A magnetic resonance fingerprint imaging method for multi-core

Abstract

The present invention discloses a magnetic resonance fingerprint imaging method for multi-nuclear. The magnetic resonance fingerprint imaging pulse sequence designed by the present invention comprises a pulse sequence set with a set number of cycles, each pulse sequence set comprises a T2 preparation pulse and a fast small-angle excitation sampling pulse set, and the fast small-angle excitation sampling pulse set comprises a set number of fast small-angle excitation sampling pulses with different flip angles FA, different repetition times TR, and different echo times TE; the T2 preparation pulse adopts an adiabatic pulse, which can be used for the quantitative measurement of relaxation parameters of multi-nuclear in a hyperpolarized state; the weight of T2 can be modulated by an adiabatic pulse, which is suitable for the precise quantitative measurement of short T2 relaxation time of multi-nuclear; T1, T2, and T2 can also be modulated by an adiabatic pulse. * Multiple relaxation parameters can be quantified simultaneously, and the quantification speed is fast.

Description

Technical Field

[0001] The present invention belongs to the technical field of multi-nuclear magnetic resonance imaging, and specifically relates to a multi-nuclear magnetic resonance fingerprint imaging method suitable for completing T1, T2, and T2 * Quantification of three relaxation parameters. Background Art

[0002] Magnetic resonance imaging can obtain high-resolution and high-contrast imaging information of tissues in a non-invasive and non-ionizing radiation manner. Magnetic resonance signals originate from atomic nuclei with non-zero spins. Currently, conventional magnetic resonance imaging methods are based on protons (spin 1 / 2) for imaging, and mainly use protons in water molecules to obtain structural and functional information of biological tissues. Proton magnetic resonance imaging includes T1-weighted, T2-weighted, proton density-weighted, diffusion imaging, and brain functional imaging.

[0003] However, proton MRI has difficulty in obtaining information about the internal microenvironment and functional metabolism of tissues; in addition, the low proton density in tissues such as the lungs makes it difficult to obtain information about the structure and function of the lungs. In vivo MRI studies have abundant multi-nuclear resources, such as 23 Na, 31 P, 13 C. 129 Xe, etc., these nuclides can all generate magnetic resonance signals, providing important information about life activities. The magnetic resonance method based on non-proton nuclei is called multi-nuclear magnetic resonance method.

[0004] The development of multi-nuclear magnetic resonance is limited by sensitivity and magnetic resonance imaging methods. Sensitivity can be improved by hyperpolarization methods. Hyperpolarization technology based on spin exchange optical pumping (SEOP) can greatly improve 129 Xe magnetic resonance detection sensitivity, achieving ultra-sensitive multi-nuclear MRI, is widely used in lung and brain application research, including evaluation of microstructure, ventilation function, and gas-blood exchange function. Dynamic nuclear polarization (DNP) and parahydrogen-induced polarization (PHIP) can also be used 13 C signal polarization enhancement. On the other hand, multi-nuclear MRI usually faces problems such as short T2 relaxation time and long sampling time. In summary, it is urgent to develop a fast and quantitative multi-nuclear MRI method under hyperpolarized state.

[0005] The magnetic resonance fingerprinting (MRF) technology proposed by Dan Ma et al. [Magnetic resonance fingerprinting[J]. Nature, 2013, 495(7440): 187-192.] is a fast multi-parameter quantitative magnetic resonance imaging method based on pattern recognition, which can obtain the T1 and T2 distribution information of tissue protons within tens of seconds to minutes.

[0006] Unlike protons, multi-nuclear MRI is often combined with hyperpolarization techniques such as SEOP and DNP, and its polarization degree cannot be restored in tissues. At the same time, multi-nuclear MRI often faces problems such as short T2 relaxation time. The MRI pulse sequence design and data reconstruction of multi-nuclear MRI are different from those of proton MRI. Currently, there is no multi-nuclear MRI fingerprint imaging method suitable for multi-parameter quantification under hyperpolarized state. Other quantitative methods, such as those based on spin echo (SE) sequence and gradient echo (GE) sequence, have slow measurement speeds and usually only obtain single parameters such as T1 or T2, which are difficult to meet the needs of rapid quantitative relaxation parameters in the hyperpolarized state. Summary of the invention

[0007] The purpose of the present invention is to provide a multi-core magnetic resonance fingerprint imaging method in view of the above problems existing in the prior art.

[0008] The above-mentioned purpose of the present invention is achieved by the following technical means:

[0009] A method for multi-core magnetic resonance fingerprint imaging, comprising the following steps:

[0010] Step 1, constructing a pulse sequence for multi-core magnetic resonance fingerprint imaging;

[0011] Step 2, constructing an objective function for optimizing the sequence parameters of the magnetic resonance fingerprint imaging pulse sequence, calculating the experimental sequence parameters according to the objective function, and then calculating the single scanning time of the magnetic resonance fingerprint imaging pulse sequence according to the experimental sequence parameters;

[0012] The sequence parameters of the MRI fingerprint imaging pulse sequence include flip angle FA, repetition time TR, and echo time TE;

[0013] Step 3, using the experimental sequence parameters in step 2 to scan the magnetic resonance fingerprint imaging pulse sequence in step 1 to obtain multi-nuclear raw magnetic resonance data;

[0014] Step 4: Based on the experimental sequence parameters obtained in step 2, use the Brook equation to simulate different T1, T2 and T2 *The signal evolution curves of the transverse magnetization vector under the experimental sequence parameters under three relaxation parameters. The signal evolution curves corresponding to all relaxation parameters constitute the fingerprint dictionary of relaxation parameters, and the singular value decomposition method is used to compress the fingerprint dictionary.

[0015] Step 5, reconstructing the multi-core raw magnetic resonance data obtained in step 3 into an image data set using non-uniform fast Fourier transform, and compressing the image data set using a singular value decomposition method;

[0016] Step 6: Match each voxel in the compressed image data set with the compressed fingerprint dictionary, and use matrix least squares method to obtain the signal fingerprint with the smallest difference from the voxel signal to obtain the voxel's T1, T2, and T2 * The value distribution of .

[0017] The magnetic resonance fingerprint imaging pulse sequence constructed in step 1 as described above is:

[0018] The magnetic resonance fingerprint imaging pulse sequence includes a pulse sequence set with a set number of cycles, each pulse sequence set includes a T2 preparation pulse and a fast small-angle excitation sampling pulse set, and the fast small-angle excitation sampling pulse set includes a set number of fast small-angle excitation sampling pulses with different flip angles FA, different repetition times TR, and different echo times TE;

[0019] The timing of the pulse sequence set is to first apply the T2 preparation pulse, and then apply the fast small-angle excitation sampling pulse set;

[0020] The sampling trajectory of the magnetic resonance fingerprint imaging pulse sequence is a variable density SPIRAL spiral trajectory with segmented excitation, combined with undersampling to further reduce the sampling time.

[0021] As mentioned above, the T2 preparation pulse adopts an adiabatic pulse, which includes a first adiabatic half-pass pulse and a second adiabatic half-pass pulse and satisfies the adiabatic condition;

[0022] The first adiabatic half-pass pulse is modulated by the hyperbolic tangent equation tank and the tangent equation tan. The pulse intensity modulation and pulse frequency modulation of the first adiabatic half-pass pulse are:

[0023]

[0024]

[0025] Where B1(t) is the variation of the pulse intensity of the first adiabatic half-pass pulse with time t, and B 1max is the maximum pulse intensity of the first adiabatic half-pass pulse, T p The total time to prepare the pulse is T2, and the time for the first adiabatic half-pass pulse is Tp / 2, Δω(t) is the change of the pulse frequency of the first adiabatic half-pass pulse with time t, ω max is the maximum frequency of the pulse frequency modulation of the first adiabatic half-pass pulse, ω off is the bias frequency set for the first adiabatic half-pass pulse, κ is the rate constant of the pulse frequency of the first adiabatic half-pass pulse changing with time t;

[0026] The pulse intensity modulation and pulse frequency modulation of the second adiabatic half-pass pulse and the pulse intensity modulation and pulse frequency modulation of the first adiabatic half-pass pulse are left-right symmetrical on the time axis; the pulse maximum intensity of the first adiabatic half-pass pulse and the pulse maximum intensity of the second adiabatic half-pass pulse are modified in each cycle.

[0027] As mentioned above, step 2 specifically includes the following steps:

[0028] Step 2.1: Construct the sequence parameters of the magnetic resonance fingerprint imaging pulse sequence under T1, T2, and T2 * The Fisher information matrix I(μ[n],θ) of

[0029] Wherein, μ[n] is the sequence parameter of the magnetic resonance fingerprint imaging pulse sequence, μ[n]=[TR n , F.A. n ,TE n ] T , TR n , FA n , and TE n are the parameter settings of the nth repetition time TR, the nth flip angle FA, and the nth echo time TE corresponding to the nth fast small-angle excitation sampling pulse in a pulse sequence set of the magnetic resonance fingerprint imaging pulse sequence, θ is the relaxation parameter, θ=[T1,T2,T2 * ] T ;

[0030] Step 2.2: Obtain the inverse matrix I of the Fisher information matrix according to the Fisher information matrix -1 (μ[n],θ), and then according to the inverse matrix I of the Fisher information matrix -1 (μ[n],θ) constructs the objective function of optimizing sequence parameters. The objective function of optimizing sequence parameters is:

[0031]

[0032] TT min ≤TR n ≤TR max ,TE min ≤TE n ≤TE max,

[0033] TE n <TR n -3,|FA n -FA n-1 |<2

[0034] In the formula, tr is the matrix to be solved (W×I -1 (μ[n],θ)), W is the weight matrix related to the relaxation parameter θ, l is the accumulation parameter, N is the number of samplings, st is the constraint condition, TR min and TR max are the minimum and maximum values ​​of the range of values ​​set for the repetition time TR, TE min and TE max They are the minimum and maximum values ​​of the echo time TE setting range, TE n <TR n -3 is the repetition time TR is at least 3ms longer than the echo time TE, |FA n -FA n-1 |<2 means the absolute value of each change of the flip angle FA is less than 2 degrees. is a numerical minimization function, and the calculation makes the function The μ[n] value when the minimum value is taken, where the μ[n] value is the value of the sequence parameter;

[0035] Step 2.3. Finally, based on the objective function of constructing the optimized sequence parameters, use MATLAB's nonlinear minimization function to calculate the optimized experimental sequence parameters of flip angle FA, repetition time TR, and echo time TE, and then calculate the single scanning time of the magnetic resonance fingerprint imaging pulse sequence based on the experimental sequence parameters.

[0036] In step 4, the use of the singular value decomposition method to compress the fingerprint dictionary specifically includes the following process:

[0037] After the fingerprint dictionary is decomposed by the singular value decomposition method, 300 columns of right eigenvectors are obtained. The fingerprint dictionary is multiplied with the first 30 columns of right eigenvectors to obtain a fingerprint dictionary compressed 10 times.

[0038] In step 5 above, the use of the singular value decomposition method to compress the image data set specifically includes the following process:

[0039] The image dataset is multiplied by the first 30 columns of right eigenvectors to obtain an image dataset compressed 10 times.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] (1) The magnetic resonance fingerprint imaging pulse sequence designed by the present invention can be used for the quantification of relaxation parameters of multiple nuclei in a hyperpolarized state;

[0042] (2) The present invention can perform T1, T2, and T2 * Simultaneous quantification of multiple relaxation parameters, more parameters are obtained, and the quantification speed is fast;

[0043] (3) The present invention uses adiabatic pulse modulation to modulate the weight of T2, which is suitable for accurate quantification of multi-nuclear short T2 relaxation time. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flow chart of the method of the present invention;

[0045] Figure 2 is a schematic diagram of a magnetic resonance fingerprint pulse sequence of the method of the present invention;

[0046] Figure 3 is a scheme of the flip angle FA, repetition time TR, and echo time TE after the sequence parameters are optimized in the method of the present invention (the brackets of the ordinate are units, the unit of the flip angle FA is degree, and the unit of the repetition time TR and the echo time TE are both milliseconds (ms));

[0047] Figure 4 is the signal fingerprint corresponding to the relaxation parameters of the three samples in the fingerprint dictionary established in Example 1 of the present invention;

[0048] Figure 5 is an example of a fingerprint dictionary matching result of the method of the present invention;

[0049] Figure 6 are T1, T2, and T2 of the three samples obtained in step 6 of Example 1 of the present invention. * Relaxation parameter distribution (the numbers 1 / 2 / 3 in the figure correspond to sample 1, sample 2, and sample 3, respectively. T1 map is the value distribution of the T1 relaxation parameter of the three samples in Example 1. T2 map is the value distribution of the T2 relaxation parameter of the three samples in Example 1. T2 * map is the T2 of the three samples in Example 1 * distribution of values ​​of the relaxation parameter). DETAILED DESCRIPTION

[0050] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with embodiments. The embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0051] Embodiment 1:

[0052] A method for multi-core magnetic resonance fingerprint imaging, comprising the following steps:

[0053] Step 1, constructing a pulse sequence for multi-core magnetic resonance fingerprint imaging;

[0054] The magnetic resonance fingerprint imaging pulse sequence includes a pulse sequence set with a set number of cycles, each pulse sequence set includes a T2 preparation pulse and a fast small-angle excitation sampling pulse set, and the fast small-angle excitation sampling pulse set includes a set number of fast small-angle (less than 15 degrees) excitation sampling pulses with different flip angles (FA), different repetition times (TR), and different echo times (TE);

[0055] Each fast small-angle excitation sampling pulse performs one excitation sampling, so the number of fast small-angle excitation sampling pulses in each fast small-angle excitation sampling pulse set is the sampling number of each fast small-angle excitation sampling pulse set, and also the sampling number of each pulse sequence set; the sampling number of the magnetic resonance fingerprint imaging pulse sequence is the sampling number of each pulse sequence set multiplied by the number of cycles;

[0056] The timing of the pulse sequence set is to first apply the T2 preparation pulse, and then apply the fast small-angle excitation sampling pulse set;

[0057] In this embodiment, the sampling times of each fast small-angle excitation sampling pulse set is set to 30, and the cycle times of the pulse sequence set is set to 10, so the magnetic resonance fingerprint imaging pulse sequence has a total of 300 excitation samplings;

[0058] The T2 preparation pulse adopts an adiabatic pulse, which includes two adiabatic half-pass pulses (a first adiabatic half-pass pulse and a second adiabatic half-pass pulse) and satisfies the adiabatic condition;

[0059] The first adiabatic half-pass pulse is modulated by the hyperbolic tangent equation tanh and the tangent equation tan. The modulations of the pulse intensity and pulse frequency of the first adiabatic half-pass pulse are:

[0060]

[0061]

[0062] Where B1(t) is the variation of the pulse intensity of the first adiabatic half-pass pulse with time t, and B 1max is the maximum pulse intensity of the first adiabatic half-pass pulse, T p The total time of T2 preparation pulse, the time of the first adiabatic half-pass pulse accounts for the total time of T2 preparation pulse T p 1 / 2, that is, T p / 2, Δω(t) is the change of the pulse frequency of the first adiabatic half-pass pulse with time t, ωmax is the maximum frequency of the pulse frequency modulation of the first adiabatic half-pass pulse, ω off is the bias frequency set for the first adiabatic half-pass pulse, and κ is the rate constant at which the pulse frequency of the first adiabatic half-pass pulse changes with time t.

[0063] In this embodiment, the first adiabatic half-pass pulse is set to: maximum frequency ω max Set to 20000Hz, bias frequency ω off The frequency was set to 2000 Hz, the rate constant κ was set to 3, and the maximum pulse intensity B 1max The 10 modifications are 30 / 27 / 24 / 21 / 18 / 15 / 12 / 10 / 8 / 6μT respectively. The total time of T2 preparation pulse T p Set to 20ms.

[0064] The pulse intensity modulation and pulse frequency modulation of the second adiabatic half-pass pulse and the pulse intensity modulation and pulse frequency modulation of the first adiabatic half-pass pulse are symmetrical on the time axis. The maximum pulse intensity of the T2 preparation pulse (including the maximum pulse intensity of the first adiabatic half-pass pulse and the maximum pulse intensity of the second adiabatic half-pass pulse) is modified in each cycle to modulate the T2 weight, and the experimental data is sampled.

[0065] T1, T2, and T2 for multi-core elements * The present invention introduces a design of relaxation mechanism based on the pulse sequence of magnetic resonance fingerprint imaging to quickly quantify three relaxation parameters. In the hyperpolarized state, the multi-nuclear element magnetic resonance signal decays toward thermal polarization. Combined with the variable flip angle FA, the variable repetition time TR, and the variable echo time TE, fast small-angle excitation sampling is used. After each excitation sampling, the transverse magnetization vector is destroyed. Then the longitudinal magnetization vector Mz signal can be expressed as:

[0066] M Z [n]=cos(FA[n])×exp(-TR[n] / T1)×Mz[n-1] (3)

[0067] Wherein, Mz[n] is the magnitude of the longitudinal magnetization vector after the nth flip angle FA[n] and the nth repetition time TR[n], and the initial longitudinal magnetization vector Mz[0] is normalized to 1. The influence of T1 relaxation is introduced into the longitudinal magnetization vector Mz signal.

[0068] Adjusting the echo time TE can adjust T2 * Weight, after the nth echo time TE, the transverse magnetization vector Mxy sampled is:

[0069] Mxy[n]=sin(FA[n])×exp(-TE[n] / T2 * )×Mz[n] (4)

[0070] Wherein, Mxy[n] is the magnitude of the transverse magnetization vector after the nth flip angle FA[n] and the nth repetition time TR[n]. The transverse magnetization vector Mxy signal introduces T2 * The influence of relaxation parameters.

[0071] Multi-core (such as 129 Xe, 23 Na) has a short T2 relaxation time, and it is difficult to introduce T2 weight. The present invention uses an adiabatic pulse to excite the magnetization vector, and uses the adiabatic pulse as a T2 preparation pulse. It is characterized in that under the action of the adiabatic pulse, the signal evolution is affected by both T1 and T2 relaxation parameters, and the weights of the influence of the T1 and T2 relaxation parameters can be modulated by the frequency and intensity of the adiabatic pulse, so that it can be applied to different T2 situations. The influence of the adiabatic pulse on the signal evolution can be simulated and calculated by the Bloch equation.

[0072] The three multi-nuclear samples simulated in this embodiment are all in a hyperpolarized state and relax to a thermally polarized state. The T1, T2, and T2 * The three relaxation parameters are set as:

[0073] Sample 1: T1 = 10000ms, T2 = 50ms, T2 * =10ms;

[0074] Sample 2: T1 = 8000ms, T2 = 30ms, T2 * =6ms;

[0075] Sample 3: T1 = 5000ms, T2 = 20ms, T2 * =3ms.

[0076] The sampling process of this embodiment uses MATLAB for simulation calculation, mainly including simulating the magnetic resonance signal obtained by sampling a multi-nuclear sample under the sequence parameters. The sampling trajectory of the magnetic resonance fingerprint imaging pulse sequence is a variable density SPIRAL spiral trajectory with segmented excitation, and combined with undersampling to further reduce the sampling time. The relevant sampling parameter settings of this embodiment are: the field of view (FOV) is set to 10×10cm 2 , the sampling matrix was set to 96 × 96, the number of segmented excitations was set to 4, the maximum gradient was set to 200 mT / m, and the maximum gradient climbing rate was set to 500 T / m / s.

[0077] Step 2, constructing an objective function for optimizing the sequence parameters of the magnetic resonance fingerprint imaging pulse sequence, calculating the experimental sequence parameters according to the objective function, and then calculating the single scanning time of the magnetic resonance fingerprint imaging pulse sequence according to the experimental sequence parameters;

[0078] The sequence parameters of the MRI fingerprint imaging pulse sequence include flip angle FA, repetition time TR, and echo time TE;

[0079] Step 2.1: According to formula (3) and formula (4), construct the sequence parameters of the magnetic resonance fingerprint imaging pulse sequence under T1, T2, and T2 * The Fisher information matrix I(μ[n],θ) of

[0080] Wherein, μ[n] is the sequence parameter of the magnetic resonance fingerprint imaging pulse sequence, μ[n]=[TR n , F.A. n ,TE n ] T , TR n , FA n , and TE n are the parameter settings of the nth repetition time TR, the nth flip angle FA, and the nth echo time TE corresponding to the nth fast small-angle excitation sampling pulse in a pulse sequence set of the magnetic resonance fingerprint imaging pulse sequence, θ is the relaxation parameter, θ=[T1,T2,T2 * ] T ;

[0081] Step 2.2: Obtain the inverse matrix I of the Fisher information matrix according to the Fisher information matrix -1 (μ[n],θ), and then according to the inverse matrix I of the Fisher information matrix -1 (μ[n],θ) constructs the objective function of optimizing sequence parameters. The objective function of optimizing sequence parameters is:

[0082]

[0083] TT min ≤TR n ≤TR max ,TE min ≤TE n ≤TE max ,

[0084] TE n <TR n -3,|FA n -FA n-1 |<2 (6)

[0086] In the formula, tr represents the solution matrix (W×I -1 (μ[n],θ)), W is the weight matrix related to the relaxation parameter θ, l is the cumulative parameter, starting from 1 and taking integer values ​​until it reaches N, N is the number of sampling times of the magnetic resonance fingerprint imaging pulse sequence, st represents the constraint condition, TR min and TR max are the minimum and maximum values ​​of the range of values ​​set for the repetition time TR, TE min and TE max They are the minimum and maximum values ​​of the echo time TE setting range, TE n <TR n -3 means that the repetition time TR is at least 3ms longer than the echo time TE, |FA n -FA n-1 |<2 means that the absolute value of each change of the flip angle FA is less than 2 degrees. is a numerical minimization function, and the calculation makes the function The μ[n] value when the minimum value is taken, where the μ[n] value is the value of the sequence parameter;

[0087] In this embodiment, the repetition time TR is set to a range of 5 to 25 ms, the echo time TE is set to a range of 2 to 12 ms, and the repetition time TR is at least 3 ms longer than the echo time TE. The absolute value of each change of the flip angle FA is less than 2 degrees.

[0088] Step 2.3: Finally, according to the objective function of constructing the optimized sequence parameters, the nonlinear minimization function (fmincon function) of MATLAB is used to calculate the optimized experimental sequence parameters of flip angle FA, repetition time TR, and echo time TE (such as Figure 3 The single scan time of the magnetic resonance fingerprint imaging pulse sequence is then calculated based on the experimental sequence parameters.

[0089] In this embodiment, the magnetic resonance fingerprint imaging pulse sequence has 300 excitation samples, and the calculated single scan time is 5842 ms.

[0090] Step 3, using the experimental sequence parameters of the magnetic resonance fingerprint imaging sequence optimized in step 2 to scan the magnetic resonance fingerprint imaging pulse sequence in step 1 to obtain multi-nuclear raw magnetic resonance data;

[0091] Step 4: Based on the experimental sequence parameters obtained in step 2, use the Brook equation to simulate different T1, T2, and T2 *The signal evolution curves (also called signal fingerprints) of the transverse magnetization vector Mxy signal under the experimental sequence parameters under three relaxation parameters correspond one to one with the relaxation parameters; the signal evolution curves corresponding to all relaxation parameters constitute the fingerprint dictionary of the relaxation parameters, and then the singular value decomposition (SVD) method is used to compress the fingerprint dictionary to reduce the matrix size of the fingerprint dictionary;

[0092] In this embodiment, the range of the relaxation parameter is set to T1 = [2000:200:12000] (T1 is 2000 from the minimum value, and increases by 200 each time until it reaches 12000; T2 and T2 * The same is true), T2 = [10:2:60], T2 * =[1:0.2:10], a total of 60996 signal evolution curves are obtained to form a fingerprint dictionary. The signal fingerprints corresponding to the relaxation parameters of the three samples are as follows: Figure 4 shown.

[0093] The singular value decomposition (SVD) method is used to compress the fingerprint dictionary. Specifically, after the fingerprint dictionary is decomposed by SVD, 300 columns of right eigenvectors are obtained. The fingerprint dictionary is multiplied with the first 30 columns of right eigenvectors to obtain a fingerprint dictionary compressed 10 times;

[0094] Step 5, reconstructing the multi-core raw magnetic resonance data obtained in step 3 into an image data set using non-uniform fast Fourier transform, and compressing the image data set using a singular value decomposition (SVD) method;

[0095] The image dataset is compressed using the singular value decomposition (SVD) method, specifically: the image dataset is multiplied by the first 30 columns of right eigenvectors in step 4 to obtain an image dataset compressed 10 times;

[0096] Step 6: Match each voxel in the compressed image data set with the compressed fingerprint dictionary, and use matrix least squares method to obtain the signal fingerprint with the smallest difference from the voxel signal, so as to obtain the T1, T2, and T2 of the voxel. * The value distribution of the three relaxation parameters completes the quantification of the three relaxation parameters.

[0097] The image data set and fingerprint dictionary are compressed using the singular value decomposition method before matching, which can reduce the matching speed by 10 times.

[0098] Embodiment 2:

[0099] A magnetic resonance fingerprint imaging device for multi-core, comprising:

[0100] A magnetic resonance fingerprint imaging pulse sequence construction module, used to implement step 1 of the above embodiment 1;

[0101] A sequence parameter optimization module, used to implement step 2 of embodiment 1;

[0102] A scanning module, used to implement step 3 of embodiment 1;

[0103] A fingerprint dictionary calculation module, used to implement step 4 of embodiment 1;

[0104] An image data set reconstruction module, used to implement step 5 of embodiment 1;

[0105] An image data set and fingerprint dictionary matching module, used to implement step 6 of embodiment 1;

[0106] A terminal comprises a processor and a storage medium;

[0107] The storage medium is used to store instructions;

[0108] The processor is used to operate according to the instruction to execute steps 1 to 6 of embodiment 1.

[0109] A computer-readable storage medium stores a computer program, which implements steps 1 to 6 of embodiment 1 when executed by a processor.

[0110] It should be noted that the embodiments described in the present invention are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the described embodiments or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the attached claims.

Claims

1. A method for multi-core magnetic resonance fingerprint imaging, characterized in that: The following steps are involved: Step 1, constructing a pulse sequence for multi-core magnetic resonance fingerprint imaging; The magnetic resonance fingerprint imaging pulse sequence is: The magnetic resonance fingerprint imaging pulse sequence includes a pulse sequence set with a set number of cycles, each pulse sequence set includes a T2 preparation pulse and a fast small-angle excitation sampling pulse set, and the fast small-angle excitation sampling pulse set includes a set number of fast small-angle excitation sampling pulses with different flip angles FA, different repetition times TR, and different echo times TE; The adiabatic pulse is used as the T2 preparation pulse; Step 2, constructing an objective function for optimizing the sequence parameters of the magnetic resonance fingerprint imaging pulse sequence, calculating the experimental sequence parameters according to the objective function, and then calculating the single scanning time of the magnetic resonance fingerprint imaging pulse sequence according to the experimental sequence parameters; The sequence parameters of the MRI fingerprint imaging pulse sequence include flip angle FA, repetition time TR, and echo time TE; The step 2 specifically includes the following steps: Step 2.1: Construct the sequence parameters of the magnetic resonance fingerprint imaging pulse sequence according to the longitudinal magnetization vector Mz and the transverse magnetization vector Mxy. * The Fisher information matrix I(μ[n],θ) of Wherein, μ[n] is the sequence parameter of the magnetic resonance fingerprint imaging pulse sequence, μ[n]=[TR n , F.A. n ,TE n ] T , TR n , FA n , and TE n are the parameter settings of the nth repetition time TR, the nth flip angle FA, and the nth echo time TE corresponding to the nth fast small-angle excitation sampling pulse in a pulse sequence set of the magnetic resonance fingerprint imaging pulse sequence, θ is the relaxation parameter, θ=[T1,T2,T2 * ] T ; The longitudinal magnetization vector Mz is calculated based on the following formula: M Z [n]=cos(FA[n])×exp(-TR[n] / T1)×Mz[n-1] Where Mz[n] is the magnitude of the longitudinal magnetization vector after the nth flip angle FA[n] and the nth repetition time TR[n], and the initial longitudinal magnetization vector Mz[0] is normalized to 1; The transverse magnetization vector Mxy is calculated based on the following formula: Mxy[n]=sin(FA[n])×exp(-TE[n] / T2 * )×Mz[n] Where Mxy[n] is the magnitude of the transverse magnetization vector after the nth flip angle FA[n] and the nth repetition time TR[n]; Step 2.2: Obtain the inverse matrix I of the Fisher information matrix according to the Fisher information matrix -1 (μ[n],θ), and then according to the inverse matrix I of the Fisher information matrix -1 (μ[n],θ) constructs the objective function of optimizing sequence parameters. The objective function of optimizing sequence parameters is: stTR min ≤TR n ≤TR max ,THE min ≤TE n ≤TE max , TEA n <TR n -3,|FA n -AGO n-1 |<2 In the formula, tr is the matrix to be solved (W×I -1 (μ[n],θ)), W is the weight matrix related to the relaxation parameter θ, l is the accumulation parameter, N is the number of samplings, st is the constraint condition, TR min and TR max are the minimum and maximum values ​​of the range of values ​​set for the repetition time TR, TE min and TE max They are the minimum and maximum values ​​of the echo time TE setting range, TE n <TR n -3 means the repetition time TR is at least 3 ms longer than the echo time TE, |FA n -FA n-1 |<2 means the absolute value of each change of the flip angle FA is less than 2 degrees. is a numerical minimization function, and the calculation makes the function The μ[n] value when the minimum value is taken, where the μ[n] value is the value of the sequence parameter; Step 2.3, finally, according to the objective function of constructing the optimized sequence parameters, the nonlinear minimization function of MATLAB is used to calculate the optimized experimental sequence parameters of the flip angle FA, repetition time TR, and echo time TE, and then the single scanning time of the magnetic resonance fingerprint imaging pulse sequence is calculated according to the experimental sequence parameters; Step 3, using the experimental sequence parameters in step 2 to scan the magnetic resonance fingerprint imaging pulse sequence in step 1 to obtain multi-nuclear raw magnetic resonance data; Step 4: Based on the experimental sequence parameters obtained in step 2, use the Brook equation to simulate different T1, T2 and T2 * The signal evolution curves of the transverse magnetization vector under the experimental sequence parameters under three relaxation parameters. The signal evolution curves corresponding to all relaxation parameters constitute the fingerprint dictionary of relaxation parameters, and the singular value decomposition method is used to compress the fingerprint dictionary. Step 5, reconstructing the multi-core raw magnetic resonance data obtained in step 3 into an image data set using non-uniform fast Fourier transform, and compressing the image data set using a singular value decomposition method; Step 6: Match each voxel in the compressed image data set with the compressed fingerprint dictionary, and use matrix least squares method to obtain the signal fingerprint with the smallest difference from the voxel signal to obtain the voxel's T1, T2, and T2 * The value distribution of .

2. A method for multi-core magnetic resonance fingerprint imaging according to claim 1, characterized in that: The timing of the pulse sequence set is to first apply a T2 preparation pulse, and then apply a fast small-angle excitation sampling pulse set; The sampling trajectory of the magnetic resonance fingerprint imaging pulse sequence is a variable density SPIRAL spiral trajectory with segmented excitation, combined with undersampling to further reduce the sampling time.

3. A method for multi-core magnetic resonance fingerprint imaging according to claim 2, characterized in that: The adiabatic pulse includes a first adiabatic half-pass pulse and a second adiabatic half-pass pulse, and satisfies an adiabatic condition; The first adiabatic half-pass pulse is modulated by the hyperbolic tangent equation tanh and the tangent equation tan. The pulse intensity modulation and pulse frequency modulation of the first adiabatic half-pass pulse are: Where B1(t) is the variation of the pulse intensity of the first adiabatic half-pass pulse with time t, and B 1max is the maximum pulse intensity of the first adiabatic half-pass pulse, T p The total time to prepare the pulse is T2, and the time for the first adiabatic half-pass pulse is T p / 2, Δω(t) is the change of the pulse frequency of the first adiabatic half-pass pulse with time t, ω max is the maximum frequency of the pulse frequency modulation of the first adiabatic half-pass pulse, ω off is the bias frequency set for the first adiabatic half-pass pulse, κ is the rate constant of the pulse frequency of the first adiabatic half-pass pulse changing with time t; The pulse intensity modulation and pulse frequency modulation of the second adiabatic half-pass pulse and the pulse intensity modulation and pulse frequency modulation of the first adiabatic half-pass pulse are left-right symmetrical on the time axis; the pulse maximum intensity of the first adiabatic half-pass pulse and the pulse maximum intensity of the second adiabatic half-pass pulse are modified in each cycle.

4. A method for multi-core magnetic resonance fingerprint imaging according to claim 3, characterized in that: The step 4 of using the singular value decomposition method to compress the fingerprint dictionary specifically includes the following process: After the fingerprint dictionary is decomposed by the singular value decomposition method, 300 columns of right eigenvectors are obtained. The fingerprint dictionary is multiplied with the first 30 columns of right eigenvectors to obtain a fingerprint dictionary compressed 10 times.

5. A method for multi-core magnetic resonance fingerprint imaging according to claim 4, characterized in that: The step 5 of compressing the image data set using the singular value decomposition method specifically includes the following process: the image data set is multiplied by the first 30 columns of right eigenvectors to obtain an image data set compressed 10 times.

Images