathematical model. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for optimizing the decomposition of base materials of dual-energy CT images, which uses all possible combinations of triple base materials to perform material decomposition. There is no need to strictly stipulate the order of triple base material elements in the triple base material list before material decomposition, and it is more conducive to reflecting the actual situation.

[0005] The technical solution of the present invention to achieve the above-mentioned purpose is: a method for optimizing the decomposition of base materials of dual-energy CT images, using the following formula as a formula for calculating the volume fraction of each base material at the position corresponding to the pixel point, and calculating the volume fraction of each base material at the position corresponding to the pixel point:

[0006]

[0007] st

[0008] μ M (E1)=α1μ1(E1)+α2μ2(E1)+μ3(E1)

[0009] μ M (E2)=α1μ1(E2)+α2μ2(E2)+α3μ3(E2)

[0010] α1+α2+α3=1

[0011] and

[0012] α1≥0, α2≥0, α3≥0

[0013] {base1,base2,base3}∈ triple base list

[0014] in,

[0015] μ M (E j ) is the energy E j Measured value of linear attenuation coefficient under ;

[0016] μ i (E j ) is the base substance i at energy E j The linear attenuation coefficient under ;

[0017] α i is the volume fraction of base material i at the corresponding position of the pixel;

[0018] i = 1, 2, 3, which are the sequential numbers of the same group of base substances;

[0019] j = 1, 2, is the sequential number of the energy level;

[0020] 0<p<1 is an adjustable factor, which can be determined based on experiments or experience.

[0021] The elements of the triple base substance list are triple base substances (or triple base substance groups), and the list includes all ternary combinations of all base substances used, wherein the total number of base substances used N≥3.

[0022] By calculating the volume fraction of each base substance in the pixel, the material composition of the area corresponding to the pixel can be determined or analyzed.

[0023] All ternary combinations of the base substances used (combinations of three base substances) should be traversed, and if there are solutions for at least some of the combinations, the solution should be The solution result under the minimum ternary basic substance combination.

[0024] Mode It refers to the function of the ternary basic substance group as a variable Find the minimum value.

[0025] When the formula for calculating the volume fraction of each base substance at the position corresponding to the pixel point has no solution using some elements of the ternary base substance list (ternary base substance group), the corresponding elements are ignored.

[0026] Since the solution is performed for specific pixels, a ternary basis material group that has no solution at one pixel may not have no solution at another pixel.

[0027] Similarly, the ternary basic substance groups obtained by solving each pixel point do not need to be the same. In fact, due to the different actual composition of human body substances, several different ternary basic substance combinations are usually obtained in the same image analysis.

[0028] Usually, the volume fraction of the base material should be solved for different pixel points separately.

[0029] Usually, the base materials corresponding to different pixels are independent of each other, and the base material groups corresponding to adjacent pixels may be the same or different.

[0030] In the case where the formula for calculating the volume fraction of each base substance at the corresponding position of the pixel point has no solution (there is no solution for all triple base substances), point D is placed in the plane (μ(E1), μ(E2)) M (μ M (E1), μ M (E2)) Projection to all points D n (μ n (E1), μ n (E2)) is the convex hull of the vertices, and the projection point D with the smallest projection distance is determined on the convex hull d (μ d (E1), μ d (E2)),

[0031] Such as D d With any vertex D in the convex hull k (μ k (E1), μ k (E2)) overlap, then:

[0032]

[0033] Such as D d Any two adjacent vertices D in the convex hull k1 (μ k1 (E1), μ k1 (E2)) and D k2 (μ k2 (E1), μ k2 (E2)), then:

[0034]

[0035] Wherein, n=1, 2, ..., N, is the sequential number of all base substances used.

[0036] Before solving the formula for calculating the volume fraction of each base material at the corresponding position of the pixel point, the left-on algorithm can be used to determine the point D in the plane (μ(E1), μ(E2)) M (μ M (E1), μ M (E2)) is located at all points D n (μ n (E1), μ n (E2)) is inside or on the convex hull of a vertex, such as point D M (μ M (E1), μ M (E2)) is located at all points D n (μ n (E1), μ n (E2)) is inside or on the convex hull of the vertex, then it is judged that the formula for calculating the volume fraction of each base substance at the corresponding position of the pixel point has a solution, such as point D M (μ M (E1), μ M (E2)) is located at all points D n (μ n (E1), μ n (E2)) is outside the convex hull of the vertex, it is judged that the formula used to calculate the volume fraction of each base substance at the corresponding position of the pixel point has no solution.

[0037] The beneficial effect of the present invention is that when decomposing the basis material components of pixel points, the present invention uses a general triple basis material list containing all triple basis material combinations for calculation, and then determines the final basis material components based on the p-times (0<p<1) norm of the basis material components obtained by decomposition. There is no need to strictly stipulate the number and sequence of triple basis material elements in the triple basis material list before material decomposition, which weakens or abandons the link of artificially using prior knowledge to stipulate the triple basis material list, and can select the optimal solution from several possible solutions, thereby improving and ensuring the accuracy and objectivity of basis material decomposition. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a schematic diagram of a distribution mode of the coordinate points of the measured values ​​in the plane (μ(E1), μ(E2)) relative to the coordinate points of each basis substance;

[0039] Figure 2 It is a schematic diagram of another distribution mode of the coordinate points of the measured values ​​in the plane (μ(E1), μ(E2)) relative to the coordinate points of each base substance. DETAILED DESCRIPTION

[0040] The present invention discloses a method for optimizing the decomposition of base materials in dual-energy CT images. The material composition of each pixel is determined or analyzed by calculating the volume fraction of each base material in the pixel (the corresponding part of the pixel). It is stipulated that a pixel can have at most three base material components, and different pixels can have different base material components. (Measure the two-dimensional vector of the linear attenuation coefficient of the pixel at two different energies through the dual-energy CT image), use each element in the triplet basis material list to solve the following equation:

[0041]

[0042] st

[0043] μ M (E1)=α1μ1(E1)+α2μ2(E1)+μ3(E1) (2)

[0044] μ M (E2)=α1μ1(E2)+α2μ2(E2)+α3μ3(E2) (3)

[0045] α1+α2+α3=1 (4)

[0046] and

[0047] α1≥0, α2≥0, α3≥0 (5)

[0048] {base 1, base 2, base 3}∈ triple base list (6)

[0049] Among them, formula (6) can also be expressed by the following formula:

[0050]

[0051] Among them, the measured value μ of the linear attenuation coefficient at different energies E1 and E2 can be M (E1) and μ M (E2) is expressed as μ M (E j ), j = 1, 2, is the sequential number of the energy levels;

[0052] The linear attenuation coefficients of the three base materials (base material 1, base material 2 and base material 3) in the same triplet base material at different energies E1 and E2 are μ1(E1) and μ1(E2), μ2(E1) and μ2(E2) and μ3(E1) and μ3(E2) respectively, which are expressed as μ i (E j ), where i = 1, 2, 3, is the sequential number of the same group of base substances;

[0053] α i is the volume fraction of base material i at the corresponding position of the pixel;

[0054] 0<p<1 is an adjustable factor, which can be determined based on experiments or empirical values.

[0055] Typically, the total number of base materials used, N≥3.

[0056] Select 3 base substances from the base substances in N and combine them, with a total of combinations, and use each combination as an element to make a triplet base material list, which includes elements.

[0057] When there is one or more feasible solutions satisfying equations (2)-(6), take |α1| p +|α2| p +|α3| p The feasible solution with the smallest value is taken as the result (corresponding to the final solution of the pixel point). When there is only one feasible solution, the feasible solution is regarded as |α1| p +|α2| p +|α3| p The feasible solution with the minimum value is taken as the result.

[0058] In formula (1), the parameter p is a positive number less than 1, for example, p=0.5, or p=0.95, to adapt to the sparse characteristics of the ternary-based material. The optimized empirical value of p can be obtained based on experiments in relevant situations.

[0059] Preferably, a method similar to the existing triplet-based substance list method is used. [2] Solve equations (2)-(5).

[0060] Although equations (2)-(5) are just a simple 3×3 linear equation system plus a non-negative constraint, in fact, equations (2)-(5) may have solutions or may not have solutions. Even if they have solutions, the solutions will be different for the three different base materials.

[0061] When there is no feasible solution, the projection method can be used to obtain the approximate solution of the pixel point, that is, the linear attenuation coefficient μ of each basis material n at two different energies n (E1) (linear attenuation coefficient at energy E1) and μ n (E2) (linear attenuation coefficient at energy E2) is defined as a number of points D on the plane (μ(E1), μ(E2)) n (μ n (E1), μ n (E2)), the measured value of the linear attenuation coefficient corresponding to the pixel point μ M (E1) (measured value of linear attenuation coefficient at energy E1) and μ M (E2) (measured value of the linear attenuation coefficient at energy E2) is defined as a point D on the plane (μ(E1), μ(E2)) M (μ M (E1), μ M (E2)), point D M Project to all points D n On the convex hull formed by the vertices, the projection point with the smallest projection distance is used to replace the pixel point, and the approximate solution of the volume fraction of each basic substance in the pixel point is obtained. The so-called plane (μ(E1), μ(E2)) is a plane in a rectangular coordinate system with μ(E1) and μ(E2) as the horizontal and vertical coordinates respectively (or a plane in an equivalent coordinate system), where μ(E1) is the linear attenuation coefficient under energy E1, and μ(E2) is the linear attenuation coefficient under energy E2, where n=1, 2, ..., N, is the number of all the basic substances used.

[0062] Specifically, it can be:

[0063] Define the following two-dimensional vector:

[0064]

[0065]

[0066]

[0067] μ n (E1) is the linear attenuation coefficient of the base substance n at energy E1, μ n (E2) is the linear attenuation coefficient of the base substance n at energy E2, is a vector with the linear attenuation coefficient of the base substance n at two different energies as the corresponding coordinate value, μ M (E1) and μ M (E2) are the measured values ​​of the linear attenuation coefficient of the pixel point in the dual-energy CT image at energy E1 and energy E2, It is a vector with the measured value of the linear attenuation coefficient at the pixel point as the corresponding coordinate value at two different energies.

[0068] The linear attenuation coefficient of the base substance n at two different energies can be expressed as a point D on the plane (μ(E1), μ(E2)) n (μ n (E1), μ n (E2)), similarly, the relevant measured value can also be expressed as a point D on the plane (μ(E1), μ(E2)) M (μ M (E1), μ M (E2)).

[0069] like Figure 1 As shown in the figure, in the quantification problem of fatty liver, it is assumed that four basic substances are used, including liver tissue A, blood B, contrast agent C and fat D, respectively. and Denote the corresponding vectors of these basis substances, in Figure 1 In the figure, points A, B, C, and D under the corresponding coordinates are used to represent the measured value vector corresponding to the two pixel points. The measured values ​​are represented by points F and G respectively. The elements in the triple basis material list are triple basis materials (three basis material combinations) in basis materials A, B, C, and D, and include all triple basis material combinations. The essence of using the triple basis material list to solve equations (2)-(5) is to use different three basis materials to solve α1, α2, and α3 for different points. When equations (2)-(5) have solutions (points F and G fall in the quadrilateral surrounded by ABCD, such as Figure 1 As shown), use the triplet basis material list to find all feasible solutions of α1, α2 and α3, and use formula (1) to get |α1| p +|α2| p +|α3| p The solution with the smallest value is taken as the final solution.

[0070] like Figure 2 As shown, the unsolvable point F or G is outside the convex hull. Project it onto the convex hull surrounded by the four points A, B, C, and D. The projection point falls on the boundary of the convex hull. Use the projection point with the smallest projection distance to replace the unsolvable point. Use the basis material at both ends of the projection point to solve the corresponding α n . There are two situations: the point closest to point F from the convex hull ABCD is F' (coincides with point C), and the point closest to point G from the convex hull ABCD is point G', and GG' is perpendicular to the boundary BC. The approximation of point F removes the volume fraction of base material C, which is 1, while the volume fractions of other base materials are all 0. Point G only has the volume fractions of base materials B and C, and the volume fractions of other base materials are all 0. Among them, the volume fraction α2 of base material B is |G′C| / |BC|, and the volume fraction α3 of base material C is |BG′| / |BC|. Among them, |·| represents the length of the line segment.

[0071] The solution can be performed for any specific pixel point (any point in the image) to obtain the result of the pixel point, or, when appropriate, for all pixels in the image or all pixels in a part of the image or a set of pixels under other conditions, to obtain the result based on the change of pixel coordinates (with the pixel coordinates as the independent variable).

[0072] For solutions to formulas and other related matters involved in this manual, please refer to the references provided in this manual and other existing relevant documents or materials.

[0073] The measured value (or similar expression) referred to in this specification is usually the value of a pixel in the image obtained by dual-energy CT image reconstruction technology at the energy E j The linear attenuation coefficient under the condition of E can also be used to reflect or characterize the energy E corresponding to the position (region) of a certain pixel before reconstruction. j Corresponding data of linear attenuation coefficient under energy E, or other data that can reflect or characterize j The data below are the actual measured data of the linear attenuation coefficient.

[0074] When using symbols (including symbols in formulas) to express functions, vectors, points, etc., in this specification, for the sake of convenience, variables, coordinates, etc. listed in brackets after the symbols referring to the functions, vectors, or points may be omitted together with the brackets. Some subscripts may also be omitted if there is no ambiguity.

[0075] References

[0076] 1.P.R.S.Mendonca,P.Lamb and D.V.Shani,“A Flexible Method for Multi-material Decomposition of Dual-Energy CT Images,”IEEE Trans.Med.Imag.vol.33,pp.99-116,2014.

[0077] 2.X.Liu,L.Yu,A.N.Primak and C.H.McCollough,“Quantitative imaging ofelement composition and mass fraction using dual-energy CT:Three-materaildecomposition,”Med.Phys.vol.36,pp.1602-1609,2009.

[0078] 3.J.O’Rourke,Computational Geometry in C,2 nd ed.Cambridge UniversityPress,1998.

[0079] 4.A.L.Peressini,F.E.Sullivan and J.J.Hul,Jr.,The Mathematics ofNonlinear Programming,Springer,1988.

Claims

1. A method for optimizing the decomposition of dual-energy CT images, characterized in that The following formula is used to calculate the volume fraction of each base substance at the corresponding position of the pixel, and the volume fraction of each base substance at the corresponding position of the pixel is calculated: S.t. m M (E1)=α1μ1(E1)+α2μ2(E1)+μ3(E1) m M (E2)=α1μ1(E2)+α2μ2(E2)+α3μ3(E2) α1+α2+α3=1 and α1≥0,α2≥0,α3≥0 {Base substance 1, Base substance 2, Base substance 3} ∈ the triple base substance list. Among them, μ M (E j ) is the energy E j Measured value of linear attenuation coefficient under ; μ i (E j ) is the base substance i at energy E j The linear attenuation coefficient under ; α i is the volume fraction of base material i at the corresponding position of the pixel; i = 1, 2, 3, is the sequential number of the base substances in the same group; j = 1, 2, is the sequential number of the energy levels; 0 < p < 1 is an adjustable factor, the elements of the triple base substance list are triple base substances, and the list includes all triple combinations of all the used base substances. Among them, the total number of used base substances N ≥ 3, When there is no solution for the formula for calculating the volume fraction of each base material at the corresponding position of the pixel point, point D is M (μM(E1), μM(E2)) is projected to all points D n (μ n (E1), μ n (E2)) is the convex hull of the vertices, and the projection point D with the smallest projection distance is determined on the convex hull d (μ d (E1), μ d (E2)), Such as D d and any vertex D in the convex hull k (μ k (E1), μ k (E2)) overlap, then: Such as D d Any two adjacent vertices D in the convex hull k1 (μ k1 (E1), μ k1 (E2)) and D k2 (μ k2 (E1), μ k2 (E2)), then: where n = 1, 2, ……, N, is the sequential number of all the used base substances.

2. The method for optimizing the decomposition of the basis material of the dual-energy CT image according to claim 1, characterized in that When the formula for calculating the volume fraction of each base substance at the corresponding position of the pixel has no solution for some elements of the triple base substance list, the corresponding elements are ignored.

3. The method for optimizing the decomposition of the basis material of the dual-energy CT image according to claim 1, characterized in that The volume fraction of the base substance is solved for different pixels respectively, and the base substances corresponding to different pixels are independent of each other.

4. The method for decomposing dual-energy CT images according to any one of claims 1 to 3, characterized in that Before solving the formula for calculating the volume fraction of each base material at the corresponding position of the pixel point, the left-on algorithm is used to determine the point D in the plane (μ(E1), μ(E2)) M (μ M (E1), μ M (E2)) is located at all points D n (μ n (E1), μ n (E2)) is inside or on the convex hull of a vertex, such as point D M (μ M (E1), μ M (E2)) is located at all points D n (μ n (E1), μ n (E2)) is inside or on the convex hull of the vertex, then it is judged that the formula for calculating the volume fraction of each base substance at the corresponding position of the pixel point has a solution, such as point D M (μ M (E1), μ M (E2)) is located at all points D n (μ n (E1), μ n (E2)) is outside the convex hull of the vertex, it is judged that the formula used to calculate the volume fraction of each base substance at the corresponding position of the pixel point has no solution.