A method for simulating magnetic resonance signals of fatty liver tissue

Abstract

The application discloses a method for simulating a magnetic resonance signal of fatty liver tissue, which comprises the following five steps: obtaining a proton density fat fraction (PDFF) and an effective transverse relaxation rate parameter of the fatty liver tissue; and the method has the beneficial effects that: a three-dimensional simulation model of the fatty liver tissue with different fat fractions is simulated, the PDFF and the predicted value corresponding to the fat fraction are obtained, and thus a correction relationship between the fat fraction and the parameter is established; meanwhile, by adjusting the parameters (magnetic field strength, etc.) in the three-dimensional simulation model of the fatty liver tissue, the influence of fat on the parameter can be corrected in different scenes; and the method can improve the quantification accuracy of liver iron and liver fat through the correlation between the fat fraction and the parameter.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of magnetic resonance signal simulation model construction, in particular to a magnetic resonance signal simulation method for fatty liver tissue. BACKGROUND

[0002] Fatty liver is the abnormal accumulation of triglycerides in hepatocytes, which is the most common chronic liver disease and affects about 25% of the world's population. As a hallmark feature of non-alcoholic fatty liver disease, it is mainly caused by liver lipid metabolism disorder caused by obesity and insulin resistance, which can gradually develop into liver fibrosis, cirrhosis and hepatocellular carcinoma. Non-alcoholic fatty liver disease is also related to systemic iron regulation, and iron regulation disorder is often observed in patients, about one-third of patients have iron overload syndrome. Iron overload not only aggravates the severity of fatty hepatitis characterized by inflammatory response and hepatocyte necrosis through oxidative overreaction, but also promotes hepatocyte steatosis by inhibiting the secretion of very low density lipoprotein.

[0003] The chemical shift encoding magnetic resonance imaging technology developed in recent years can simultaneously measure the proton density fat fraction (Proton Density Fat Fraction, PDFF) and effective transverse relaxation rate parameter , which are used to reflect the concentrations of triglycerides and iron, respectively. Clinical studies have shown that there is a certain correlation between liver steatosis and , but the correlation has not been mathematically characterized, and the biophysical mechanism between the two is not clear. It is very important to accurately correct the quantification of liver iron deposition and fat deposition. On the one hand, for patients with high steatosis and mild iron overload, , the overestimation of the parameter will lead to a false positive diagnosis; on the other hand, when iron deposition is severe, it will affect the accuracy of the fat quantification result. Understanding the influence of liver fat on the magnetic resonance parameter will help more accurately quantify liver iron deposition and fat deposition, which will be of great significance to the diagnosis, staging and treatment of patients with fatty liver and iron overload. The correlation between the fat fraction and the parameter can be corrected through a large number of clinical experiments, but the process will consume a lot of manpower and material resources due to the influence of factors such as magnetic field strength, manufacturer equipment and scanning scheme.

[0004] Therefore, in view of the deficiencies in the prior art, it is necessary to provide a magnetic resonance signal simulation method for fatty liver tissue to solve the deficiencies in the prior art. SUMMARY

[0005] The present application aims to avoid the deficiencies of the prior art and provides a magnetic resonance signal simulation method for fatty liver tissue. The magnetic resonance signal simulation method for fatty liver tissue can improve liver iron and liver fat quantification accuracy through the correlation between fat fraction and

[0006] The above-mentioned object of the present application is achieved by the following technical measures:

[0007] The present application provides a magnetic resonance signal simulation method for fatty liver tissue, comprising the following steps:

[0008] Step (1), a three-dimensional simulation model of fatty liver tissue with different fat fractions is constructed according to liver biopsy samples and three-dimensional morphological distribution of fat droplets, and then model parameters of the three-dimensional simulation model of fatty liver tissue are set, and step (2) is entered;

[0009] Step (2), the three-dimensional simulation model of fatty liver tissue is simulated under the action of a main magnetic field to generate a non-uniform magnetic field, and step (3) is entered;

[0010] Step (3), a simulation proton is put into the three-dimensional simulation model of fatty liver tissue, and the phase accumulation of each simulation proton is calculated, and step (4) is entered;

[0011] Step (4), the water signal and fat signal of all simulation protons of the three-dimensional simulation model of fatty liver tissue are synthesized, and then the magnetic resonance signal is obtained by superposition;

[0012] Step (5), the proton density fat fraction (PDFF) and effective transverse relaxation rate parameter T2* are obtained by analyzing and calculating the magnetic resonance signal obtained in step (4);

[0013] Preferably, step (2) is specifically that under the action of a simulation main magnetic field, a simulation fat droplet of the three-dimensional simulation model of fatty liver tissue generates a dipole field, and the dipole fields of all simulation fat droplets are superimposed to obtain a non-uniform magnetic field.

[0014] The dipole field of each simulation fat droplet at different positions in the three-dimensional simulation model of fatty liver tissue is obtained by formula (I):

[0015]

[0016] Where, ΔB is the non-uniform magnetic field, B0 is the strength of the simulation main magnetic field, χ L is the fat magnetic susceptibility coefficient, R is the radius of the simulation fat droplet, r is the distance from the observation point to the center of the simulation fat droplet in the three-dimensional simulation model of fatty liver tissue, and θ is the included angle between the center line of the simulation fat droplet and the observation point and the direction of the simulation main magnetic field.

[0017] In step (3), the initial position of the simulated proton in the three-dimensional simulated model of fatty liver tissue is randomly generated by computer software.

[0018] Preferably, the motion of the simulated proton in the three-dimensional simulated model of fatty liver tissue is isotropic Gaussian diffusion, and the average absolute displacement is where D is the diffusion coefficient of the simulated proton, and δ is the motion time interval of the simulated proton.

[0019] The simulated proton with an initial position outside the simulated fat droplet is defined as a water-simulated proton, and the diffusion coefficient D of the water-simulated proton is set to be 2 / ms, and 0.1 μm 2 / ms 2 / ms.

[0020] The simulated proton with an initial position inside the simulated fat droplet is defined as a simulated fat droplet-simulated proton, and the diffusion coefficient D of the simulated fat droplet-simulated proton is set to be 2 / ms.

[0021] When the water-simulated proton moves in the three-dimensional simulated model of fatty liver tissue, if the water-simulated proton contacts the boundary of the three-dimensional simulated model of fatty liver tissue, the water-simulated proton is made to be injected from the symmetric point of the other side boundary of the three-dimensional simulated model of fatty liver tissue to the inside of the three-dimensional simulated model of fatty liver tissue according to the symmetry principle.

[0022] If the water-simulated proton penetrates the simulated fat droplet, the water-simulated proton is moved to the point where the surface of the simulated fat droplet intersects with the motion path of the water-simulated proton.

[0023] In step (3), the phase accumulation of each simulated proton at time t is calculated by formula (II),

[0024]

[0025] where γ is the magnetic spin ratio constant 2.675×10 8 rad·s -1 ·T -1 , i is the step number of the simulated proton, p(i) is the position coordinate of the simulated proton at the i-th step, and t is the time of the motion of the simulated proton.

[0026] Preferably, step (4) is specifically superimposing the water signal of the water-simulated proton and the fat signal of the simulated fat droplet-simulated proton in the three-dimensional simulated model of fatty liver tissue to obtain the magnetic resonance signal.

[0027] Preferably, the water signal S wThe fat signal S of the simulated protons in the simulated fat droplet is obtained from equation (III). f It is obtained from equation (IV);

[0028]

[0029]

[0030] Where S(0) is the initial signal value, R20 is the empirical value of the liver's transverse relaxation rate, P is the number of lipid spectral peaks, and α p f represents the relative amplitude of the fat signal. F，p denoted by , p represents the frequencies of multiple spectral peaks of the fat signal relative to the spectral peaks of water, where p is the index of the spectral peak.

[0031] Preferably, the above magnetic resonance signal is obtained by equation (V);

[0032] S(t)=S w (t)+S f (t) Equation (V).

[0033] The magnetic resonance signal simulation method for fatty liver tissue of the present invention uses a spin echo sequence, given the intensity B0 of the simulated main magnetic field, and sets different echo times to obtain magnetic resonance signals with different echo times.

[0034] Preferably, step (5) above includes:

[0035] Step 5.1: For the magnetic resonance signal S(t) obtained in step (4), perform curve fitting using the chemical shift encoded imaging signal model to obtain ρ. w , ρ f and The predicted values ​​of the parameters, where the chemical shift encoded imaging signal model is represented by equation (VI);

[0036]

[0037] Where, ρ w ρ represents the amplitude of the water signal. f The amplitude of the fat signal is φ0, the initial phase is f. B This is a frequency shift caused by magnetic field inhomogeneity;

[0038] Step 5.2, based on ρ obtained in Step 5.1 w , ρ f Equation (VII) yields PDFF.

[0039]

[0040] Preferably, the above-mentioned fat liver tissue three-dimensional simulation model can be considered to be composed of a plurality of simulated fat droplets and water;

[0041] Preferably, the above-mentioned a is 0.76 μm 2 / ms.

[0042] Preferably, the above-mentioned model parameters are at least one of a fat magnetic susceptibility coefficient, a diffusion coefficient of simulated protons in a medium, TE or R20.

[0043] Preferably, the above-mentioned number of simulated protons is 5000-20000.

[0044] The fat liver tissue magnetic resonance signal simulation method of the present application comprises the following steps: step (1), constructing a fat liver tissue three-dimensional simulation model with different fat fractions according to the three-dimensional morphological distribution of liver biopsy samples and fat droplets, and then setting the model parameters of the fat liver tissue three-dimensional simulation model to enter step (2); step (2), constructing a fat liver tissue three-dimensional simulation model with different fat fractions according to the three-dimensional morphological distribution of liver biopsy samples and fat droplets, and then setting the model parameters of the fat liver tissue three-dimensional simulation model to enter step (3); step (3), putting simulated protons into the fat liver tissue three-dimensional simulation model, accumulating the phase of each simulated proton, and entering step (4); step (4), synthesizing the water signal and fat signal of all simulated protons in the fat liver tissue three-dimensional simulation model, and then superimposing to obtain a magnetic resonance signal; and step (5), analyzing and calculating the magnetic resonance signal obtained in step (4) to obtain proton density fat fraction PDFF and effective transverse relaxation rate parameters The fat liver tissue magnetic resonance signal simulation method has the following beneficial effects: for a fat liver tissue three-dimensional simulation model with different fat fractions, PDFF and correction value corresponding to the fat fraction can be obtained through simulation, so that a correction relationship between the fat fraction and parameters can be established. At the same time, by adjusting various parameters (magnetic field strength, etc.) in the fat liver tissue three-dimensional simulation model, the influence of fat on parameters can be corrected in different scenarios. Through the correlation between the fat fraction and parameters, the fat liver tissue magnetic resonance signal simulation method can improve the quantification accuracy of liver iron and liver fat. BRIEF DESCRIPTION OF DRAWINGS

[0045] The present application will be further described with reference to the accompanying drawings, but the content in the drawings does not constitute any limitation on the present application.

[0046] Figure 1 The flowchart of the fat liver tissue magnetic resonance signal simulation method of the present application.

[0047] Figure 2A three-dimensional simulation model of fatty liver tissue with a fat fraction of 5%, 15%, and 20%.

[0048] Figure 3 A non-uniform magnetic field distribution of three random layers of a three-dimensional simulation model of fatty liver tissue with a fat fraction of 20% at 1.5T and 3.0T field strength.

[0049] Figure 4 A water signal, a fat signal, and a magnetic resonance signal of different echo times of fatty liver tissue with a fat fraction of 3% at 1.5T and 3.0T field strength.

[0050] Figure 5 A magnetic resonance signal of different echo times of fatty liver tissue with a fat fraction of 5%, 15%, and 20% at 1.5T and 3.0T field strength.

[0051] Figure 6 A relationship between a simulated and predicted value and a fat fraction at 1.5T and 3.0T field strength.

[0052] Figure 7 A relationship between a simulated and predicted simulated proton density fat fraction and a fat fraction at 1.5T and 3.0T field strength. DETAILED DESCRIPTION

[0053] The technical solutions of the present application are further described in conjunction with the following examples.

[0054] Example 1.

[0055] A magnetic resonance signal simulation method for fatty liver tissue, as shown in Figure 1 , includes the following steps:

[0056] Step (1), a three-dimensional simulation model of fatty liver tissue with different fat fractions is constructed according to liver biopsy samples and three-dimensional morphological distribution of fat droplets, and then model parameters of the three-dimensional simulation model of fatty liver tissue are set, and step (2) is entered;

[0057] Step (2), a main magnetic field acts on the three-dimensional simulation model of fatty liver tissue to generate a non-uniform magnetic field, and step (3) is entered;

[0058] Step (3), a simulation proton is put into the three-dimensional simulation model of fatty liver tissue, and the phase accumulation of each simulation proton is accumulated, and step (4) is entered, wherein the number of simulation protons is 5000-20000;

[0059] Step (4), the water signal and the fat signal of all simulation protons of the three-dimensional simulation model of fatty liver tissue are synthesized, and then the magnetic resonance signal is obtained by superposition;

[0060] Step (5), analyzing and calculating the magnetic resonance signal obtained in step (4) to obtain the proton density fat fraction PDFF and the effective transverse relaxation rate parameter

[0061] It should be noted that the construction of the fatty liver tissue three-dimensional simulation model is the prior art, and is not the focus of the present application. The specific method should be known to those skilled in the art, and can be referred to the application number 202110168473.2, a three-dimensional simulation model construction method of fatty liver tissue. The model parameters of the present application are at least one of the fat magnetic susceptibility, the diffusion coefficient of the simulation proton in the medium, TE or R20;

[0062] It should be noted that the number of simulation protons depends on the size of the fatty liver tissue three-dimensional simulation model. Too few simulation protons will cause instability of the simulation signal, and too many simulation protons will result in a long calculation time. When the number of simulation protons is in the range of 5000-20000, a relatively stable simulation result can be obtained. Because the principle of magnetic resonance imaging is to use hydrogen simulation protons for imaging, on the basis of the simulation of the non-uniform magnetic field, the isotropic Gaussian diffusion of the simulation protons, i.e. the change of the spatial position of the simulation protons, will cause the change of the phase in the non-uniform magnetic field, thereby synthesizing the magnetic resonance signal.

[0063] Specifically, step (2) is to generate a dipole field of the simulation fat droplet of the fatty liver tissue three-dimensional simulation model under the action of the simulation main magnetic field, and to superimpose the dipole fields of all simulation fat droplets to obtain a non-uniform magnetic field.

[0064] The dipole field of each simulation fat droplet at different positions in the fatty liver tissue three-dimensional simulation model is obtained by formula (I):

[0065]

[0066] Where, ΔB is the non-uniform magnetic field, B0 is the strength of the simulation main magnetic field, χ L is the fat magnetic susceptibility, R is the radius of the simulation fat droplet, r is the distance from the observation point to the center of the simulation fat droplet in the fatty liver tissue three-dimensional simulation model, and θ is the included angle between the center of the simulation fat droplet and the observation point and the direction of the simulation main magnetic field.

[0067] It should be noted that the simulation fat droplet has magnetic susceptibility, and generates a dipole field under the action of the simulation main magnetic field, resulting in a non-uniform simulation main magnetic field. The size of the non-uniform magnetic field of the fatty liver tissue three-dimensional simulation model is the result of the joint action of the simulation main magnetic field and the dipole fields of all simulation fat droplets. The observation point referred to in the present application is a random point in the fatty liver tissue three-dimensional simulation model.

[0068] In step (3), the initial position of the simulated proton in the three-dimensional simulation model of the fatty liver tissue is randomly generated by computer software.

[0069] It should be noted that the computer software of the present application can be MATLAB software.

[0070] The motion of the simulated proton in the three-dimensional simulation model of the fatty liver tissue is isotropic Gaussian diffusion, and the average absolute displacement is where D is the diffusion coefficient of the simulated proton, and δ is the motion time interval of the simulated proton.

[0071] The simulated proton with an initial position outside the simulated fat droplet is defined as a simulated proton in water, and the diffusion coefficient D of the simulated proton in water is set to a μm 2 / ms, and 0.1 μm 2 / ms 2 / ms; the simulated proton with an initial position inside the simulated fat droplet is defined as a simulated proton in the simulated fat droplet, and the diffusion coefficient D of the simulated proton in the simulated fat droplet is set to 0 μm 2 / ms.

[0072] It should be noted that the initial position of the simulated proton in the present application is randomly distributed in the three-dimensional simulation model of the fatty liver tissue, and a part of the initial position of the simulated proton is located inside the simulated fat droplet, and another part of the simulated proton is located in the space between the three-dimensional simulation model of the fatty liver tissue and the outside of the simulated fat droplet, which is water. The simulated proton located between the three-dimensional simulation model of the fatty liver tissue and the outside of the simulated fat droplet can move, and the spatial position of the simulated proton in water changes when it does isotropic Gaussian diffusion, and there may be cases of crossing the boundary of the three-dimensional simulation model of the fatty liver tissue or penetrating the simulated fat droplet. The present application limits the above two cases as follows:

[0073] If the simulated proton in water contacts the boundary of the three-dimensional simulation model of the fatty liver tissue, the simulated proton in water is shot from the symmetric point of the other side boundary of the three-dimensional simulation model of the fatty liver tissue into the inside of the three-dimensional simulation model of the fatty liver tissue according to the symmetry principle;

[0074] If the simulated proton in water penetrates the simulated fat droplet, the simulated proton in water is moved to the point where the surface of the simulated fat droplet intersects with the motion path of the simulated proton in water.

[0075] wherein in step (3), the phase accumulation of each simulated proton at time t is calculated by formula (II),

[0076]

[0077] wherein γ is the magnetic spin ratio constant 2.675×10 8 rad·s-1 ·T -1 , i is the step number of the simulation proton, p(i) is the position coordinate of the simulation proton at the i-th step, and t is the time of the movement of the simulation proton.

[0078] It should be noted that each simulation proton moves to a new position after a time interval, and the phase accumulated by each simulation proton after time t can be calculated by formula (II).

[0079] Step (4) of the present application specifically superimposes the water signal of the simulation proton in water and the fat signal of the simulation proton in the simulation fat droplet in the three-dimensional simulation model of fatty liver tissue to obtain a magnetic resonance signal.

[0080] The water signal S w of the simulation proton in water is obtained from formula (III). f The fat signal S p of the simulation proton in the simulation fat droplet is obtained from formula (IV).

[0081]

[0082]

[0083] wherein S(0) is an initial signal value, R20 is an empirical value of the transverse relaxation rate of the liver, P is the number of fat spectral peaks, a p is the relative amplitude of the fat signal, f F，p is the frequency of the plurality of spectral peaks of the fat signal relative to the spectral peak of water, and p is the serial number of the spectral peak.

[0084] Because the simulation proton moves once every 0.5 μs, and the phase changes after each movement, Φ(t) represents the cumulative phase of the simulation proton at time t by adding the phase changes after each movement of the simulation proton within time t. It should be noted that P of the present application can be 1, 2, 3, 6, 8, etc., and can be determined according to actual conditions. The value of p corresponds to P, and it represents the serial number of the spectral peak. a p may be 0% to 100%, and the sum of all a p values is 100%. f F，p corresponds to the value of a p .

[0085] The magnetic resonance signal is obtained from formula (V).

[0086] S(t) = S w (t) + S f (t) formula (V).

[0087] The fat liver tissue magnetic resonance signal simulation method of the present application adopts a spin echo sequence, gives the strength B0 of the simulation main magnetic field, sets different echo times, and obtains the magnetic resonance signals at different echo times.

[0088] The step (5) of the present application comprises:

[0089] Step 5.1, curve fitting is performed on the magnetic resonance signal S(t) obtained in step (4) using a chemical shift encoding imaging signal model to obtain the prediction values of ρ w , ρ f and parameters, wherein the chemical shift encoding imaging signal model is represented by formula (VI);

[0090]

[0091] wherein ρ w is the amplitude of the water signal, ρ f is the amplitude of the fat signal, φ0 is the initial phase, f B is the frequency offset caused by the magnetic field inhomogeneity;

[0092] Step 5.2, according to ρ w , ρ f and formula (VII) obtained in step 5.1, PDFF is obtained,

[0093]

[0094] The fat liver tissue magnetic resonance signal simulation method has the beneficial effects that: for the three-dimensional simulation model of the fat liver tissue with different fat fractions, the PDFF and prediction values corresponding to the fat fractions can be obtained through simulation, so that the correction relationship between the fat fraction and the parameters can be established. Meanwhile, the influence of fat on the parameters can be corrected in different scenarios by adjusting the parameters (magnetic field strength, etc.) in the three-dimensional simulation model of the fat liver tissue. The fat liver tissue magnetic resonance signal simulation method can improve the liver iron and liver fat quantification accuracy through the correlation between the fat fraction and the parameters.

[0095] Example 2.

[0096] An application of a fat liver tissue magnetic resonance signal simulation method comprises the following steps:

[0097] Step (1), according to the liver biopsy sample and the three-dimensional morphological distribution of fat droplets, three-dimensional simulation models of fatty liver tissue with fat fractions of 3%, 5%, 7%, 10%, 15%, and 20% are obtained, and then the fat magnetic susceptibility coefficient of the three-dimensional simulation model of fatty liver tissue is set to 0.5ppm, and step (2) is entered;

[0098] Step (2), under the action of a uniform simulation main magnetic field with a simulation main magnetic field strength of 1.5T or 3.0T, the three-dimensional simulation model of fatty liver tissue generates a non-uniform magnetic field, and step (3) is entered, such as Figure 3 ;

[0099] Step (3), 10000 simulation protons are put into the three-dimensional simulation model of fatty liver tissue, and the phase accumulation of each simulation proton is calculated, and step (4) is entered;

[0100] Step (4), the water signal and fat signal of all simulation protons of the three-dimensional simulation model of fatty liver tissue are synthesized, and then the magnetic resonance signal is obtained by superposition;

[0101] Step (5), the proton density fat fraction PDFF and effective transverse relaxation rate parameter

[0102] According to the three-dimensional simulation model construction method of fatty liver tissue of the prior art, and the correlation between the simulation fat droplet size, nearest neighbor distance and regional anisotropy Gamma distribution function fitting parameters (γ and β) and the fat fraction, the fitting parameters under the corresponding fat fraction are calculated, the simulation fat droplet size, nearest neighbor distance and regional anisotropy obeying the Gamma distribution function are generated, and the distribution in the three-dimensional simulation model of fatty liver tissue is determined under the joint action of size, nearest neighbor distance and regional anisotropy. Among them, the regional anisotropy is used to determine the fat fraction in each small cube, the size is used to determine the radius of the simulation fat droplet, and the nearest neighbor distance is used to determine the distance between adjacent simulation fat droplets in the three-dimensional simulation model of fatty liver tissue. The size of the three-dimensional simulation model of fatty liver tissue is 480μm×480μm×480μm, and the fat magnetic susceptibility coefficient is set to 0.5ppm, wherein the size of the three-dimensional simulation model of fatty liver tissue is 480μm×480μm×480μm, such as Figure 2 , unit: μm.

[0103] In step (2), under the action of the simulation main magnetic field, the simulation fat droplets of the three-dimensional simulation model of fatty liver tissue generate a dipole field, and the dipole fields of all simulation fat droplets are superimposed to obtain a non-uniform magnetic field.

[0104] The dipole field of each simulation fat droplet at different positions in the three-dimensional simulation model of fatty liver tissue is obtained by formula (I):

[0105]

[0106] wherein, ΔB is the inhomogeneous magnetic field, B0 is the strength of the simulated main magnetic field, χ is the fat magnetic susceptibility coefficient, R is the radius of the simulated fat droplet, r is the distance from the observation point in the three-dimensional simulated model of the fatty liver tissue to the centroid of the simulated fat droplet, and θ is the included angle between the centroid of the simulated fat droplet and the observation point and the direction of the simulated main magnetic field. L

[0107] In step (3), the initial position of the simulated proton in the three-dimensional simulated model of the fatty liver tissue is randomly generated by MATLAB software.

[0108] The motion of the simulated proton in the three-dimensional simulated model of the fatty liver tissue is isotropic Gaussian diffusion, and the average absolute displacement is wherein D is the diffusion coefficient of the simulated proton, and δ is the motion time interval of the simulated proton, and δ is specifically 0.5 μs.

[0109] The simulated proton with the initial position outside the simulated fat droplet is defined as a simulated proton in water, and the diffusion coefficient D of the simulated proton in water is set to 0.76 μm 2 / ms; the simulated proton with the initial position inside the simulated fat droplet is defined as a simulated proton in the simulated fat droplet, and the diffusion coefficient D of the simulated proton in the simulated fat droplet is set to 0 μm 2 / ms.

[0110] When the simulated proton in water moves in the three-dimensional simulated model of the fatty liver tissue, if the simulated proton in water contacts the boundary of the three-dimensional simulated model of the fatty liver tissue, the simulated proton in water is shot into the inside of the three-dimensional simulated model of the fatty liver tissue from the symmetrical point of the other side boundary of the three-dimensional simulated model of the fatty liver tissue according to the symmetry principle.

[0111] If the simulated proton in water penetrates the simulated fat droplet, the simulated proton in water is moved to the point where the surface of the simulated fat droplet intersects with the motion path of the simulated proton in water.

[0112] wherein, in step (3), the phase accumulation of each simulated proton at time t is calculated by formula (II),

[0113]

[0114] wherein γ is the magnetic spin ratio constant 2.675×10 8 rad·s -1 ·T -1 , i is the step number of the simulated proton, p(i) is the position coordinate of the simulated proton at the i th step, and t is the motion time of the simulated proton.

[0115] ​The step (4) of the present application is specifically superimposing water signals of all simulated protons in water and fat signals of simulated protons in simulated fat droplets in the three-dimensional simulation model of fatty liver tissue to obtain a magnetic resonance signal.

[0116] The magnetic resonance signal simulation method of the fatty liver tissue of the present application adopts a spin echo sequence, sets different echo times under 1.5T magnetic field intensity or 3.0T magnetic field intensity, when the simulation main magnetic field is 1.5T, TE=1.2, 1.7, 2.2, …, 11.2ms; when the simulation main magnetic field is 3.0T, TE=1.1, 2.2, …, 6.2ms, to obtain magnetic resonance signals of different echo times.

[0117] The water signal S of the simulated protons in water is obtained from formula (III) w The fat signal S of the simulated protons in the simulated fat droplets is obtained from formula (IV) f obtained from formula (IV);

[0118]

[0119]

[0120] wherein S(0) is an initial signal value, R20 is an empirical value of a liver transverse relaxation rate (wherein 20s -1 is set as 20s -1 under 1.5T field strength and 35s p under 3.0T field strength); P is the number of fat spectral peaks and is set as 6; α F，p is the relative amplitude of the fat signal and is 4.7%, 3.9%, 0.6%, 12%, 70% and 8.8%; f w is the frequency of the multiple spectral peaks of the fat signal relative to the spectral peaks of water, which are 0.6ppm, -0.5ppm, -1.95ppm, -2.6ppm, -3.4ppm and -3.8ppm respectively.

[0121] The magnetic resonance signal is obtained from formula (V) as follows Figure 4 and 5 ;

[0122] S(t) = S w (t) + S f (t) formula (V).

[0123] The step (5) of the present application comprises:

[0124] Step 5.1, curve fitting is performed on the magnetic resonance signal S(t) obtained in step (4) using a chemical shift encoding imaging signal model to obtain ρ w , ρ f and a predicted value of a parameter, wherein the chemical shift encoded imaging signal model is represented by equation (VI);

[0125]

[0126] wherein, ρ w is the amplitude of the water signal, ρ f is the amplitude of the fat signal, φ0is the initial phase, f B is the frequency offset caused by the magnetic field inhomogeneity;

[0127] Step 5.2, obtaining PDFF from ρ w , ρ f and equation (VII) according to ρ

[0128]

[0129] The technical feasibility and technical effects of the present application are illustrated below with six different fat fractions: three-dimensional simulation models of fatty liver tissue with fat fractions of 3%, 5%, 7%, 10%, 15%, and 20% are generated, with a model size of 480 μm x 480 μm x 480 μm, a fat magnetic susceptibility of 0.5 ppm, a simulated proton diffusion coefficient in water molecules of D = 0.76 μm 2 / ms, a diffusion coefficient in fat of D = 0 μm 2 / ms, and intrinsic tissue T20(R20= 1 / T 20 ) of 50 ms and 1000 / 35 ms at 1.5 T and 3.0 T, respectively. When the simulation main magnetic field is 1.5 T, TE = 1.2, 1.7, 2.2, …, 11.2 ms; when the simulation main magnetic field is 3.0 T, TE = 1.1, 2.2, …, 6.2 ms.

[0130] For each three-dimensional simulation model of fatty liver tissue corresponding to a fat fraction, it is placed in a 1.5 T or 3.0 T uniform magnetic field, and the inhomogeneous magnetic field of the three-dimensional simulation model of fatty liver tissue is the superposition of the dipole fields generated by all the simulated fat droplets.

[0131] 10000 simulated protons are placed in the model, the initial positions of the simulated protons are randomly generated by MATLAB, the simulated protons are divided into simulated protons in fat molecules and simulated protons in water molecules by detecting the positional relationship between the simulated proton positions and the simulated fat droplet positions, and the time interval of the simulated proton movement is 0.5 μs. Using the inhomogeneous magnetic field obtained in step (2), the cumulative phase of each simulated proton at different times is calculated according to the phase calculation formula.

[0132] The simulation of this invention uses a spin echo sequence to acquire signals at different echo times. At each echo time, the cumulative phase of the simulated protons in the water calculated in step (3) is substituted into the signal synthesis formula of the simulated protons in the water, and the cumulative phase of the simulated protons in the fat calculated in step (3) is substituted into the signal synthesis formula of the simulated protons in the fat. Then, the signals in the water and the fat are superimposed to obtain magnetic resonance signals at different echo times.

[0133] The magnetic resonance signal S(t) obtained in step (4) is curve-fitted using a chemical shift encoded imaging signal model to obtain ρ. w , ρ f and The predicted value of the parameter, based on ρ w and ρ f PDFF can be calculated, such as Figure 6 and 7 In the signal model, f F，p The frequencies of multiple spectral peaks of the fat signal relative to the water spectral peaks are 0.6 ppm, -0.5 ppm, -1.95 ppm, -2.6 ppm, -3.4 ppm, and -3.8 ppm, respectively. α p The relative amplitudes of the fat signal are 4.7%, 3.9%, 0.6%, 12%, 70%, and 8.8%, respectively.

[0134] Example 3.

[0135] A method for constructing a three-dimensional simulation model of fatty liver tissue is provided for reference, including the following steps:

[0136] Step (1): Stain the liver biopsy sample;

[0137] Step (2): Based on the liver biopsy samples, statistical analysis was performed to obtain the two-dimensional size distribution histogram and the two-dimensional nearest neighbor distance distribution histogram of the real simulated fat droplets.

[0138] Step (3): Calculate the three-dimensional size distribution histogram of the simulated fat droplets based on the two-dimensional size distribution histogram of the real simulated fat droplets obtained in step (2);

[0139] Step (4): Construct grid points for the shape and scale parameters of multiple simulated fat droplets in three-dimensional nearest neighbor distance distribution, and combine them with the simulated fat droplet three-dimensional size distribution histogram obtained in step (3) to construct multiple three-dimensional simulation models of fatty liver tissue;

[0140] Step (5), respectively, to each fat liver tissue three-dimensional simulation model obtained in step (4) is cut cross section, corresponding to obtain a plurality of two-dimensional simulation cross section, and step (2) of the real simulation fat droplet two-dimensional nearest neighbor distance distribution is compared, the simulation fat droplet two-dimensional nearest neighbor distance distribution is selected, and the shape parameter and scale parameter of the simulation fat droplet three-dimensional nearest neighbor distance distribution are obtained according to the simulation fat droplet three-dimensional nearest neighbor distance, the simulation fat droplet three-dimensional nearest neighbor distance is defined as the final simulation fat droplet three-dimensional nearest neighbor distance distribution;

[0141] Step six, according to the simulation fat droplet three-dimensional size distribution obtained in step (3) and the final simulation fat droplet three-dimensional nearest neighbor distance distribution obtained in step (5), a three-dimensional simulation model of fat liver tissue is constructed.

[0142] Among them, step (1) includes:

[0143] Step 1.1, H&E staining is carried out on the liver biopsy sample;

[0144] Step 1.2, the simulation fat droplet after staining is subjected to image binarization processing.

[0145] Among them, the simulation fat droplet three-dimensional size distribution in step (3) is obtained by formula (I):

[0146]

[0147] Wherein, i and j are the number of groups of the real simulation fat droplet two-dimensional size distribution histogram in step (2), and are natural numbers; NA(i) is the frequency of the i th group in the real simulation fat droplet two-dimensional size distribution histogram, NV(j) is the frequency of the j th group in the simulation fat droplet three-dimensional size distribution histogram, and the coefficient a ij is the element of the inverse matrix of matrix [k ij ].

[0148] Wherein, [k ij ] is obtained by formula (II):

[0149]

[0150] The step (4) of the application specifically includes:

[0151] Step 4.1, the shape parameter γ of the simulation fat droplet three-dimensional nearest neighbor distance distribution is taken as the x axis, the scale parameter β of the simulation fat droplet three-dimensional nearest neighbor distance distribution is taken as the y axis, and a rectangular coordinate system is established with the (0, 0) point as the origin, and the range of γ is 0~a, the range of β is 0~b, and the x axis and the y axis are both spaced by c, a plurality of grid points are constructed, the coordinates of the grid points are (p, q), and 0

[0152] Step 4.2: Using the x-coordinate p of the grid points obtained in Step 4.1 as the shape parameter and the y-coordinate q of the grid points as the scale parameter, and based on the simulated three-dimensional size distribution histogram of fat droplets obtained in Step (3), construct multiple three-dimensional simulation models of fatty liver tissue.

[0153] Step 4.2 of the present invention specifically includes:

[0154] Step 4.2.1: Using the x-coordinate p of the grid points obtained in Step 4.1 as the shape parameter and the y-coordinate q of the grid points as the scale parameter, obtain multiple simulated three-dimensional nearest neighbor distance distribution histograms of fat droplets through the Gamma distribution function.

[0155] Step 4.2.2: Based on the three-dimensional nearest neighbor distance distribution histogram of the simulated fat droplet obtained in step 4.2.1, randomly generate a set of three-dimensional nearest neighbor distance sequences that conform to the three-dimensional nearest neighbor distance distribution histogram;

[0156] Step 4.2.3: Distribute the simulated fat droplets one by one into the simulation space according to the three-dimensional size distribution histogram of simulated fat droplets obtained in step 4.2.1. The distance between each simulated fat droplet and the previous simulated fat droplet follows the three-dimensional nearest neighbor distance sequence obtained in step 4.2.2. In the three-dimensional simulation model, the simulated fat droplets do not overlap with other simulated fat droplets, thus obtaining a three-dimensional simulation model of fatty liver tissue.

[0157] Step (5) specifically involves:

[0158] Step 5.1: Cut sections from each three-dimensional simulation model of fatty liver tissue obtained in step (4) to obtain multiple two-dimensional simulation sections;

[0159] Step 5.2: Obtain the two-dimensional nearest neighbor distance distribution histogram of the simulated fat droplet based on the cross section obtained in Step 5.1;

[0160] Step 5.3: Based on the simulated two-dimensional nearest neighbor distance distribution histogram of the fat droplet obtained in Step 5.2, obtain multiple sets of shape parameters γ using the Gamma distribution function. simu and scale parameter β simu ;

[0161] Step 5.4: Obtain the shape parameter γ from the two-dimensional nearest neighbor distance distribution histogram of the real simulated fat droplets obtained in step (2) using the Gamma distribution function. real and scale parameter β real ;

[0162] Step 5.5: Use the shape parameter γ obtained in step 5.3 for each set of shapes. simu The shape parameter γ obtained in step 5.4 real Calculate the root mean square difference rγ the corresponding scale parameter β simu the scale parameter β real the root mean square difference r β

[0163] Step 5.6, selecting the root mean square difference r γ and the root mean square difference r β the grid point of the three-dimensional simulation model of fatty liver tissue corresponding to the minimum is the target grid point;

[0164] Step 5.7, selecting the x-coordinate of the target grid point obtained in step 5.6 as the shape parameter of the three-dimensional nearest neighbor distance distribution of the simulated fat droplets, and the y-coordinate of the target grid point as the scale parameter of the three-dimensional nearest neighbor distance distribution of the simulated fat droplets, and obtaining the three-dimensional nearest neighbor distance of the simulated fat droplets according to the target grid point, and defining the three-dimensional nearest neighbor distance of the simulated fat droplets as the final three-dimensional nearest neighbor distance distribution of the simulated fat droplets.

[0165] wherein the Gamma distribution function GDF is as shown in formula (III)

[0166]

[0167] wherein x is the two-dimensional or three-dimensional nearest neighbor distance of the simulated fat droplets, γ is the shape parameter, β is the scale parameter, and Γ(γ) is the Gamma function.

[0168] The step six of the present application comprises:

[0169] Step 6.1, randomly generating a set of three-dimensional size sequences conforming to the three-dimensional size distribution histogram of the simulated fat droplets in step (3);

[0170] Step 6.2, randomly generating a set of three-dimensional nearest neighbor distance sequences conforming to the final three-dimensional nearest neighbor distance distribution of the simulated fat droplets in step (5);

[0171] Step 6.3, distributing the simulated fat droplets to the simulation space one by one according to the three-dimensional size distribution of the simulated fat droplets obtained in step 6.1, and the distance between each simulated fat droplet and the previous simulated fat droplet conforms to the three-dimensional nearest neighbor distance sequence obtained in step 6.2, and the simulated fat droplets in the three-dimensional simulation model do not overlap with other simulated fat droplets, thereby obtaining a three-dimensional simulation model of fatty liver tissue.

[0172] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit the protection scope of the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the essence and scope of the technical solutions of the present application.​

Claims

1. A method of simulating magnetic resonance signals of fatty liver tissue, characterized by, It comprises the following steps: Step (1), constructing a three-dimensional simulation model of fatty liver tissue with different fat fractions according to the three-dimensional morphological distribution of liver biopsy samples and fat droplets, and then setting the model parameters of the three-dimensional simulation model of fatty liver tissue, entering step (2); Step (2), simulating the main magnetic field acting on the three-dimensional simulation model of fatty liver tissue to generate a non-uniform magnetic field, entering step (3); Step (3), putting the simulation protons into the three-dimensional simulation model of fatty liver tissue, accumulating the phase of each simulation proton, and entering step (4); Step (4), synthesizing the water signal and fat signal of all simulation protons in the three-dimensional simulation model of fatty liver tissue, and then superimposing to obtain the magnetic resonance signal; Step (5), analyzing and calculating the magnetic resonance signal obtained in step (4) to obtain proton density fat fraction (PDFF) and effective transverse relaxation rate parameter ; The step (2) is specifically that under the action of the simulation main magnetic field, the simulation fat droplets of the three-dimensional simulation model of fatty liver tissue generate a dipole field, and the dipole fields of all simulation fat droplets are superimposed to obtain a non-uniform magnetic field; The dipole field of each simulation fat droplet at different positions in the three-dimensional simulation model of fatty liver tissue is obtained by formula (I): Formula (I); wherein, inhomogeneous magnetic field, is the strength of the simulated main magnetic field, is the fat magnetic susceptibility coefficient, R is the radius of the simulated fat droplet, r is the distance from the observation point to the centroid of the simulated fat droplet in the three-dimensional simulated model of the fatty liver tissue, and θ is the included angle between the connecting line between the centroid of the simulated fat droplet and the observation point and the direction of the simulated main magnetic field. 2.The method of simulating a magnetic resonance signal of fatty liver tissue according to claim 1, characterized in that: In the step (3), the initial position of the simulation proton in the three-dimensional simulation model of fatty liver tissue is randomly generated by computer software; The motion of the simulated protons in the three-dimensional simulated model of fatty liver tissue is isotropic Gaussian diffusion, and the average absolute displacement is where D is the diffusion coefficient of the simulated protons, and δ is the motion time interval of the simulated protons. 3.The method of simulating magnetic resonance signals of fatty liver tissue according to claim 2, characterized in that: Simulated protons having an initial position outside the simulated fat droplet are defined as water protons, the diffusion coefficient D of which is set to a = 0.1 μm 2 / ms, and 0.1 μm 2 / ms < a < 2 μm 2 / ms; simulated protons having an initial position inside the simulated fat droplet are defined as simulated protons in the simulated fat droplet, which are set to have a diffusion coefficient D of 0 pm / s 2 / ms; When the water simulation proton moves in the three-dimensional simulation model of fatty liver tissue, if the water simulation proton contacts the boundary of the three-dimensional simulation model of fatty liver tissue, the water simulation proton is made to penetrate the three-dimensional simulation model of fatty liver tissue from the other side boundary symmetry point of the three-dimensional simulation model of fatty liver tissue according to the symmetry principle; If the water simulation proton penetrates the simulation fat droplet, the water simulation proton is moved to the point where the surface of the simulation fat droplet intersects with the motion path of the water simulation proton. 4.The method of simulating magnetic resonance signals of fatty liver tissue according to claim 3, characterized in that: In the step (3), the phase accumulation of each simulation proton at time t is calculated by formula (II), Formula (II); wherein is the magnetic gyromagnetic ratio constant i is the step number of the simulated proton, is the position coordinate of the simulated proton at the i-th step, t is the time of the simulated proton motion. 5.The method of simulating magnetic resonance signals of fatty liver tissue according to claim 4, characterized in that: The step (4) is specifically that the water signal of all water simulation protons and the fat signal of simulation protons in the simulation fat droplets in the three-dimensional simulation model of fatty liver tissue are superimposed to obtain the magnetic resonance signal; water signals of simulated protons in the water fat signals of simulated protons in the simulated fat droplets from formula (IV) Formula (III); Formula (IV); wherein, is an initial signal value, is an empirical value of the transverse relaxation rate of the liver, P is the number of spectral peaks of fat spectrum, is the relative amplitude of the fat signal, is the frequency of the plurality of spectral peaks of the spectral peak of the fat signal relative to water, and p is the serial number of the spectral peak. The magnetic resonance signal is obtained by formula (V); Formula (V). 6.The method of simulating magnetic resonance signals of fatty liver tissue according to claim 5, characterized in that: The spin echo sequence is used to give a given simulation main magnetic field strength and to set different echo times to obtain magnetic resonance signals of different echo times. 7.The method of simulating magnetic resonance signals of fatty liver tissue according to claim 6, characterized in that: The step (5) comprises: Step 5.1, obtaining magnetic resonance signals from the step (4) using a chemical shift encoding imaging signal model for curve fitting to obtain and predicted values of the parameters, wherein the chemical shift encoding imaging signal model is represented by equation (VI); Formula (VI); wherein is the amplitude of the water signal, is the amplitude of the fat signal, is the initial phase, is the frequency offset caused by magnetic field inhomogeneity; Step 5.

2. PDFF was obtained from the product of Step 5.1 and formula (VII) Formula (VII). 8.The method of simulating magnetic resonance signals of fatty liver tissue according to claim 7, characterized in that: The three-dimensional simulation model of fatty liver tissue is composed of a plurality of simulation fat droplets and water; said a is 0.76 pm 2 / ms; the model parameters are at least one of a fat magnetic susceptibility, a simulated proton diffusion coefficient in the medium, TE, or a T1 relaxation time. The number of simulation protons is 5000-20000.

Images