A method of fast multi-parametric spectroscopy and functional imaging based on magnetic resonance fingerprinting

Abstract

The application discloses a kind of fast multi-parameter spectrum and functional imaging method based on magnetic resonance fingerprinting.It includes: constructing selective magnetic resonance spectrum fingerprinting imaging sequence;Optimizing acquisition parameters in selective magnetic resonance spectrum fingerprinting imaging sequence;Using the optimized selective magnetic resonance spectrum fingerprinting imaging sequence to collect water signal and metabolite comprehensive signal, reconstruct the spatial fingerprint of water signal according to water signal, and reconstruct the spatial fingerprint spectrum of each metabolite according to metabolite comprehensive signal;Match the spatial fingerprint of water signal with water signal simulation dictionary, and obtain the parameter map of water signal;Match the spatial fingerprint spectrum of each metabolite with metabolite simulation dictionary, and obtain the parameter map of metabolite.The application can realize high-speed, multi-target, multi-parameter, absolute quantitative spectrum imaging and metabolism-function combined imaging.

Description

Technical Field

[0001] This invention belongs to the field of magnetic resonance spectroscopy imaging, and in particular relates to a fast multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting. Background Technology

[0002] Magnetic resonance spectroscopy (MRI) is a non-invasive in vivo examination technique used to detect the chemical properties and structural characteristics of various metabolite molecules in living organisms. The relaxation time of metabolites (including longitudinal relaxation time T1 and transverse relaxation time T2) is not only a key parameter for the absolute quantitative analysis of their concentration but also serves as a bioindicator reflecting the cellular microenvironment. Traditional measurement methods (such as variable repetition time or inversion recovery sequence for T1 measurement and multi-echo measurement for T2) can obtain metabolite relaxation times, but the long scan times required make it very difficult to obtain relaxation time mappings in MRI. Therefore, there are currently no clinically applicable MRSI sequences that can simultaneously estimate the longitudinal relaxation time T1, transverse relaxation time T2, and high-resolution absolute concentration of metabolites.

[0003] Functional magnetic resonance imaging (fMRI) is the primary tool for studying neuroimaging, monitoring brain activity by examining blood oxygen levels. Current fMRI and magnetic resonance spectroscopy techniques use completely different acquisition methods, and therefore are typically achieved through two separate scans.

[0004] Magnetic resonance fingerprinting is an important method in the field of quantitative magnetic resonance imaging (MRI), providing a general framework for rapid multi-parameter quantitative imaging. The emergence of monomorphic magnetic resonance spectroscopy fingerprinting, based on the principle of magnetic resonance fingerprinting, has paved the way for multi-parameter measurement and absolute quantification of monomorphic spectra. However, compared to multi-voxel imaging, monomorphic spectroscopy has disadvantages such as a smaller measurement range and the tendency to overlook lesions.

[0005] Therefore, existing imaging technologies cannot simultaneously achieve rapid, multi-parameter measurement and absolutely quantitative spectral and functional imaging to map the relaxation time and spatial distribution of absolute concentrations of metabolites and brain functional imaging. Summary of the Invention

[0006] To address the problems existing in the background art, the present invention aims to provide a rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting. The present invention utilizes a novel rapid spectral imaging acquisition sequence to simultaneously construct multi-dimensional spectral fingerprints of water and metabolite signals, as well as corresponding fingerprint signal reconstruction, thereby simultaneously mapping high-resolution, multi-parameter-measured, absolutely quantitative images and functional images of multiple metabolites in vivo.

[0007] The technical solution adopted in this invention is:

[0008] The rapid multi-parameter spectral and functional imaging method includes the following steps:

[0009] Step S1: Construct a selective magnetic resonance spectroscopy fingerprinting sequence. The selective magnetic resonance spectroscopy fingerprinting sequence includes at least one cyclic structure, which mainly consists of multiple time-series cascaded basic units with identical structures but different acquisition parameters. The basic units are used to simultaneously acquire water and metabolite signals, and the cyclic structure is used to achieve fingerprint encoding and spatial encoding of the water and metabolite signals.

[0010] In the selective magnetic resonance spectroscopy fingerprint imaging sequence, fingerprint encoding is achieved by changing the repetition time TR, echo time TE, and excitation pulse flip angle FA in each basic unit within the cyclic structure.

[0011] Step S2: Optimize the acquisition parameters in the selective magnetic resonance spectroscopy fingerprinting sequence based on the relaxation characteristics of all target metabolites to obtain the optimized selective magnetic resonance spectroscopy fingerprinting sequence. The acquisition parameters include the repetition time TR, echo time TE, and excitation pulse flip angle FA in each basic unit within each cyclic structure.

[0012] Step S3: Use the optimized selective magnetic resonance spectroscopy fingerprinting sequence to acquire water signals and metabolite composite signals, reconstruct the spatial fingerprint of the water signal based on the water signal, and reconstruct the spatial fingerprint spectrum of each metabolite based on the composite metabolite signal.

[0013] Step S4: Match the spatial fingerprint of the water signal with the water signal simulation dictionary to obtain the optimal parameter estimate of the water signal, and then obtain the parameter map of the water signal. Match the spatial fingerprint spectrum of each metabolite with the metabolite simulation dictionary to obtain the optimal parameter estimate of the metabolite, and then obtain the parameter map of the metabolite.

[0014] The selective magnetic resonance spectroscopy fingerprinting sequence includes an EPSFI sequence, and the basic unit of the EPSFI sequence includes, in sequence:

[0015] The fat inhibition module applies frequency-selective fat inhibition pulses (selecting only the frequency range where fat is located) to reduce fat contamination of metabolite spectra.

[0016] The layer-selective excitation module simultaneously applies a Z-axis gradient and a radio frequency pulse to excite the region of interest.

[0017] The phase encoding module performs spatial phase encoding.

[0018] The water signal acquisition module is used to acquire water signals. It can use gradient echo, planar gradient echo, multi-echo gradient echo, EPI readout, or a combination thereof to acquire water signals.

[0019] A frequency-selective refocusing module applies two adiabatic pulses with variable center frequencies, coupled with disruptive gradients. Specifically, a pair of disruptive gradients is applied before and after the application of each adiabatic pulse with a variable center frequency. Each pair of disruptive gradients includes one disruptive gradient applied along the X-axis and another simultaneously applied along the Y-axis. This module refocuses metabolite signals using a pair of frequency-selective adiabatic pulses, excluding water signals from the refocusing frequency band. The center frequency and bandwidth of the frequency-selective refocusing module cover the frequency bands of all target metabolites. The disruptive gradients are used to reduce interference from water peaks, lipid peaks, or non-target frequency band signals on the spectrum of the target metabolites, and can also be replaced by at least one of a lipid suppression module, a water suppression module, or a water signal avoidance module.

[0020] The planar multi-echo metabolite readout module reads out metabolite signals through planar gradient echo.

[0021] The selective magnetic resonance spectroscopy fingerprinting sequence includes qT2. * -EPSFI sequence: For the EPSFI sequence, the water signal acquisition module is configured to acquire water signals using planar gradient echo, and acquires water signals with at least two different echo times to obtain qT2. * -EPSFI sequence.

[0022] By fitting or matching the water signals with at least two different echo times, the T2 of the water signal can be obtained. * picture.

[0023] The EPSFI sequence and qT2 * In the phase encoding module of the EPSFI sequence, the gradient distribution of the Y-axis and Z-axis is successively phase encoded.

[0024] The selective magnetic resonance spectral fingerprinting sequence includes the BOLD-EPSFI sequence: For the EPSFI sequence, the water signal acquisition module is changed to be located after the layer-selective excitation module and before the frequency-selective refocusing module, and is configured to read out the water signal gradient through EPI. The phase encoding module is changed to be located after the frequency-selective refocusing module and before the planar multi-echo metabolite readout module, and is configured to perform successive phase encoding on the Y-axis gradient.

[0025] When imaging using the BOLD-EPSFI sequence, the method further includes:

[0026] The spatial fingerprint of the water signal is matched with the water signal simulation dictionary to obtain the excitation field inhomogeneity B1 map. The parameter map of the water signal includes the water signal relaxation time. Using the water signal relaxation time and the excitation field inhomogeneity B1 map, a functional imaging map is reconstructed.

[0027] The reconstruction process of functional imaging maps is as follows:

[0028] B1 plot and relaxation time plot T obtained from the water signal obtained in the first cycle 1w As a correction parameter, it corrects the water signal S measured in the current cycle. meas Then, the corrected water signal S corr Correlation analysis was performed with the stimulus paradigm to reconstruct the functional imaging map;

[0029] Corrected water signal S corr It is obtained through the following formula:

[0030] S corr =S meas / S sim ;

[0031] Among them, S corr S represents the corrected water signal for the current cycle. meas S represents the water signal measured in the current cycle. sim This represents the simulated signal obtained through Bloch simulation based on the correction parameters and the sequence parameters of the current loop;

[0032] The Bloch simulation is represented as follows:

[0033] ;

[0034] Where TR, TE, and FA represent the values ​​of the repetition time, echo time, and excitation pulse flip angle in the current loop, respectively. 1w B1 and B1 represent the B1 diagram and relaxation time diagram matched with the water signal obtained in the first cycle, respectively.

[0035] The parameter maps of the metabolites include at least one of the following: metabolite longitudinal relaxation time T1 map, metabolite transverse relaxation time T2 map, metabolite concentration map, and metabolite relaxation-corrected concentration map.

[0036] The dimensions of the water signal include X-space dimension, Y-space dimension, fingerprint encoding dimension, coil channel dimension, and average dimension.

[0037] The dimensions of the metabolite signal include X-space dimension, Y-space dimension, time dimension, fingerprint encoding dimension, coil channel dimension, and average dimension.

[0038] In step S2, a genetic algorithm is used for optimization.

[0039] During the optimization process, the fitness function is set according to the following formula:

[0040] ;

[0041] ;

[0042] In the formula, For fitness, The mean root mean square error is given, and M is the number of target metabolites. The deviation between the true value and the estimated value. The variance of the estimated value is given by T, which represents the relaxation time, indicating the longitudinal relaxation time T1 or the transverse relaxation time T2, and k represents the sequence number of the target metabolite.

[0043] During the optimization process, each individual is composed of the repetition time TR, echo time TE, and excitation pulse flip angle FA of all basic units within each loop structure.

[0044] In step S3, the water signal undergoes phase correction, spatial filtering, spatial Fourier transform, and receiving coil channel merging to obtain the spatial fingerprint of the water signal.

[0045] The metabolite signals were subjected to odd-even echo merging, spatial filtering, spatial Fourier transform, receiving coil channel merging, and time Fourier transform to obtain the spatial fingerprint spectrum of each metabolite.

[0046] Step S4 includes:

[0047] Step S4.1: Use the spatial fingerprint of the water signal obtained in the first loop as the spatial fingerprint of the water signal, and match it with the water signal simulation dictionary to obtain the excitation field inhomogeneity B1 map and the water longitudinal relaxation time T. 1w picture.

[0048] Step S4.2: Fit the spatial fingerprint spectrum of each metabolite to obtain the spatial fingerprint of the metabolite. Using the excitation field inhomogeneity B1 diagram as a known condition, match the metabolite fingerprint with the metabolite simulation dictionary to obtain the metabolite longitudinal relaxation time T1 diagram and the metabolite transverse relaxation time T2 diagram.

[0049] Step S4.3: Calculate the ratio between the spatial fingerprint of the metabolite and the spatial fingerprint of the water signal, and after relaxation and water concentration factor correction, obtain the absolute concentration map of the metabolite.

[0050] In summary, this invention achieves rapid multi-parameter imaging and functional imaging of various metabolites through a novel signal acquisition method, reconstruction, and processing approach. The first step involves acquiring high-dimensional water and metabolite signals using a selective magnetic resonance spectroscopy fingerprinting sequence with optimized acquisition parameters. The second step involves processing the raw water and metabolite signals separately according to the reconstruction process to obtain fingerprint data for water and metabolites. The third step involves performing dot product matching between the fingerprint data for water and metabolites and dictionaries obtained from Bloch simulation to obtain relaxation time maps, absolute quantification maps of metabolites, and functional images for water and metabolites.

[0051] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0052] 1. This invention is the first to apply the magnetic resonance fingerprint framework to the field of spectral imaging, enabling real-time mapping of the three-dimensional spatial distribution of relaxation times (including T1 and T2) of various metabolites.

[0053] 2. The sequence structure provided by this invention can simultaneously measure the relaxation time of water and various metabolites, achieving absolute quantification of metabolites. It is more sensitive in detecting lesions than traditional relative concentration indicators, and the collection time is significantly shortened.

[0054] 3. The acquisition and reconstruction method provided by this invention can simultaneously achieve spectral imaging and functional imaging. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the overall process of the present invention.

[0056] Figure 2 This is a schematic diagram of the selective magnetic resonance spectral fingerprinting sequence of the present invention, wherein (A) is a schematic diagram of the basic unit of the EPSFI sequence; (B) is a schematic diagram of the filling of metabolite signals in various dimensions under the EPSFI cyclic structure, including the relationship between fingerprint coding cycle, spatial coding cycle and spectral temporal acquisition; (C) is qT2 * - A schematic diagram of the basic unit of the EPSFI sequence; (D) is a schematic diagram of the basic unit of the BOLD-EPSFI sequence.

[0057] Figure 3 The results are Bloch simulations of selective refocusing adiabatic pulses.

[0058] Figure 4 For qT2 * - Results of phantom experiments using EPSFI sequences; where (A) represents the T1 and T2 values ​​of water obtained in a uniform spherical magnetic resonance spectroscopy phantom. *Figure B1 shows (B) T1, T2 and absolute concentration plots of NAA, Cr and Cho, (C) dictionary matching results of magnetic resonance spectral fingerprints of representative voxels in the phantom, and (D) comparison of the measured absolute concentrations after normalization to water with the phantom label concentration (gold standard).

[0059] Figure 5 For qT2 * - A schematic diagram of the multimodal output of two representative healthy subjects under EPSFI sequence, showing the water T1 plot and T2 plot. * Figure B1, metabolite concentration graph, and metabolite relaxation graph (T1 and T2).

[0060] Figure 6 Flowchart for reconstructing water and metabolite signals.

[0061] Figure 7 The diagram shows the output of the BOLD-EPSFI sequence, where (A) is the BOLD signal after relaxation correction, (B) is the delay correction analysis diagram, and (C) is the signal diagram after delay correction. Detailed Implementation

[0062] The present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0063] This invention provides a rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting.

[0064] The method of the present invention includes the following steps:

[0065] Step S1: Construct a selective magnetic resonance spectroscopy fingerprinting sequence; the selective magnetic resonance spectroscopy fingerprinting sequence includes at least one cyclic structure, which is mainly composed of multiple basic units with the same structure but different acquisition parameters connected in time series; the basic units are used to simultaneously acquire water and metabolite signals, and the cyclic structure is used to realize fingerprint encoding and spatial encoding of water and metabolite signals.

[0066] In selective magnetic resonance spectroscopy fingerprinting sequences, fingerprint encoding is achieved by changing the repetition time TR, echo time TE, and excitation pulse flip angle FA in each basic unit within the cyclic structure.

[0067] Selective magnetic resonance spectroscopy fingerprinting sequences include EPSFI sequences. The basic units of an EPSFI sequence include: a fat suppression module, which applies frequency-selective fat suppression pulses (selecting only the frequency range where fat is located) to reduce fat contamination of metabolite spectral lines; a layer-selective excitation module, which simultaneously applies a Z-axis gradient and radio frequency pulses to excite the region of interest; a phase encoding module, which performs spatial phase encoding; a water signal acquisition module, used to acquire water signals; gradient echo, planar gradient echo, multi-echo gradient echo, EPI readout, or a combination thereof can be used to acquire water signals; a frequency-selective refocusing module, which applies two adiabatic pulses with variable center frequencies and, in conjunction with a disruptive gradient, achieves selective refocusing of metabolite signals; and a planar multi-echo metabolite readout module, which reads out metabolite signals through planar gradient echo. In EPSFI sequences, water signal readout can be performed immediately after excitation, preserving a strong signal-to-noise ratio for the water signal. Water signal readout can be performed using single-echo gradient echo readout, multi-echo gradient echo readout, or EPI readout suitable for functional imaging. Water signal readout shares the same excitation with subsequent metabolite readout, thus exhibiting spatial and temporal consistency. Metabolite signal readout employs planar multi-echo readout. Within each phase encoding step, multiple echoes are acquired through oscillating readout gradients, thereby simultaneously obtaining spatial encoding and spectral temporal information. The metabolite signal, relative to the water signal, possesses an additional spectral temporal dimension; after time-Fourier transform, the target metabolite spectrum can be obtained.

[0068] Furthermore, the selective magnetic resonance spectroscopy fingerprinting sequence includes qT2. * -EPSFI Sequence: For the EPSFI sequence, the water signal acquisition module is configured to acquire water signals using planar gradient echo, and water signals with at least two different echo times are acquired to obtain qT2. * -EPSFI sequence; by fitting or matching water signals with at least two different echo times, the apparent transverse relaxation time T2 of the water signal can be obtained. * Figure. This sequence can simultaneously output the water longitudinal relaxation time T1 diagram, the excitation field inhomogeneity diagram B1 diagram, and the water surface apparent transverse relaxation time T2. * The graph includes two or more of the following: a longitudinal relaxation time (T1) graph of metabolites, a transverse relaxation time (T2) graph of metabolites, a metabolite concentration graph, and a metabolite relaxation-corrected concentration graph. The fitting method includes, but is not limited to, one or a combination of exponential decay fitting, dictionary matching, nonlinear least squares fitting, and linearized fitting. qT2 * - The EPSFI mode retains the ability of EPSFI to quantify multiple parameters of metabolites, while also increasing the ability to measure water T2. * Characterization of relevant tissue microenvironment information. This model is suitable for analyzing changes in the microenvironment of brain tumor regions, segmenting multimodal abnormal regions, and assessing spatial heterogeneity.

[0069] EPSFI sequence and qT2 * In the phase encoding module of the EPSFI sequence, the gradient distribution of the Y-axis and Z-axis is successively phase encoded.

[0070] Furthermore, the selective magnetic resonance spectroscopy fingerprinting sequence includes the BOLD-EPSFI sequence: For the EPSFI sequence, the water signal acquisition module is repositioned after the layer-selective excitation module and before the frequency-selective refocusing module, and configured to read out the water signal gradient via EPI; the phase encoding module is repositioned after the frequency-selective refocusing module and before the planar multi-echo metabolite readout module, and configured to perform successive phase encoding on the Y-axis gradient. In the BOLD-EPSFI sequence, the water signal acquisition module acquires time-series water signals related to functional blood oxygenation levels that repeat over time, while metabolite signal data is used to obtain metabolite concentration maps and / or metabolite relaxation parameter maps. First, B1 and water T1 are estimated using the initial fingerprint cycle, and then used for relaxation correction, emission field correction, and delay correction of the functional water signal; subsequently, the functional BOLD response map is obtained through stimulus paradigm correlation analysis, generalized linear models, or other functional analysis methods. Thus, metabolic maps and functional response maps can be obtained in the same spatial location and time frame.

[0071] Specifically, the basic units of the BOLD-EPSFI sequence include: a fat suppression module, which applies frequency-selective fat suppression pulses (selecting only the frequency range where fat is located) to reduce fat contamination of metabolite spectral lines; a layer-selective excitation module, which simultaneously applies a Z-axis gradient and a radio frequency pulse to excite the region of interest; a water signal acquisition module, which reads out the water signal through an EPI gradient; a frequency-selective re-convergence module, which applies two adiabatic pulses with variable center frequencies and uses a destructive gradient to selectively re-converge metabolite signals; a phase encoding module, which performs successive phase encoding on the Y-axis gradient; and a planar multi-echo metabolite readout module, which reads out the metabolite signal through a planar gradient echo. The water signal is a freely decaying signal acquired immediately after excitation using a planar gradient echo, while the metabolite signal is an echo signal acquired after re-convergence of metabolites within the frequency range of the selective re-convergence pulse (adiabatic pulse with variable center frequency) in the planar multi-echo readout gradient. The metabolite signal has an additional time dimension compared to the water signal, and the construction method of the basic modules and the loop structure allows both water and metabolite signals to be fingerprinted and spatially encoded.

[0072] In the frequency-selective refocusing module, disruptive gradients are used to reduce interference from water peaks, lipid peaks, or non-target frequency band signals on the spectrum of the target metabolite. Applying two adiabatic pulses with variable center frequencies, in conjunction with the disruptive gradients, specifically means applying a pair of disruptive gradients before and after the application of each adiabatic pulse at the variable center frequency; each pair of disruptive gradients includes one disruptive gradient applied along the X-axis and another simultaneously applied along the Y-axis. This module refocuses the metabolite signal using a pair of frequency-selective adiabatic pulses, excluding the water signal from the refocusing frequency band.

[0073] In practice, the center frequency and bandwidth of the frequency-selective re-aggregation module cover the frequency bands of all target metabolites.

[0074] Furthermore, the disruptive gradient can be replaced with modules such as fat suppression modules, water suppression modules, or water signal avoidance modules.

[0075] In the fat suppression module, frequency-selective saturation, inversion recovery, Dixon-like methods, or other fat suppression methods can be used.

[0076] The stratification excitation pulse can be a hard pulse, a sinc pulse, an SLR pulse, an adiabatic pulse, or other RF pulses that meet the stratification requirements.

[0077] Step S2: Optimize the acquisition parameters in the selective magnetic resonance spectroscopy fingerprinting sequence based on the relaxation characteristics of all target metabolites to obtain the optimized selective magnetic resonance spectroscopy fingerprinting sequence; the acquisition parameters include the repetition time TR, echo time TE, and excitation pulse flip angle FA in each basic unit within each cycle structure. For example... Figure 2 As shown in (A), TR refers to the time from the start to the end of each cycle basic module, TE refers to the time from the center of the excitation pulse to the center of the first gradient echo of the metabolite signal readout section, and FA refers to the reversal angle of the excitation pulse.

[0078] In step S2, a genetic algorithm is preferably used for optimization. During the optimization process, each individual is composed of the repetition time TR, echo time TE, and excitation pulse flip angle FA of all basic units within each cycle structure. The fitness function is set according to the following formula during the optimization process:

[0079] ;

[0080] ;

[0081] In the formula, For fitness, The mean root mean square error is given, and M is the number of target metabolites. The deviation between the true value and the estimated value. The variance of the estimated value is given by T, which is the relaxation time, representing the longitudinal relaxation time T1 or the transverse relaxation time T2, and k represents the sequence number of the target metabolite.

[0082] Furthermore, the optimization methods that can be used include, but are not limited to, one or a combination of grid search, Bayesian optimization, simulated annealing, and gradient optimization.

[0083] Furthermore, the optimization objectives may also include the estimation errors of metabolite T1, T2, concentration, B1, and water T2. * One or more of the following: estimation error, scan time, signal-to-noise ratio, and dictionary discrimination.

[0084] Step S3: Use the optimized selective magnetic resonance spectroscopy fingerprinting sequence to acquire water signals and metabolite composite signals, reconstruct the spatial fingerprint of the water signal based on the water signal, and reconstruct the spatial fingerprint spectrum of each metabolite based on the composite metabolite signal.

[0085] The dimensions of water signals include X-space dimension, Y-space dimension, fingerprint encoding dimension, coil channel dimension, and average dimension; the dimensions of metabolite signals include X-space dimension, Y-space dimension, time dimension, fingerprint encoding dimension, coil channel dimension, and average dimension.

[0086] In step S3, the water signal is subjected to phase correction, spatial filtering, spatial Fourier transform, and receiving coil channel merging to obtain the spatial fingerprint of the water signal; the metabolite signal is subjected to odd-even echo merging, spatial filtering, spatial Fourier transform, receiving coil channel merging, and time Fourier transform to obtain the spatial fingerprint spectrum of each metabolite.

[0087] Phase correction refers to correcting the first-order and zero-order phase errors of the odd and even rows in the X space using a reference signal, thereby aligning the odd and even rows of the K space phase code acquired by the plane gradient echo.

[0088] Spatial filtering refers to multiplying the data by the positions of each point in k-space as independent variables to generate a three-dimensional low-pass filter, thereby reducing voxel crosstalk caused by the point spread function.

[0089] Both water and metabolite signals require spatial Fourier transform to decode spatial information, while metabolite signals require time Fourier transform to decode spectral information.

[0090] Among them, receiving coil channel merging refers to merging the weights and phase diagrams of each receiving coil. The weights and phase diagrams of each receiving coil are obtained based on the amplitude and phase of the water signal that has undergone spatial Fourier transform in each receiving coil.

[0091] Step S4: Match the spatial fingerprint of the water signal with the water signal simulation dictionary to obtain the optimal parameter estimate of the water signal, and then obtain the parameter map of the water signal; match the spatial fingerprint spectrum of each metabolite with the metabolite simulation dictionary to obtain the optimal parameter estimate of the metabolite, and then obtain the parameter map of the metabolite.

[0092] Furthermore, the parameter maps of the water signal include the excitation field inhomogeneity B1 map and the water longitudinal relaxation time T. 1w Figure and apparent lateral relaxation time T2 * At least one of the figures shown.

[0093] Furthermore, the parameter maps of metabolites include at least one of the following: metabolite longitudinal relaxation time T1 map, metabolite transverse relaxation time T2 map, metabolite concentration map, and metabolite relaxation-corrected concentration map.

[0094] Step S4 includes:

[0095] Step S4.1: Use the spatial fingerprint of the water signal obtained in the first loop as the spatial fingerprint of the water signal, and match it with the water signal simulation dictionary to obtain the excitation field inhomogeneity B1 map and the water longitudinal relaxation time T. 1w picture.

[0096] Step S4.2: Fit the spatial fingerprint spectrum of each metabolite to obtain the spatial fingerprint of the metabolite. Using the excitation field inhomogeneity B1 diagram as a known condition, match the metabolite fingerprint with the metabolite simulation dictionary to obtain the metabolite longitudinal relaxation time T1 diagram and the metabolite transverse relaxation time T2 diagram.

[0097] Preferably, the amplitude of the spatial fingerprint spectrum of the target metabolite is obtained by fitting and used as the spatial fingerprint.

[0098] In this step, the fitting methods include, but are not limited to, LCModel, AMARES, QUEST, peak area fitting, basis function fitting, sparse spectrum fitting, deep learning spectrum fitting, or combinations thereof, with LCModel being the preferred method.

[0099] Step S4.3: Calculate the ratio between the spatial fingerprint of the metabolite and the spatial fingerprint of the water signal, and after relaxing and correcting the obtained ratio with the water concentration factor, obtain the absolute concentration map of the metabolite.

[0100] A simulation dictionary refers to a fingerprint set obtained through Bloch simulation based on known acquisition parameters and a set of parameters to be solved. Acquisition parameters include TR, TE, and FA, while parameters to be solved include longitudinal relaxation time T1, lateral relaxation time T2, and excitation field inhomogeneity B1. Specifically, two dictionaries need to be generated for the same sequence: a water signal simulation dictionary D1 and a metabolite simulation dictionary D2, used for fingerprint matching of water and metabolites, respectively. The sequence of the water signal simulation dictionary D1 is determined by TR and FA, and the parameters to be solved include the water longitudinal relaxation time T1. 1w And the excitation field inhomogeneity B1. The sequence of the metabolite simulation dictionary D2 is determined by TR, TE, and FA, and the parameters to be solved include the longitudinal relaxation time T1 and the transverse relaxation time T2. The metabolite simulation dictionary D2 is used as a known condition for simulation based on the excitation field inhomogeneity B1 estimated from the water signal. The dictionary formula is as follows:

[0101] ;

[0102] ;

[0103] in, This indicates a Bloch simulation.

[0104] Furthermore, the simulation method can also employ the equivalent signal evolution simulation method.

[0105] Preferably, a vector dot product matching method is used for matching. The specific process includes: first, taking the dot product of the fingerprint of each voxel of the water signal with the simulated water signal D1, and taking the parameter to be measured corresponding to the maximum value as the optimal parameter estimate. The excitation field inhomogeneity B1 map and the water longitudinal relaxation time T can be obtained immediately and simultaneously. 1w Figure 1. Using the non-uniformity of the excitation field B1 as a known condition for the metabolite simulation dictionary D2, and then taking the dot product of the fingerprint signal of each metabolite and the metabolite simulation dictionary D2, we can obtain the longitudinal relaxation time T1 map and the transverse relaxation time T2 map of each metabolite. Dividing the obtained corrected metabolite map by the corrected water signal map, we can obtain the absolute concentration map of each metabolite.

[0106] Furthermore, matching can be performed using any one or more combinations of vector dot product matching, normalized correlation matching, minimum distance matching, maximum likelihood estimation, low-rank subspace matching, sparse representation matching, machine learning regression, and deep learning inference.

[0107] Furthermore, when using the BOLD-EPSFI sequence for imaging, the method also includes step S5:

[0108] Step S5: Match the spatial fingerprint of the water signal with the water signal simulation dictionary to obtain the excitation field inhomogeneity B1 map; the parameter map of the water signal includes the water signal relaxation time; reconstruct the functional imaging map using the water signal relaxation time and the excitation field inhomogeneity B1 map.

[0109] The reconstruction process of functional imaging maps is as follows:

[0110] B1 plot and relaxation time plot T obtained from the water signal obtained in the first cycle 1w As a correction parameter, it corrects the water signal S measured in the current cycle. meas Then, the corrected water signal S corr Correlation analysis was performed with the stimulus paradigm to reconstruct the functional imaging map;

[0111] Corrected water signal S corr It is obtained through the following formula:

[0112] S corr =S meas / S sim

[0113] Among them, S corr S represents the corrected water signal for the current cycle. meas S represents the water signal measured in the current cycle. sim This represents the simulated signal obtained through Bloch simulation based on the correction parameters and the sequence parameters of the current loop;

[0114] Bloch simulation is represented as:

[0115] ;

[0116] Where TR, TE, and FA represent the values ​​of the repetition time, echo time, and excitation pulse flip angle in the current loop, respectively. 1w B1 and B1 represent the B1 diagram and relaxation time diagram matched with the water signal obtained in the first cycle, respectively.

[0117] Furthermore, the method of the present invention also includes at least one of the following: performing B0 drift correction, phase correction, frequency drift correction, motion correction, coil sensitivity estimation, coil merging weight estimation, or metabolite spectrum localization based on the water parameter map.

[0118] The target metabolites applicable to the method of the present invention include, but are not limited to, metabolites such as NAA, Cr, Cho, lactic acid, inositol, glutamate, glutamine, and lipids.

[0119] The method of this invention is applicable to obtaining multi-parameter metabolic profiles, functional profiles, microenvironment parameter profiles, or multimodal abnormal region profiles of the brain, tumor regions, muscles, liver, prostate, or other tissue regions for non-therapeutic purposes.

[0120] The method of this invention is also applicable to multimodal characterization of brain tumors, outputting NAA concentration, Cho concentration, Cr concentration, lactate concentration, inositol concentration, glutamate concentration, glutamine concentration, lipid concentration, metabolite T2, and water T2. * It uses at least two of the functional BOLD responses and generates anomaly region masks or spatial heterogeneity indices based on at least two parameters.

[0121] Specific embodiments of the present invention are as follows:

[0122] Example 1

[0123] This embodiment uses qT2. * -The EPSFI sequence underwent multi-parameter spectroscopy, such as Figure 1 and Figure 6 As shown, this embodiment includes the following steps:

[0124] Step S1: Construct a selective magnetic resonance spectroscopy fingerprint imaging sequence qT2 * -EPSFI.

[0125] In step S1, the development platform is a 3.0T Tesla Siemens Prisma magnetic resonance scanner. Experimental subjects include phantom experiments and human brain experiments. The phantom is a uniform spherical magnetic resonance spectroscopy brain phantom provided by General Electric, with target metabolites NAA, Cr, and Cho at labeled concentrations of 12.5 mM, 10 mM, and 3 mM, respectively. Subjects in the human experiments were divided into two groups: healthy individuals and patients, with three participants in each group.

[0126] Step S2: Optimize the acquisition parameters in the selective magnetic resonance spectroscopy fingerprinting sequence based on the relaxation characteristics of all target metabolites.

[0127] The process of step S2 is as follows:

[0128] Step S2.1: The loop structure consists of 5 basic units. The parameters to be optimized include 5 TR values, 5 TE values, and 5 FA values. The order cannot be changed arbitrarily. MATLAB's built-in genetic algorithm is used for optimization estimation to find the parameter combination that minimizes the objective function. The objective function is:

[0129] ;

[0130] ;

[0131] In the formula, For fitness, The mean root mean square error is given, and M is the number of target metabolites. The deviation between the true value and the estimated value. The variance of the estimated value is given by T, where T is the relaxation time, T1 represents the longitudinal relaxation time or T2 represents the transverse relaxation time, and k represents the sequence number of the target metabolite.

[0132] The optimization range of FA is limited to 30 to 90 degrees, TR is limited to 700 to 4000 milliseconds, and TE is limited to 70 to 180 milliseconds.

[0133] The optimization of the true relaxation time of different target metabolites varies. For example, in this case, the optimization targets for phantom measurements are: T1 and T2 values ​​of 756 and 508 ms for NAA, 381 and 273 ms for Cr, and 228 and 197 ms for Cho. In humans, the T1 and T2 values ​​are 1365 and 264 ms for NAA, 12400 and 152 ms for Cr, and 1300 and 207 ms for Cho.

[0134] The genetic algorithm uses a population size of 500 and a maximum number of generations of 150 to avoid local optima. The final sequence parameters are the optimal values ​​obtained from three iterations of the genetic algorithm.

[0135] Step S2.2: In vivo water and metabolite signals are acquired using selective magnetic resonance spectroscopy fingerprinting sequences, and optimized cyclic parameters are used to achieve fingerprint encoding of water and metabolites. The acquisition area covers the entire brain (≥210 mm × 210 mm), and selective inversion recovery pulses are used to suppress fat signals near the skull.

[0136] S3. Signal Reconstruction: Separate water and metabolite signals of different dimensions from the original data and process them separately to obtain the original water signal spectrum and the spectrum of each metabolite.

[0137] The process of step S3 is as follows:

[0138] Step S3.1: The first step in processing the metabolite signal is to merge the odd and even echoes. Since the odd and even echoes are collected in different directions in K-space, it is necessary to reverse the even echoes and perform first-order phase correction on the even echoes before merging the odd and even echoes.

[0139] Step S3.2: Filter the merged data by multiplying the K-space data by a Hamming window to achieve low-pass filtering, thereby suppressing the influence of the point spread function and reducing signal crosstalk between voxels.

[0140] Step S3.3: Perform a multidimensional fast Fourier transform along the spatial encoding dimension to restore the signal's location information.

[0141] Step S3.4: Obtain the amplitude and phase of each receiving coil at each voxel position by using the receiving coil sensitivity obtained by reference scanning, and combine the signals of 64 channels.

[0142] Step S3.5: Perform a fast Fourier transform on the time dimension to obtain the spectral signal at various points in space.

[0143] Step S3.6: To reconstruct the water signal acquired using planar gradient echo, after inverting the dual echo signal, first-order and zero-order phase corrections are performed by reference scanning to align the odd and even rows in K space. Then, spatial filtering is performed, and the spatial Fourier transform and receiving coil are combined to obtain the water signal at various locations in space that changes with time.

[0144] S4. Parameter Estimation: The reconstructed signal with attached parameter weights is fitted. The relaxation times and absolute concentrations of water and metabolites are obtained using dictionary matching.

[0145] The process of step S4 is as follows:

[0146] Step S4.1, Bloch's equations in the rotating coordinate system:

[0147] ;

[0148] ;

[0149] ;

[0150] This formula quantitatively describes the macroscopic magnetization vector after introducing a rotating coordinate system. Along Nucleation and relaxation processes T1 and T2 under the action of radio frequency field B1 in the directional direction, wherein, For radio frequency, For the precession frequency of Lamor, The initial macroscopic magnetization vector, It is the gyrometry.

[0151] Based on this, sequence simulation code was constructed to simulate the evolution of water and metabolite signals, and dictionaries D1 and D2 were generated for matching water and metabolite relaxation, respectively.

[0152] Step S4.2: Take the absolute value of the water signal and take the signal of the first fingerprint cycle (considered to be in a steady state) as the water fingerprint. Perform dot product matching between the water signal fingerprint and the dictionary D1 to obtain the excitation field inhomogeneity B1 map and the water longitudinal relaxation time T. 1w picture.

[0153] Step S4.3: Use the open-source software LCModel to fit the fingerprint spectrum of the metabolite to obtain the metabolite fingerprint. Using the B1 image as a known condition, match the metabolite fingerprint with the dot product of the dictionary D2 to obtain the longitudinal relaxation time T1 image and the transverse relaxation time T2 image of the metabolite.

[0154] Step S4.4: Calculate the metabolite fingerprint and the longitudinal relaxation time T of water. 1w The ratio of the graphs was used to obtain the absolute concentration graph of metabolites after relaxation and water concentration factor correction.

[0155] The table below shows the absolute concentrations of NAA, Cr, and Cho, as well as T1 and T2 (mean ± standard deviation) in the gray matter, white matter, and tumor regions of the images obtained in this embodiment.

[0156]

[0157] Figure 4 and Figure 5 Specifically, in this embodiment, that is, in qT2 * - A schematic diagram of the multimodal output under the EPSFI sequence, showing the longitudinal relaxation time T1 and the apparent transverse relaxation time T2. * Figures: B1 diagram of excitation field inhomogeneity, metabolite concentration diagram, and metabolite relaxation diagram (T1 and T2).

[0158] Example 2

[0159] This embodiment uses the BOLD-EPSFI sequence for functional imaging. The water signal acquisition module uses EPI to read out the gradient, and a single acquisition can obtain a water signal image.

[0160] When imaging is performed using a BOLD-EPSFI sequence, the method further includes step S4.5:

[0161] Step S4.5: Functional graph imaging, based on the matching results of the water fingerprint from the first cycle (water longitudinal relaxation time T). 1w (Figure), correcting the longitudinal relaxation time T of water obtained from other cycles. 1w The brain activation map can be obtained by calculating the correlation between signals and stimulus paradigms.

[0162] Figure 7 This is a schematic diagram of the functional output of this embodiment, showing the function BOLD response diagram.

[0163] The above description is only a preferred embodiment of the present invention. Therefore, all equivalent changes or modifications made to the structure, features and principles described in the claims of this patent application are included in the scope of this patent application.

Claims

1. A rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting, characterized in that, Includes the following steps: Step S1: Construct a selective magnetic resonance spectroscopy fingerprint imaging sequence; the selective magnetic resonance spectroscopy fingerprint imaging sequence includes at least one cyclic structure, which is mainly composed of multiple time-series cascaded basic units; In the selective magnetic resonance spectroscopy fingerprint imaging sequence, fingerprint encoding is achieved by changing the repetition time TR, echo time TE, and excitation pulse flip angle FA in each basic unit within the cyclic structure. Step S2: Optimize the acquisition parameters in the selective magnetic resonance spectroscopy fingerprinting sequence based on the relaxation characteristics of all target metabolites to obtain an optimized selective magnetic resonance spectroscopy fingerprinting sequence; the acquisition parameters include the repetition time TR, echo time TE, and excitation pulse flip angle FA in each basic unit within each cyclic structure; Step S3: Use the optimized selective magnetic resonance spectroscopy fingerprinting sequence to acquire water signals and metabolite composite signals, reconstruct the spatial fingerprint of the water signal based on the water signal, and reconstruct the spatial fingerprint spectrum of each metabolite based on the composite metabolite signal. Step S4: Match the spatial fingerprint of the water signal with the water signal simulation dictionary to obtain the parameter map of the water signal; match the spatial fingerprint spectrum of each metabolite with the metabolite simulation dictionary to obtain the parameter map of the metabolite.

2. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 1, characterized in that: The selective magnetic resonance spectroscopy fingerprinting sequence includes an EPSFI sequence, and the basic unit of the EPSFI sequence includes, in sequence: The fat inhibition module applies frequency-selective fat inhibition pulses; The layer-selective excitation module simultaneously applies a Z-axis gradient and a radio frequency pulse to excite the region of interest. The phase encoding module performs spatial phase encoding; A water signal acquisition module is used to acquire water signals. The frequency-selective refocusing module applies two adiabatic pulses with variable center frequencies, in conjunction with gradient destruction. The planar multi-echo metabolite readout module reads out metabolite signals through planar gradient echo.

3. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 2, characterized in that: The selective magnetic resonance spectroscopy fingerprinting sequence includes qT2. * -EPSFI sequence: For the EPSFI sequence, the water signal acquisition module is configured to acquire water signals using planar gradient echo, and acquires water signals with at least two different echo times to obtain qT2. * -EPSFI sequence; By fitting or matching the water signals with at least two different echo times, the T2 of the water signal can be obtained. * picture.

4. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 2, characterized in that: The selective magnetic resonance spectral fingerprinting sequence includes the BOLD-EPSFI sequence: For the EPSFI sequence, the water signal acquisition module is changed to be set after the layer-selective excitation module and before the frequency-selective refocusing module, and is set to read out the water signal gradient through EPI; the phase encoding module is changed to be set after the frequency-selective refocusing module and before the planar multi-echo metabolite readout module, and is set to perform successive phase encoding on the Y-axis gradient. When imaging using the BOLD-EPSFI sequence, the method further includes: The spatial fingerprint of the water signal is matched with the water signal simulation dictionary to obtain the excitation field inhomogeneity B1 map; the parameter map of the water signal includes the water signal relaxation time; the functional imaging map is reconstructed using the water signal relaxation time and the excitation field inhomogeneity B1 map.

5. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 4, characterized in that: The reconstruction process of the functional imaging map is specifically as follows: the B1 map and relaxation time map T are matched with the water signal obtained in the first cycle. 1w As a correction parameter, it corrects the water signal S measured in the current cycle. meas Then, the corrected water signal S corr Correlation analysis was performed with the stimulus paradigm to reconstruct the functional imaging map; Corrected water signal S corr It is obtained through the following formula: S corr =S meas / S sim ; Among them, S corr S represents the corrected water signal of the current cycle. meas S represents the water signal measured in the current cycle. sim This represents the simulated signal obtained through Bloch simulation based on the correction parameters and the sequence parameters of the current loop; The Bloch simulation is represented as follows: ； Where TR, TE, and FA represent the values ​​of the repetition time, echo time, and excitation pulse flip angle in the current loop, respectively. 1w B1 and B1 represent the B1 diagram and relaxation time diagram matched with the water signal obtained in the first cycle, respectively.

6. The fast multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 1, characterized in that: The parameter diagram of the water signal includes the excitation field inhomogeneity B1 diagram and the water longitudinal relaxation time T. 1w Figure and apparent lateral relaxation time T2 * At least one of the figures; The parameter maps of the metabolites include at least one of the following: metabolite longitudinal relaxation time T1 map, metabolite transverse relaxation time T2 map, metabolite concentration map, and metabolite relaxation-corrected concentration map.

7. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 1, characterized in that: The dimensions of the water signal include X-space dimension, Y-space dimension, fingerprint encoding dimension, coil channel dimension, and average dimension; The dimensions of the metabolite signal include X-space dimension, Y-space dimension, time dimension, fingerprint encoding dimension, coil channel dimension, and average dimension.

8. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 1, characterized in that: In step S2, a genetic algorithm is used for optimization; During the optimization process, the fitness function is set according to the following formula: ； ； In the formula, For fitness, The mean root mean square error is given, and M is the number of target metabolites. The deviation between the true value and the estimated value. The variance of the estimated value is given by T, which is the relaxation time, representing the longitudinal relaxation time T1 or the transverse relaxation time T2, and k represents the sequence number of the target metabolite. During the optimization process, each individual is composed of the repetition time TR, echo time TE, and excitation pulse flip angle FA of all basic units within each loop structure.

9. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 1, characterized in that: In step S3, the water signal undergoes phase correction, spatial filtering, spatial Fourier transform, and receiving coil channel merging to obtain the spatial fingerprint of the water signal. The metabolite signals were subjected to odd-even echo merging, spatial filtering, spatial Fourier transform, receiving coil channel merging, and time Fourier transform to obtain the spatial fingerprint spectrum of each metabolite.

10. The rapid multi-parameter spectral and functional imaging method based on magnetic resonance fingerprinting according to claim 1, characterized in that: Step S4 includes: Step S4.1: Use the spatial fingerprint of the water signal obtained in the first loop as the spatial fingerprint of the water signal, and match it with the water signal simulation dictionary to obtain the excitation field inhomogeneity B1 map and the water longitudinal relaxation time T. 1w picture; Step S4.2: Fit the spatial fingerprint spectrum of each metabolite to obtain the spatial fingerprint of the metabolite. Using the excitation field inhomogeneity B1 diagram as a known condition, match the metabolite fingerprint with the metabolite simulation dictionary to obtain the metabolite longitudinal relaxation time T1 diagram and the metabolite transverse relaxation time T2 diagram. Step S4.3: Calculate the ratio between the spatial fingerprint of the metabolite and the spatial fingerprint of the water signal, and after relaxation and water concentration factor correction, obtain the absolute concentration map of the metabolite.

Images