Method for eliminating cest fat artifacts, medium, device and magnetic resonance imaging apparatus

Abstract

The application discloses a CEST fat artifact elimination method, medium, equipment and a magnetic resonance imaging device, and belongs to the technical field of magnetic resonance. The application iteratively solves a model by introducing a multi-peak model of fat into a simplified water-magnetization transfer (MT) two-pool Bloch-McConnell equation, simultaneously fits multi-peak signals of water, magnetization transfer and fat, obtains a reference signal of CEST effect without direct fat artifact, and finally eliminates the indirect influence of fat signals on CEST signals in a normalized manner. The Bloch-McConnell equation containing fat adopted by the application can more truly reflect the composition of the collected signals, thereby obtaining more accurate CEST reference signals. The application can eliminate fat artifacts in CEST signals without adjusting the acquisition sequence, and improves the accuracy of CEST imaging.

Description

Technical Field

[0001] This invention belongs to the field of magnetic resonance imaging, specifically relating to the removal of fat artifacts and the extraction of contrast in chemical exchange saturation transfer imaging. Background Technology

[0002] Chemical Exchange Saturation Transfer (CEST) imaging in magnetic resonance imaging provides molecular-level information reflecting various pathological changes. Currently, CEST has considerable application in the brain, for example, using CEST-provided protein and metabolite information for the detection and grading of gliomas. However, when applying CEST to the body, especially to fatty tissues such as the breast, it is affected by strong fat artifacts. CEST analysis typically involves subtracting the reference signal from the labeled signal, using the difference to reflect the CEST effect of the target group. Asymmetric magnetization transfer rationasymmetry (MTR) is commonly used in CEST analysis. asym The MTR analysis uses the signal at the frequency symmetrical about the water resonance frequency of the target group as a reference signal. This method assumes that the signal at the symmetrical frequency does not contain the CEST effect and that other effects are the same as the labeled signal; the difference between the two is the CEST signal. However, in tissues containing fat, the resonance frequency range of fat usually coincides with the frequency range of the reference signal, and the fat signal is not symmetrical about the water resonance frequency. This means that the MTR... asym The influence of fat could not be excluded during the analysis, resulting in fat artifacts in the results; in addition, magnetization transfer (MT) also affects MTR to some extent. asym Analysis

[0003] Current methods for eliminating fat artifacts can be broadly categorized into (1) methods based on acquisition sequence design and (2) methods based on post-processing analysis. Acquisition sequence design methods include fat suppression sequences, water excitation sequences, and the CEST-Dixon water-lipid separation method. These methods require additional sequence design, which hinders their widespread clinical application. Fat suppression sequences have limited effectiveness in suppressing fat and may leave residual fat signals. Water excitation sequences increase the energy absorbed by the human body, limiting their acquisition speed and application scope. The CEST-Dixon water-lipid separation method requires significantly more data for analysis, resulting in a substantial increase in acquisition time. Post-processing analysis methods are primarily based on fitting analysis using CEST signal normalization. The specific fitting method used is mainly multi-pool Lorentz fitting, but single-pool Lorentz difference methods and extrapolated magnetization transfer signal fitting methods can also be used for analysis. The single-pool Lorentz difference method and the extrapolation magnetization transfer signal fitting method do not incorporate the fat signal into the model, resulting in artifacts in the processing results. The multi-pool Lorentz fitting method assumes that low power saturation reaches a steady state, which may not be satisfied in practical applications. In addition, this method has many fitting parameters and requires a large amount of data for fitting, resulting in longer acquisition time and lower robustness. Summary of the Invention

[0004] The main objective of this invention is to overcome the shortcomings of existing methods and provide a more complete and robust differential analysis method based on magnetization transfer and fat signal fitting, which is based on post-processing, has a wider range of applications, and is more robust. di differential analysiswith fitted ma g net i zation t ransfer a nd l The CEST method for eliminating fat artifacts uses DIGITAL signals. This method incorporates fat signals into a model based on the Bloch-McConnell equation, obtains a reference signal through fitting, and finally normalizes it to eliminate fat artifacts.

[0005] To achieve the above objectives, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides a CEST fat artifact removal method based on magnetization transfer (MT) and fat signal fitting difference. This method removes fat artifacts and extracts signals from each voxel in the raw data acquired by CEST (Chemical Exchange Saturation Transfer). The processing steps for each voxel include:

[0007] Step 1: Perform linear interpolation correction on the main magnetic field frequency B0 field shift of the raw data collected by CEST to obtain the corrected Z spectrum corresponding to the target voxel;

[0008] Step 2: Using data containing only MT and data containing both direct water saturation and direct fat saturation effects from the corrected Z-spectrum as fitting samples, fit the model parameters in the water-MT-fat three-pool Bloch-McConnell equation model within their respective given upper and lower bounds; the water-MT-fat three-pool Bloch-McConnell equation model is obtained by incorporating the multi-peak model of fat into the water-MT two-pool Bloch-McConnell equation.

[0009] Step 3: Substitute the fitted model parameters into the water-MT-fat three-pool Bloch-McConnell equation model, and generate the CEST reference signal for each frequency point in the corrected Z-spectrum. Subtract the actual acquired signal in the corrected Z-spectrum from the reference signal corresponding to each frequency point to obtain the CEST effect value of the corresponding frequency point in the CEST effect curve.

[0010] Step 4: Normalize the CEST effect curve by dividing it by (1-FF), where FF is the fitted fat fraction; the value at the frequency of interest in the normalized CEST effect curve is the CEST effect value after eliminating fat artifact interference.

[0011] As a preferred embodiment, the water-MT-fat three-pool Bloch-McConnell equation model is as follows:

[0012]

[0013] Where exp represents an exponential function with the natural constant e as the base; Δt represents the increment at time t; Let be the total magnetization vector of the water-grease at time t+Δt, which serves as the reference signal. Let represent the magnetization vectors of the water component at time t. Magnetization vector of the fat portion, in Matrix M w =[M xw M yw M zw M zm ] T M xw M yw M zw M zmLet M represent the x-direction magnetization components of the water group, the y-direction magnetization components of the water group, the z-direction magnetization components of the water group, and the z-direction magnetization components of the MT group, respectively. f =[M xf1 M yf1 M zf1 M xf2 M yf2 M zf2 ,…,M xf7 M yf7 M zf7 ] T M xf1 M xf2 M xf7 These are the x-direction magnetization components of the 1st to 7th groups in fat, M. yf1 M yf2 M yf7 These are the y-direction magnetization components of the 1st to 7th groups in fat, M zf1 M zf2 M zf7 These are the z-direction magnetization components of the 1st to 7th fat groups; fat fraction FF = M 0f / (M 0f +M 0w ) represents the proportion of fat signal to total water and fat signal, where M 0f The magnetization components M of the first to seventh groups in fat under steady state. 0f1 M 0f2 M 0f7 The sum of M 0w The magnetization components of the water group under steady-state conditions; coefficient matrix Where the matrix Vector C w =[0,0,R 1w M 0w ,R 1m M 0m ] T ;Δω w =ω RF -ω w -ΔB0, ω RF ω represents the frequency of the applied radio frequency pulse. w Let represent the resonant frequency of the water group, ΔB0 represent the residual inhomogeneity of the main magnetic field after correction, ω1 represent the intensity of the applied radio frequency pulse, and k wm k mw R represents the exchange rate from the water group to the MT group and from the MT group to the water group, respectively. efm R represents the absorption rate of MT Group to radio frequency pulses. 1w =1 / T1w R 2w =1 / T 2w R 1m =1 / T 1m R 2m =1 / T 2m Let M represent the longitudinal relaxation rate of the water group, the transverse relaxation rate of the water group, the longitudinal relaxation rate of the MT group, and the transverse relaxation rate of the MT group, respectively. 0m The magnetization of the MT group in steady state; matrix Vector C f =[0,0,R 1f1 M 0f1 ,0,0,R 1f2 M 0f2 ,…,0,0,R 1f7 M 0f7 ] T , where matrix A f1 ~A f7 The nth matrix in n = 1, 2, ..., 7; the relative amplitudes M of the 7 fat groups 0f1 M 0f2 M 0f7 The relationship between them is estimated using three parameters: CL, ndb, and nmidb. CL represents the fatty acid chain length, ndb represents the number of double bonds per molecule, and nmidb represents the number of methyl olefin double bonds. Δω fn =ω RF -ω fn -ΔB0, ω fn R represents the resonant frequency of the nth group in fat. 1fn =1 / T 1fn R 2fn =1 / T 2fn These represent the longitudinal relaxation rate and the transverse relaxation rate of the nth fat group, respectively; the longitudinal relaxation time of all 7 fat groups is set to be the same: T. 1f =T 1f1 =T 1f2 =…=T 1f7 The lateral relaxation time for the seven fat groups was also set to the same value: T 2f =T 2f1 =T 2f2 =…=T 2f7 ;

[0014] The parameters to be fitted in the model are: T 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f,T 2f ,ΔB0,CL,ndb,nmidb.

[0015] Preferably, the longitudinal relaxation rate R of the MT 1m The resonant frequency of the fat peak is also input into the model as a priori condition, with preset values.

[0016] Preferably, the intensity ω1 of the applied radio frequency pulse is calculated from the additionally acquired radio frequency field intensity map and then used as a priori condition input into the model.

[0017] Preferably, in step 2, the data containing only MT are data points in the corrected Z-spectrum with frequencies in the range of 80 to 6 ppm; the data containing the direct saturation effect of water are data points in the corrected Z-spectrum with frequencies in the range of 1.5 to -1.5 ppm; and the data containing the direct saturation effect of fat are data points in the corrected Z-spectrum with frequencies in the range of -1 to -6 ppm.

[0018] Preferably, in step 2, the data used as fitting samples are data in the corrected Z-spectrum with frequencies in the range of 80 to 10 ppm and 0.75 to -6 ppm.

[0019] Preferably, in step 2, the parameter T to be fitted in the model is... 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f ,T 2f The upper bounds of ΔB0,CL,ndb,nmidb during fitting are [1.50, 0.100, 5e], respectively. -6 The lower bounds are [0.4500, 40, 1, 0.400, 0.150, 64, 18, 5.5, 2.5], respectively. -6 The initial values ​​for the fitted values ​​are [1.35, 0.015, 1e], 0.0001, 20, 0, 0.250, 0.025, -64, 17, 1, 0. -6 ,0.0001,20,1,0.400,0.025,0,17,1,0).

[0020] Preferably, in step 2, the absorption rate of the MT group to the radio frequency pulse is... Where g(·) is a linear function, including Gaussian, Lorentz, and super-Lorentz types; in breast data, the super-Lorentz type is preferred.

[0021] Preferably, in step 2, data in the range of 80 to 10 ppm are given a higher weight during fitting, which is twice the weight of data in the range of 0.75 to -6 ppm.

[0022] Preferably, in step 4, the frequency of interest is 3.5 ppm, and the CEST effect value at 3.5 ppm in the normalized CEST effect curve is APT. # value.

[0023] In a second aspect, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in any of the above methods.

[0024] Thirdly, the present invention provides a computer electronic device, which includes a processor and a storage medium, wherein a computer program is stored on the storage medium, and when the computer program is executed by the processor, the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference is implemented as described in any of the above methods.

[0025] Fourthly, the present invention provides a magnetic resonance imaging device for removing CEST fat artifacts, comprising a magnetic resonance scanner and a control unit. The magnetic resonance scanner is used to acquire CEST images through magnetic resonance CEST imaging. The control unit is capable of acquiring the CEST images and stores a computer program. When the computer program is executed, it is used to implement the CEST fat artifact removal method based on magnetization transfer and fat signal fitting difference as described in any of the above methods, and outputs the CEST effect value for each voxel to remove fat artifacts.

[0026] The present invention has the following beneficial effects:

[0027] This invention incorporates a multi-peak model of fat into the Bloch-McConnell equation for iterative solution, simultaneously fitting multi-peak signals from water, magnetization transfer, and fat. The Bloch-McConnell equation including fat more accurately reflects the composition of the acquired signal, thus obtaining a more accurate CEST reference signal. This eliminates fat artifacts in the CEST signal without adjusting the acquisition sequence, improving the accuracy of CEST imaging. Attached Figure Description

[0028] Figure 1 This is a schematic diagram illustrating the specific implementation process of the present invention;

[0029] Figure 2The data curves collected and fitted in the examples are shown below;

[0030] Figure 3 The images shown are the dynamic enhancement maps collected in the examples and the APT obtained by this method. # picture;

[0031] Figure 4 The fat fraction map is obtained by fitting the method proposed in this invention in the example. Detailed Implementation

[0032] The present invention will now be further described in conjunction with specific embodiments and accompanying drawings.

[0033] This invention provides a post-processing-based, widely applicable, more complete and robust differential analysis method based on magnetization transfer and fat signal fitting. di differential analysis with fitted ma g net i zation t ransfer a nd l This invention presents a method for eliminating CEST fat artifacts using DIGITAL (Digital Identifier for Specific Analytical Signals). The seven resonance peaks of fat are incorporated into a simplified water-MT two-cell Bloch-McConnell equation model. By fitting the signals of water, MT, and fat, a reference signal without direct fat artifacts is obtained. Finally, normalization is used to eliminate the indirect influence of fat signals on the CEST signal.

[0034] The core of this method lies in rationally incorporating fat into the simplified two-pool Bloch-McConnell equation model of water-MT-fat, thus establishing a three-pool Bloch-McConnell equation model of water-MT-fat. Simultaneously, this model is reasonably simplified to reduce the computational resource requirements of the method, thereby broadening its application scope. The three-pool Bloch-McConnell equation of water-MT-fat in this method will be further explained below:

[0035] The simplified Bloch-McConnell equation for the water-MT two-pool (water pool w, MT pool w) commonly used in the field can be expressed as:

[0036]

[0037]

[0038]

[0039]

[0040] Where M xw M yw m zw M zm These represent the x-direction magnetization components of the water group, the y-direction magnetization components of the water group, the z-direction magnetization components of the water group, and the z-direction magnetization components of the MT group, respectively, Δω. w =ω RF -ω w ω RF ω represents the frequency of the applied radio frequency pulse. w Let ω1(t) represent the resonant frequency of the water group, and k represent the intensity of the radio frequency pulse applied at time t. wm k mw R represents the exchange rate from the water group to the MT group and from the MT group to the water group, respectively. rfm R represents the absorption rate of MT Group to radio frequency pulses. 1w =1 / T 1w R 2w =1 / T 2w R 1w =1 / T 1m R 2m =1 / T 2m Let M represent the longitudinal relaxation rate of the water group, the transverse relaxation rate of the water group, the longitudinal relaxation rate of the MT group, and the transverse relaxation rate of the MT group, respectively. 0w M 0m These represent the magnetization of the water group and the magnetization of the MT group under steady-state conditions, respectively.

[0041] MT Group's absorption rate R of radio frequency pulses rfb It can be represented as Where g(·) is the linear function, including Gaussian, Lorentz, and super-Lorentz types. Currently, it is generally believed in the field that the super-Lorentz linear function is closer to the MT signal in the human body.

[0042] For super-Lorentzian linear functions:

[0043]

[0044] Where Δω m =ω RF -ω m ω m This indicates the resonant frequency of the water group.

[0045] The simplified water-MT two-cell Bloch-McConnell equation described above can only describe the exchange process between water and magnetization transfer signal. When a fat signal is present, the CEST data curve will contain a fat resonance peak. When fitting the two cells, the fat signal may be present in the fitted sample data, resulting in a deviation in the fitting results. In addition, the preprocessing of CEST data is also affected by fat, and the CEST signal is distorted by the modulation of fat content.

[0046] For a single resonance peak signal of fat, the Bloch equation can be used to describe it:

[0047]

[0048]

[0049]

[0050] Where M xf M yf M zf These are the magnetization components in the x, y, and z directions, respectively, Δω f =ω RF -ω f ω RF ω represents the frequency of the applied radio frequency pulse. f R represents the resonant frequency of the resonance peak, ω1(t) represents the intensity of the radio frequency pulse applied at time t, and R 1f =1 / T 1f R 2f =1 / T 2f M represents the longitudinal relaxation rate and the lateral relaxation rate, respectively. 0f The magnetization intensity is in steady state.

[0051] The Bloch-McConnell equations for a two-pool water-MT system can be written in matrix form as follows:

[0052]

[0053] Where M w =[M xw M yw M zw M zm ] T ,

[0054]

[0055] C w =[0,0,R 1w M 0w ,R 1m M 0m ]T

[0056] Further simplification yields:

[0057]

[0058] in,

[0059] The solution to the above equation is:

[0060]

[0061] in Let be the magnetization vector of the Bloch-McConnell equation model for the water and MT pools at time (t+Δt).

[0062] Similarly, for a single peak in fat, the Bloch equation, after simplification and solution using the above method, yields the following solution:

[0063]

[0064] in Let be the magnetization vector of a single resonance peak of fat at time (t+Δt). The superscript "1" in the formula represents a fat signal with one resonance peak.

[0065] Since there are no chemical exchanges or other interactions between the various resonance peaks of fat cells, summing the signals of multiple fat peaks according to their amplitude proportions yields a multi-peak signal of fat cells. This summation can be accomplished through matrix operations, letting...

[0066] in Vector C f =[0,0,R 1f1 M 0f1 ,0,0,R 1f2 M 0f2 ,…,0,0,R 1f7 M 0f7 ] T , where matrix A f1 ~A f7 The nth matrix in The relative amplitude M of the seven fat groups 0f1 M 0f2 M 0f7The relationships between these factors may differ within an organization, and can be estimated using three parameters: fatty acid chain length (CL), the number of double bonds per molecule (ndb), and the number of methyl olefin double bonds (nmidb). Specifically, the number of hydrogen protons in each of the seven groups within a fatty acid molecule can be expressed using these three parameters as follows:

[0067] Group #1 contains 9 hydrogen protons (M1 = 9).

[0068] The number of hydrogen protons in Group #2 is M2 = ((CL-4)×6)-(ndb×8)+(nmidb×2)

[0069] Group #3 contains 6 hydrogen protons (M3 = 6).

[0070] The number of hydrogen protons in Group #4 is M4 = 6 + (ndb - nmidb) × 4

[0071] The number of hydrogen protons in Group #5 is M5 = nmidb × 2

[0072] Group #6 contains 4 hydrogen protons (M6 = 4).

[0073] The number of hydrogen protons in Group #7 is M7 = ndb × 2 + 1

[0074] The relationship between the relative amplitudes of the seven fat groups mentioned above is determined by the chemical composition of fat. The number of hydrogen protons M in any nth group is... n Dividing by the total number of hydrogen protons in a fat molecule yields the relative amplitudes between the groups, n = 1, 2, ..., 7.

[0075] Δω fn =ω RF -ω fn -ΔB0, ω fn R represents the resonant frequency of the nth group in fat. 1fn =1 / T 1fn R 2fn =1 / T 2fn Let represent the longitudinal relaxation rate of the nth fat group and the transverse relaxation rate of the nth fat group, respectively.

[0076] Where M f =[M xf1 M yf1 M zf1 M xf2 M yf2 M zf2 ,…,M xf7 M yf7 M zf7 ] T Mxf1 M xf2 M xf7 These are the x-direction magnetization components of the 1st to 7th groups in fat, M. yf1 M yf2 M yf7 These are the y-direction magnetization components of the 1st to 7th groups in fat, M zf1 M zf2 M zf7 These represent the z-direction magnetization components of the 1st to 7th groups in the fat.

[0077] Based on the above changes, the sum of the multi-peak signals from fat can be expressed as:

[0078]

[0079] Using formulas (1) and (3), the values ​​of water and fat at a specific saturation pulse frequency ω can be calculated iteratively. RF The signal M after saturation w (ω RF M f (ω RF Since there is no chemical exchange between water and lipids, the final acquired signal can be simply represented as the sum of the water and lipid signals, i.e., M. acq (ω RF ) = M w (ω RF )+M f (ω RF ).

[0080] In CEST data processing, the acquired data needs to be divided by the unsaturated image data to calculate the Z-spectrum, and the fitted signal divided by the unsaturated data yields the reference Z-spectrum M. ref :

[0081]

[0082] Let the fat fraction FF be the proportion of unsaturated fat to the unsaturated water-lipid signal, FF = M 0f / (M 0f +M 0w The above formula can be simplified to:

[0083]

[0084] M in the above formula 0w With M 0f This can be eliminated by setting the steady-state value during fitting to 1. It's important to note that the steady-state value during fitting does not need to be equal to the steady-state value in the fat fraction. Thus, the above formula can be further simplified to:

[0085] M ref (ω RF ) = M w (ω RF )·(1-FF)+M f (ω RF )·FF#(4)

[0086] This is equivalent to directly fitting the percentage of water and lipid signals that are unsaturated after saturation, i.e., directly fitting the Z-spectrum, and then adding them together according to the ratio of water and lipid unsaturated signals.

[0087] From formulas (1), (3), and (4), the formula for the change of the reference signal with time can be obtained as follows:

[0088]

[0089] Consider a specific frequency of interest ω i The CEST effect at the reference signal. The CEST effect value CEST(ω) is calculated directly from the reference signal. i This can be represented as:

[0090] CEST(ω i ) = M ref (ω i )-M acq (ω i )

[0091] Where M acq (ω i ) is the saturation frequency ω i The data from the Z-spectrum. The CEST effect is caused by the exchange of solute molecules with water, so the signal in the fatty portion of the reference signal and the acquired signal are the same. Therefore:

[0092] CEST(ω i )=(M wref -M wacq )·(1-FF)

[0093] Where M wref M wacq These represent the water components in the reference and acquired data, respectively. According to the above formula, the directly calculated CEST effect is affected by (1-FF). Dividing both sides of the above formula by (1-FF) eliminates the scaling effect of fat fraction on the CEST effect. The fat fraction-corrected CEST effect is denoted as CEST′(ω). i ), which may include:

[0094]

[0095] Formula (5) is the model used in this invention, and Formula (6) is the algorithm for calculating the CEST effect using this model.

[0096] Since the relaxation rates of fat cells are relatively similar, we assume T 1f =T 1f1 =T 1f2 =…=T 1f7 T 2f =T 2f1 =T 2f2 =…=T 2f7 To reduce the computational burden of fitting. The longitudinal relaxation rate R of MT 1m The resonance frequency of the fat peak, based on data from the literature, was input into the model as a priori condition. Ultimately, the model of this invention contains 12 unknowns, T. 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f ,T 2f ΔB0,CL,ndb,nmidb, and all other parameters are known. By fitting the model, a reference signal that includes both fat and MT effects can be obtained, eliminating the direct influence of fat on the CEST effect. Then, the CEST effect is renormalized according to the CEST effect calculation formula to eliminate the scaling of the CEST effect by fat, thus obtaining the CEST effect value that excludes fat artifacts.

[0097] Therefore, based on the above theoretical discussion, the specific implementation process of the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference in this invention can be described as follows:

[0098] Fat artifact removal and signal extraction were performed on each voxel in the raw data acquired by CEST. The processing steps for each voxel included:

[0099] Step 1: Perform linear interpolation correction on the raw data acquired by CEST to offset the main magnetic field frequency B0 field, and obtain the corrected Z spectrum corresponding to the target voxel.

[0100] Step 2: Using data containing only MT and data containing both direct water saturation and direct fat saturation effects from the corrected Z-spectrum as fitting samples, fit the model parameters of the water-MT-fat three-pool Bloch-McConnell equation model within their respective given upper and lower bounds. The above water-MT-fat three-pool Bloch-McConnell equation model is as described in formula (5), and will not be repeated here.

[0101] It should be noted that the specific frequency ranges selected for extracting data points from the corrected Z-spectrum as fitting samples depend on the frequency ranges of the three effects. Generally, data containing only MT are data points in the corrected Z-spectrum with frequencies in the range of 80–6 ppm; data containing the direct water saturation effect are data points in the corrected Z-spectrum with frequencies in the range of 1.5–-1.5 ppm; and data containing the direct fat saturation effect are data points in the corrected Z-spectrum with frequencies in the range of -1–-6 ppm. Theoretically, data points extracted from these three frequency ranges in the corrected Z-spectrum can be used as fitting samples. However, considering the fitting effect, the frequencies of these data points should ideally not fall around the final frequency of interest. For example, if the frequency of interest is selected as 3.5 ppm, data points near 3.5 ppm should ideally not be used as fitting samples. Therefore, it is preferable to set the data used as fitting samples to have frequencies in the corrected Z-spectrum in the ranges of 80–10 ppm and 0.75–-6 ppm.

[0102] Step 3: Substitute the fitted model parameters into the water-MT-fat three-pool Bloch-McConnell equation model, and generate the CEST reference signal for each frequency point in the corrected Z-spectrum. Subtract the actual acquired signal in the corrected Z-spectrum from the reference signal corresponding to each frequency point to obtain the CEST effect value of the corresponding frequency point in the CEST effect curve.

[0103] Step 4: Normalize the CEST effect curve by dividing it by (1-FF), where FF is the fitted fat fraction; the value at the frequency of interest in the normalized CEST effect curve is the CEST effect value after eliminating fat artifact interference.

[0104] It should be noted that the frequency of interest can be chosen arbitrarily, but in this field, it is typically selected as the resonance frequency of the group containing the CEST effect to reflect the signal strength of that group at that frequency. For example, when the frequency of interest is selected as 3.5 ppm, the CEST effect value calculated using this method is APT. # (Amideproton transfer signal from DIGITAL analysis, obtained by the CEST fat artifact removal method based on the difference between magnetization transfer and fat signal fitting) value. Other frequencies that can reflect the CEST effect can also be selected as frequencies of interest; this method does not impose restrictions on the selection of frequencies of interest.

[0105] The method proposed in this invention will be further illustrated below through the following non-limiting examples:

[0106] Example

[0107] In this embodiment, the specific implementation process of the DIGITAL method is as follows: Figure 1 As shown. The method proposed in this invention requires first acquiring CEST image data. For each voxel in the acquired CEST image data, the DIGITAL method is performed voxel by voxel, that is, steps 1 to 4 are performed sequentially to calculate the APT of each voxel after eliminating fat artifacts. # The specific implementation process of steps 1 to 4 will be described in detail below.

[0108] Step 1: Perform linear interpolation correction on the original data collected by CEST to offset the main magnetic field frequency B0 field, and obtain the corrected Z spectrum.

[0109] Step 2: Data containing only MT in the Z-spectrum, as well as data containing both water and fat direct saturation effects, are used as fitting samples. The pre-defined water-MT-fat three-pool Bloch-McConnell equation model is fitted according to given upper and lower bounds. The fitted model is used to calculate the reference signal for the Z-spectrum. As mentioned above, the water-MT-fat three-pool Bloch-McConnell equation model used in this method is as follows:

[0110]

[0111] Where exp represents an exponential function with the natural constant e as the base; Δt represents the increment at time t; Let be the total magnetization vector of the water-grease at time t+Δt, which serves as the reference signal. Let represent the magnetization vectors of the water component at time t. magnetization vector of the fat portion in Matrix M w =[M xw M yw M zw M zm ] T M xw M yw M zw M zm These represent the x-direction magnetization component of the water group, the y-direction magnetization component of the water group, the z-direction magnetization component of the water group, and the z-direction magnetization component of the MT group, respectively. f =[M xf1 M yf1 M zf1 M xf2 M yf2 M zf2 ,…,Mxf7 M yf7 M zf7 ] T M xf1 M xf2 M xf7 These are the x-direction magnetization components of the 1st to 7th groups in fat, M. yf1 M yf2 M yf7 These are the y-direction magnetization components of the 1st to 7th groups in fat, M zf1 M zf2 M zf7 These are the z-direction magnetization components of the 1st to 7th fat groups; fat fraction FF = M 0f / (M 0f +M 0w ) represents the proportion of fat signal to total water and fat signal, where M 0f The magnetization components M of the first to seventh groups in fat under steady state. 0f1 M 0f2 M 0f7 The sum of M 0w The magnetization components of the water group under steady-state conditions; coefficient matrix Where the matrix Vector C w =[0,0,R 1w M 0w ,R 1m M 0m ] T Δω w =ω RF -ω w -ΔB0, ω RF ω represents the frequency of the applied radio frequency pulse. w Let represent the resonant frequency of the water group, ΔB0 represent the residual inhomogeneity of the main magnetic field after correction, ω1 represent the intensity of the applied radio frequency pulse, and k wm k mw This indicates the exchange rate from the water group to the MT group, and from the MT group back to the water group. Where g(·) represents the super-Lorentz type, indicating the absorption rate of the MT group for radio frequency pulses, R 1w =1 / T 1w R 2w =1 / T 2w R 1m =1 / T 1m R 2m =1 / T 2m Let M represent the longitudinal relaxation rate of the water group, the transverse relaxation rate of the water group, the longitudinal relaxation rate of the MT group, and the transverse relaxation rate of the MT group, respectively. 0mThe matrix represents the magnetization of the MT group in steady state. Vector C f =[0,0,R 1f1 M 0f1 ,0,0,R 1f2 M 0f2 ,…,0,0,R 1f7 M 0f7 ] T , where matrix A f1 ~A f7 The nth matrix in The relative amplitude M of the seven fat groups 0f1 M 0f2 M 0f7 The composition of fatty acids may vary within a single organization, and the relationships between them can be estimated using three parameters: the fatty acid chain length (CL), the number of double bonds per molecule (ndb), and the number of methyl olefin double bonds (nmidb). Specifically, the number of hydrogen protons in each of the seven groups within a fatty acid molecule can be expressed using these three parameters as follows:

[0112] Group #1 contains 9 hydrogen protons (M1 = 9).

[0113] The number of hydrogen protons in Group #2 is M2 = ((CL-4)×6)-(ndb×8)+(nmidb×2)

[0114] Group #3 contains 6 hydrogen protons (M3 = 6).

[0115] The number of hydrogen protons in Group #4 is M4 = 6 + (ndb - nmidb) × 4

[0116] The number of hydrogen protons in Group #5 is M5 = nmidb × 2

[0117] Group #6 contains 4 hydrogen protons (M6 = 4).

[0118] The number of hydrogen protons in Group #7 is M7 = ndb × 2 + 1

[0119] The relationship between the relative amplitudes of the seven fat groups mentioned above is determined by the chemical composition of fat. The number of hydrogen protons M in any nth group is... n Dividing by the total number of hydrogen protons in a fat molecule yields the relative amplitudes between groups, n = 1, 2, ..., 7. Δω fn =ω RF -ω fn -ΔB0, ω fn R represents the resonant frequency of the nth group in fat. 1fn =1 / T 1fn R 2fn =1 / T2fn Let represent the longitudinal relaxation rate of the nth fat group and the transverse relaxation rate of the nth fat group, respectively.

[0120] In the above model, because fat contains seven resonance peaks, there are too many unknown parameters. Therefore, we simplified the seven-peak model of fat. The main differences between the seven peaks of fat lie in their resonance frequencies and relative amplitudes; their relaxation rates do not differ significantly. Therefore, we assume that the relaxation rates of different resonance peaks of fat are the same in the model, i.e., T0. 1f =T 1f1 =T 1f2 =…=T 1f7 T 2f =T 2f1 =T 2f2 =…=T 2f7 In addition, this method requires the acquisition of radio frequency field intensity maps to calculate the actual radio frequency pulse intensity ω1 received by each voxel.

[0121] After the above simplification, the model used in this method contains 12 undetermined parameters, namely T 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f ,T 2f ΔB0,CL,ndb,nmidb. These undetermined parameters are fitted using data from the Z-spectrum containing only MT, as well as data containing both direct water saturation and direct fat saturation effects. In this example, the selected data are those with frequencies in the ranges of 80–10 ppm and 0.75–-6 ppm. In this example, it is preferable to set the weight of the data in the 80–10 ppm range to twice the weight of the data in the 0.75–-6 ppm range to balance the difference in the number of data in the two ranges.

[0122] In this embodiment, there are 12 undetermined parameters T 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f ,T 2fΔB0,CL,ndb,nmidb can be directly estimated from the fitted samples to obtain the fitted value that minimizes the total fitting error in the model used by this method within the upper and lower bounds. The total fitting error can be represented by the minimum mean square error, and the weights can be set by changing the coefficients of the error when calculating the minimum mean square error. Specific fitting methods can be implemented using software such as MATLAB and Python or other existing technologies.

[0123] During fitting, the undetermined parameter T 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f ,T 2f The upper and lower bounds of ΔB0,CL,ndb,nmidb can be adjusted based on the fitted data. When fitting different human tissues or water phantoms, the parameters to be determined need to be adjusted according to the normal parameter range of the tissue or water phantom. In this example, breast data is fitted, and the parameter T... 1w ,T 2w ,T 2m M 0m ,k wm ,FF,T 1f ,T 2f The upper bounds of ΔB0,CL,ndb,nmidb are set to [1.50, 0.100, 5e] respectively. -6 The lower bounds are set to [0.4500, 40, 1, 0.400, 0.150, 64, 18, 5.5, 2.5], respectively. -6 The initial values ​​for the fitted values ​​are [1.35, 0.015, 1e], 0.0001, 20, 0, 0.250, 0.025, -64, 17, 1, 0. -6 ,0.0001,20,1,0.400,0.025,0,17,1,0]

[0124] Step 3: Substitute the parameters obtained from the fitting process above into the model proposed in this method to calculate the reference curve that includes the MT effect, the direct water saturation effect, and the direct fat saturation effect. The horizontal axis of this curve represents the input fitting frequency ω, and the vertical axis represents the reference signal. The longitudinal magnetization vector. The CEST effect curve is obtained by subtracting the Z-spectrum curve (corrected by the main magnetic field B0) point by point from this reference curve. It is important to note that since the corrected Z-spectrum curve is actually composed of discrete data points, when calculating the CEST effect curve, the corresponding reference signal should be calculated based on the frequency of each data point in the Z-spectrum curve, substituted into the model with the fitting parameters. Then, the original value of the corresponding frequency on the Z-spectrum curve is subtracted from the reference signal to obtain the CEST effect value of the corresponding frequency in the CEST effect curve.

[0125] Step 4: Divide the CEST effect curve calculated in Step 3 by (1-FF) obtained from the voxel fitting and renormalize to eliminate the scaling of the CEST effect by the fat signal, obtaining the corrected CEST effect curve. The data at a frequency of 3.5 ppm in this curve is the APT. # Effect size.

[0126] Of course, when calculating the reference curve, it is not necessary to calculate the complete reference curve. It is only necessary to calculate the value at 3.5 ppm. Comparing this value with the value at 3.5 ppm of the data curve after correction of the main magnetic field B0 will give the CEST effect value at 3.5 ppm. Dividing this value by (1-FF) will give the corrected CEST effect value, which can then be used as the APT value. # Effect size.

[0127] After performing steps 1 to 4 above on all voxels in the CEST plot, each voxel corresponds to an APT. # Values, these APTs # By matching each value with a voxel in the CEST plot, an APT plot can be obtained. # picture.

[0128] The aforementioned DIGITAL method was tested on a breast cancer patient to further demonstrate its technical effectiveness. In this embodiment, the imaging sequence for acquiring CEST data included a CEST saturation module and a fast spin echo acquisition module. The CEST saturation module contained 10 Gaussian saturation pulses with a duration of 100 ms, an effective amplitude of 1 μT, and an interval of 10 ms between the saturation pulses. It is important to note that this method does not require fat suppression during CEST imaging, therefore a fat suppression module is not needed during CEST acquisition. Additionally, extra main magnetic field B0 and radio frequency field B1 maps were acquired for B0 correction and parameter fitting.

[0129] like Figure 1 As shown, the Z-spectrum data acquired from CEST are corrected for main magnetic field B0 shift based on the acquired main magnetic field B0 map. The acquired Z-spectrum is as follows: Figure 2The experimental data are shown below. Then, DIGITAL fitting was performed on all voxels. The fitting samples were data with frequencies in the range of 80–10 ppm and 0.75–-6 ppm. The acquired RF field B1 plot was also input into the fitting process to calculate the RF ω1. The upper and lower bounds and initial values ​​of the parameters are the same as in step 2 above. The samples used for fitting and the reference curve obtained from the fitting are shown below. Figure 2 As shown.

[0130] The experimental data obtained in this example are as follows: Figure 3 As shown in the figure, APT # In dynamically enhanced images, there are corresponding high-signal regions within the high-signal regions. Figure 3 (As indicated by the black arrow in the middle), the color bar on the right only indicates APT. # Figure 1. The fat fraction graph obtained by fitting in this experiment is shown below. Figure 4 As shown, a significant change in fat fraction can be observed. (Comparison) Figure 3 APT in # The diagram shows the APT (Advanced Perceptions Map) of glandular regions with low and high fat fractions. # The effect size showed no significant difference, suggesting that the fat artifact had been eliminated.

[0131] Similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer electronic device corresponding to the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference provided in the above embodiment, which includes a memory and a processor.

[0132] The memory is used to store computer programs;

[0133] The processor is configured to implement the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described above when executing the computer program.

[0134] Similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a computer-readable storage medium corresponding to the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference provided in the above embodiments. The storage medium stores a computer program, which, when executed by a processor, implements the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described above.

[0135] It is understood that the aforementioned storage media and memory may include Random Access Memory (RAM) or Non-Volatile Memory (NVM), such as at least one disk storage device. Furthermore, the storage media can also be various media capable of storing program code, such as USB flash drives, external hard drives, magnetic disks, or optical discs. Of course, with the widespread application of cloud servers, the aforementioned software programs can also be hosted on cloud platforms to provide corresponding services; therefore, computer-readable storage media are not limited to the form of local hardware.

[0136] It is understood that the processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0137] It should also be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. In the embodiments provided in this application, the division of steps or modules in the device and method is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules or steps may be combined or integrated together, and a module or step may also be split.

[0138] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0139] Similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a magnetic resonance imaging device corresponding to the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference provided in the above embodiment, which includes a magnetic resonance scanner and a control unit.

[0140] The magnetic resonance scanner is used to acquire CEST images through magnetic resonance CEST imaging;

[0141] The control unit can acquire the CEST image and stores a computer program. When the computer program is executed, it is used to implement the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described above, and outputs the CEST effect value of each voxel for eliminating fat artifacts.

[0142] It should be noted that the magnetic resonance imaging (MRI) device can be any MRI scanner capable of implementing parallel imaging methods. Its structure is existing technology, and mature commercial products can be used; the specific model is not limited. Furthermore, in addition to storing the aforementioned computer programs, the control unit of the MRI device should also contain the imaging sequences and other software programs necessary for implementing CEST imaging.

[0143] The modules and functions involved in the above embodiments of the present invention can be implemented by circuits, other hardware, or executable program code, as long as the corresponding functions can be achieved. If code is used, the code can be stored in a storage device and executed by corresponding elements in a computing device. The implementation of the present invention is not limited to any specific hardware and software combination. All hardware models in the present invention can be commercially available products, and can be selected according to actual user needs. Of course, the magnetic resonance CEST imaging sequence and device also require other necessary hardware or software, which will not be elaborated here.

[0144] The embodiments described above are merely implementations of this application and are not intended to limit the invention. Any changes and improvements made to the invention without departing from its spirit and scope, or direct or indirect applications in other related technical fields, are included in this invention and are protected by the appended claims.

Claims

1. A CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference, characterized in that, Fat artifact removal and signal extraction were performed on each voxel in the raw data acquired by CEST. The processing steps for each voxel included: Step 1: Perform linear interpolation correction on the main magnetic field frequency B0 field shift of the raw data collected by CEST to obtain the corrected Z spectrum corresponding to the target voxel; Step 2: Using data containing only MT and data containing both direct water saturation and direct fat saturation effects from the corrected Z-spectrum as fitting samples, fit the model parameters in the water-MT-fat three-pool Bloch-McConnell equation model within their respective given upper and lower bounds; the water-MT-fat three-pool Bloch-McConnell equation model is obtained by incorporating the multi-peak model of fat into the water-MT two-pool Bloch-McConnell equation. Step 3: Substitute the fitted model parameters into the water-MT-fat three-pool Bloch-McConnell equation model, and generate the CEST reference signal for each frequency point in the corrected Z-spectrum. Subtract the actual acquired signal in the corrected Z-spectrum from the reference signal corresponding to each frequency point to obtain the CEST effect value of the corresponding frequency point in the CEST effect curve. Step 4: Divide the CEST effect curve by Normalization is performed, where The obtained fat fraction is used for fitting; in the normalized CEST effect curve, the value at the frequency of interest is the CEST effect value after eliminating fat artifact interference.

2. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 1, characterized in that, The water-MT-fat three-pool Bloch-McConnell equation model is as follows: in Represented by natural constant An exponential function with base 0; Indicates time The increment; for The total magnetization vector of the lipid at time t is used as a reference signal. , They represent magnetization vector of the water part at time Magnetization vector of the fat portion ;in , ,matrix , , , , They are respectively water groups Directional magnetization components, water group Directional magnetization components, water group Directional magnetization components, MT group Directional magnetization components, matrix , , … These represent the magnetization components along the x-axis of the 1st to 7th groups in fat. , … These represent the y-direction magnetization components of the 1st to 7th groups in the fat. , … These represent the z-direction magnetization components of the 1st to 7th groups in the fat; Fat Score This indicates the proportion of fat signal to total water and fat signal, among which The magnetization components of the first to seventh groups in fat under steady state. , … The sum of The magnetization components of the water group under steady-state conditions; coefficient matrix , , where the matrix ,vector ; , Indicates the frequency of the applied radio frequency pulse. Indicates the resonant frequency of the water group. This indicates the residual inhomogeneity of the main magnetic field after correction. Indicates the intensity of the applied radio frequency pulse. , These represent the exchange rates from the water group to the MT group and from the MT group to the water group, respectively. This indicates the absorption rate of MT Group for radio frequency pulses. , , , Let represent the longitudinal relaxation rate of the water group, the lateral relaxation rate of the water group, the longitudinal relaxation rate of the MT group, and the lateral relaxation rate of the MT group, respectively. The magnetization of the MT group in steady state; matrix ,vector , where the matrix ~ The nth matrix in , n =1,2,…,7; Relative amplitudes of the seven fat groups , … The relationship between them is estimated using three parameters: CL, ndb, and nmidb. Parameter CL represents the fatty acid chain length, parameter ndb represents the number of double bonds per molecule, and parameter nmidb represents the number of methyl olefin double bonds. , Indicates the first of the fats The resonant frequency of the group , These represent the first and second parts of fat. The longitudinal relaxation rate of each group, the first in fat The lateral relaxation rate of each group; the longitudinal relaxation time of the seven fat groups is set to be the same: The lateral relaxation time for the seven fat groups was also set to be the same: ; The parameters to be fitted in the model are: .

3. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 2, characterized in that, The longitudinal relaxation rate of the MT The resonant frequency of the fat peak is also input into the model as a priori condition, with preset values.

4. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 2, characterized in that, The intensity of the applied radio frequency pulse The data is calculated from additionally acquired radio frequency field intensity maps and then used as prior conditions input into the model.

5. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 1, characterized in that, In step 2, data containing only MT are data points in the corrected Z-spectrum with frequencies in the range of 80 to 6 ppm; data containing the direct saturation effect of water are data points in the corrected Z-spectrum with frequencies in the range of 1.5 to -1.5 ppm; and data containing the direct saturation effect of fat are data points in the corrected Z-spectrum with frequencies in the range of -1 to -6 ppm.

6. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 5, characterized in that, The data used as fitting samples are those with frequencies in the corrected Z-spectrum ranging from 80 to 10 ppm and from 0.75 to -6 ppm.

7. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 2, characterized in that, The parameters to be fitted in the model in step 2 The upper bounds for fitting are [1.50, 0.100, 5e]. -6 The lower bounds are [0.4500, 40, 1, 0.400, 0.150, 64, 18, 5.5, 2.5], respectively. -6 The initial values ​​for the fitted values ​​are [1.35, 0.0001, 20, 0, 0.250, 0.025, -64, 17, 1, 0]. -6 , 0.0001, 20, 1, 0.400, 0.025, 0, 17, 1, 0).

8. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 2, characterized in that, In step 2, data in the range of 80 to 10 ppm are given a higher weight during fitting, which is twice the weight of data in the range of 0.75 to -6 ppm.

9. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 2, characterized in that, In step 2, the absorption rate of the MT group to the radio frequency pulse ;in These are linear functions, including Gaussian, Lorentz, and super-Lorentz types.

10. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 8, characterized in that, In breast data, the linear function is of the super-Lorentz type.

11. The CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in claim 1, characterized in that, In step 4, the frequency of interest is 3.5 ppm, and the CEST effect value at 3.5 ppm in the normalized CEST effect curve is APT. # value.

12. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in any one of claims 1 to 11.

13. A computer electronic device, characterized in that, The device includes a processor and a storage medium, wherein a computer program is stored on the storage medium, and when the computer program is executed by the processor, it implements the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in any one of claims 1 to 11.

14. A magnetic resonance imaging device for removing CEST fat artifacts, characterized in that, The device includes a magnetic resonance scanner and a control unit. The magnetic resonance scanner is used to acquire CEST images through magnetic resonance CEST imaging. The control unit is capable of acquiring the CEST images and stores a computer program. When the computer program is executed, it is used to implement the CEST fat artifact elimination method based on magnetization transfer and fat signal fitting difference as described in any one of claims 1 to 11, and outputs the CEST effect value of fat artifact elimination for each voxel.

Images