A b0 robust cardiac t1 quantification imaging method

Abstract

The application discloses a B0 robust cardiac T1 quantitative imaging method, and belongs to the field of magnetic resonance quantitative imaging, and comprises the following steps: designing a dual-phase increment cardiac T1 quantitative imaging acquisition sequence, collecting original data based on the dual-phase increment cardiac T1 quantitative imaging acquisition sequence, constructing a simulation signal dictionary based on a magnetic resonance physical model, matching the original data with the simulation signal dictionary to obtain the best dictionary entry corresponding to the original data, and obtaining a T1 quantitative image and a B0 quantitative image based on the best dictionary entry. The application encodes B0 information by using a dual-phase increment sequence design, realizes B0 robust T1 quantitative imaging, reduces the influence of B0 field inhomogeneity under high field strength, and reduces bSSFP black band artifact interference. In combination with the dictionary matching technology, T1 and B0 are jointly estimated, more accurate T1 estimation and B0 quantitative images are provided, the method is easy to deploy and apply on an existing MRI platform, and has high implementation and migration feasibility.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of magnetic resonance quantitative imaging, and particularly relates to a B0 robust cardiac T1 quantitative imaging method. BACKGROUND

[0002] Magnetic resonance quantitative imaging technology provides objectivity and repeatability beyond traditional magnetic resonance weighted imaging by directly measuring the intrinsic physical properties of tissues, which mainly relies on the relative difference of image signal intensity to indirectly reflect the tissue contrast. Cardiac magnetic resonance T1 quantitative imaging is one of the magnetic resonance quantitative imaging technologies, which is a technology capable of accurately measuring the longitudinal relaxation time of myocardial tissue. The T1 value of myocardial tissue is its inherent physical property, which can reflect the changes of physical properties such as tissue water content and macromolecular environment. In the field of magnetic resonance imaging, accurate quantification of T1 value has important significance for objectively characterizing the physical state of the tissue. For example, the T1 value before administration can be used as a basic index of the physical properties of the tissue; the change of T1 value before and after administration can indirectly reflect the physical properties of the extracellular space (such as extracellular volume fraction ECV).

[0003] The accurate measurement of cardiac T1 relaxation time is usually based on the principle of the classic Look-Locker sequence, the core of which is to apply a single inversion pulse and then quickly and continuously sample the signal in the longitudinal magnetization recovery process at multiple different effective inversion time (TI) points, so as to be able to depict the T1 recovery curve. Modified Look-Locker inversion recovery (MOLLI) is one of the most widely used and mature technical solutions of this principle in clinical cardiac T1 quantitative imaging. The MOLLI sequence (for example, a 5-(3)-3 scheme using 5 cardiac cycles for acquisition, 3 cardiac cycles for recovery, and 3 cardiac cycles for acquisition again) uses balanced steady-state free precession (bSSFP) as the signal readout module. Due to its high signal-to-noise ratio and high acquisition efficiency characteristics, bSSFP can support fast image data acquisition at each TI point, thereby effectively obtaining multiple sampling points on the T1 recovery curve within a limited time.

[0004] However, despite the significant advantages of bSSFP-based MOLLI sequences in acquisition speed and signal strength, their T1 quantification accuracy is severely limited by the inhomogeneity of the main magnetic field (B0). Due to its physiological characteristics, such as organ pulsation, respiratory movements, intracavitary blood flow, and significant differences in magnetic susceptibility between the cardiac region and surrounding tissues (e.g., the air-filled lungs, liver, etc.), the cardiac region often exhibits an unavoidable and spatially uneven B0 field shift. The bSSFP sequence itself is a special type of fast gradient echo sequence, its core technical feature being gradient balance, where the total gradient moment of all applied gradient pulses (including slice selection, phase encoding, and readout gradients) is zero in all directions within a repetition time (TR). This allows the sequence to reach and maintain a magnetization steady state. However, it is precisely this extremely high sensitivity to phase and the coherent superposition of the steady state that makes the bSSFP signal extremely sensitive to B0 field shifts. When local B0 field shifts result in an accumulated distance of approximately 180 degrees from the resonance phase within a TR (for the commonly used 180° phase cycle), the bSSFP signal undergoes severe destructive interference, leading to a sharp decrease or even complete disappearance of signal intensity. This results in characteristic banded signal loss on the image, known as "banding artifacts." When bSSFP is used for data acquisition in MOLLI sequences, these banding artifacts directly contaminate the raw image data used for subsequent T1 fitting, causing severe distortion or complete failure of T1 value measurements in the artifact region and its vicinity. This significantly limits the accuracy and clinical reliability of bSSFP-based T1 quantification techniques such as MOLLI in regions of B0 field inhomogeneity.

[0005] Therefore, how to effectively overcome the inhomogeneity of the B0 field, especially to reduce or eliminate the negative impact of the black band artifact introduced when using the MOLLI and other T1 quantitative techniques with bSSFP readout, in order to obtain robust cardiac T1 quantitative maps under inhomogeneous B0 fields, remains a key technical problem that urgently needs to be solved in the field of magnetic resonance imaging. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a B0-robust quantitative cardiac T1 imaging method to resolve the issues present in the prior art.

[0007] To achieve the above objectives, the present invention provides a B0-robust quantitative cardiac T1 imaging method, comprising:

[0008] Design a dual-phase increment cardiac T1 quantitative imaging acquisition sequence, and acquire raw data based on the dual-phase increment cardiac T1 quantitative imaging acquisition sequence;

[0009] The dual-phase incremental cardiac T1 quantitative imaging acquisition sequence uses an ECG-gated method. In the first cardiac cycle of the first round of five cardiac cycles and the first cardiac cycle of the second round of five cardiac cycles, an adiabatic inversion pulse is applied to invert the longitudinal magnetization vector. No adiabatic inversion pulse is applied in the other four cardiac cycles of the current round. After the acquisition of the first round of five cardiac cycles is completed, the acquisition of the signal for the second round of five cardiac cycles is paused for three cardiac cycles.

[0010] A simulation signal dictionary was constructed based on a magnetic resonance physical model.

[0011] The original data is matched with the simulation signal dictionary to obtain the optimal dictionary entry corresponding to the original data, and the T1 quantitative map and B0 quantitative map are obtained based on the optimal dictionary entry.

[0012] Optionally, after each cardiac cycle gating trigger and a waiting period, signal acquisition is performed. After entering steady state using pre-pulses with five increasing flip angles, data acquisition is performed using radio frequency pulses with a fixed flip angle.

[0013] Optionally, the waiting time is calculated based on the preset reversal time, and the waiting time for the first five cardiac cycles is less than that for the second five cardiac cycles. The signal acquisition is performed while the subject is holding its breath.

[0014] Optionally, during signal acquisition, the layer selection gradient is applied synchronously with the radio frequency pulse, followed by the application of the phase encoding gradient and the readout gradient; the layer selection gradient applies a pre-coalescing phase gradient before the application of the radio frequency pulse and a recoalescing phase gradient after the application of the radio frequency pulse; the readout gradient applies a pre-coalescing phase gradient before signal acquisition and a recoalescing phase gradient after data acquisition is completed; the phase encoding gradient is applied before signal acquisition and a recoalescing phase gradient of equal magnitude and opposite polarity is applied after signal acquisition is completed.

[0015] Optionally, the phase increment of the radiofrequency pulses in the first five cardiac cycles and the second five cardiac cycles can be set to different values.

[0016] Optionally, the process of constructing a simulation signal dictionary based on the magnetic resonance physical model includes:

[0017] The parameter combination design is determined, and signal simulation is performed based on the parameter combination design, the magnetic resonance physical model, and the phase distribution map magnetic resonance simulation tool to obtain the set of simulation signals under different parameter combinations; a simulation signal dictionary is constructed based on the set of simulation signals; wherein, the parameter combination design includes parameter composition and parameter range, and the parameter composition includes longitudinal relaxation time, transverse relaxation time, and B0 field offset.

[0018] Optionally, the process of obtaining the optimal dictionary entry corresponding to the original data includes:

[0019] Image reconstruction is performed on the original data to obtain multi-contrast images corresponding to each cardiac cycle. The amplitude of the complex signal of each pixel in the multi-contrast images is extracted as the real signal. The inner product of the real signal of each pixel with each dictionary signal in the simulation signal dictionary is calculated. For each pixel, the dictionary entry that maximizes the inner product of the corresponding real signal and the dictionary signal is found as the optimal dictionary entry. The inner product calculation is performed after normalizing the real signal and all dictionary signals in the simulation signal dictionary.

[0020] The present invention provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0021] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0022] Compared with the prior art, the present invention has the following advantages and technical effects:

[0023] This invention achieves robust T1 quantitative imaging of B0 by encoding B0 information through a dual-phase increment sequence design, reduces the influence of B0 field inhomogeneity on T1 quantitative imaging under high field strength, and mitigates the interference of bSSFP black band artifacts.

[0024] This invention uses dictionary matching technology to jointly estimate T1 and B0. Compared with traditional fitting methods, dictionary matching can better handle B0 information in the sequence to provide more accurate T1 estimation. It can also generate a quantitative B0 map at the same time to provide more comprehensive quantitative information.

[0025] This invention is an improvement on existing mature sequences and combines dictionary matching technology. It is technically easy to deploy and apply on existing MRI platforms, and has high feasibility for implementation and migration. Attached Figure Description

[0026] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0027] Figure 1 This is a flowchart illustrating the implementation of the B0 robust cardiac T1 quantitative imaging method according to an embodiment of the present invention.

[0028] Figure 2 This is a schematic diagram of a dual-phase increment cardiac T1 quantitative imaging sequence according to an embodiment of the present invention;

[0029] Figure 3This is a schematic diagram of the simulation dictionary generation and T1 and B0 quantitative graph reconstruction based on dictionary matching in an embodiment of the present invention.

[0030] Figure 4 This is a schematic diagram of the phantom imaging results and the T1 and B0 quantitative map reconstruction results according to an embodiment of the present invention;

[0031] Figure 5 This is a schematic diagram of cardiac imaging results of volunteers after actual time-reversal rearrangement, according to an embodiment of the present invention.

[0032] Figure 6 This is a schematic diagram of the reconstruction results of the standard short-axis T1 and B0 quantitative images of the heart of volunteers according to an embodiment of the present invention;

[0033] Figure 7 This is a standard short-axis basal T1 quantitative image of a volunteer's heart and a partially enlarged schematic diagram, according to an embodiment of the present invention. Detailed Implementation

[0034] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0035] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0036] Example 1

[0037] like Figure 1 As shown, this embodiment provides a B0 (main magnetic field) robust quantitative imaging method for cardiac T1 (longitudinal relaxation time), including:

[0038] Step S1: As Figure 2 As shown, a dual-phase increment B0 robust cardiac T1 quantitative imaging acquisition sequence is constructed, and XCAT (4D Digital Extended Cardiac-Torso) digital cardiac phantom simulation is performed based on the sequence:

[0039] The XCAT digital heart phantom has accurate and diverse cardiac anatomy. In this embodiment, the T1 quantitative map of the intermediate layer of the heart generated by XCAT is used as the target image, and a variable B0 quantitative map is defined. These images are input into the simulation tool along with the sequence to obtain 10 simulated heart images with B0 non-uniformity and multiple contrasts.

[0040] The sequence parameters were set according to in vivo quantitative cardiac imaging. The field of view (FOV) in the readout direction was set to 340 mm, the FOV in the phase encoding direction was set to 300 mm, the slice thickness was 8 mm, the voxel size was 2.21 × 1.77 × 8 cubic mm, the readout direction resolution was 192, the phase encoding direction resolution was 80%, the repetition time / echo time (TR / TE) was 2.86 ms / 1.45 ms, and the flip angle of the bSSFP (balanced steady-state free precession) readout was 35°. The phase cycle increment 1 of the bSSFP was set to 180°, and the phase increment 2 was set to 270°. The inversion time 1 of the first five cardiac cycles of the adiabatic inversion pulse was set to 150 ms, and the inversion time 2 of the second five cardiac cycles was set to 250 ms. The R-wave interval of the cardiac cycle in the simulation was set to a fixed 1000 ms.

[0041] The sequence acquisition was performed using ECG gating for 13 cardiac cycles, following a 5-(3)-5 acquisition scheme: 5 cardiac cycles were acquired first, followed by a pause, a 3-cycle rest, and then another 5 cardiac cycles. After an R wave in the ECG, and following a certain trigger delay, data acquisition was performed during mid-diastole while the subject held their breath. In the first of the five acquired cardiac cycles, a 180° adiabatic inversion pulse was applied to invert the longitudinal magnetization vector; this was not applied in the other four cardiac cycles. The waiting time was calculated using the set inversion time. After the waiting time, the sequence was readout using bSSFP for data acquisition. A pre-pulse with 5 inverted angles was used to enter steady state, during which no data was acquired. Then, a radiofrequency pulse with a fixed inversion angle was used for data acquisition, resulting in a total of 71 lines. Within one TR cycle, the slice selection gradient was applied synchronously with the radiofrequency pulse to select the excitation slice, followed by the application of the phase encoding gradient and the readout gradient. To ensure complete balance of the gradient magnetic field within a TR cycle, i.e., the total gradient moments on all three logic axes are zero, the layer selection gradient applies a pre-converging gradient before the RF pulse is applied and a reconverging gradient after the RF pulse is applied; the readout gradient applies a pre-converging gradient before data acquisition and a reconverging gradient after data acquisition; the phase encoding gradient applies before acquisition and a reconverging gradient of equal magnitude but opposite polarity after acquisition. In the next TR cycle, the phase of the RF pulse increases by a phase increment, for example, phase increment 1 is 180° and phase increment 2 is 270°. In the first 5 acquired cardiac cycles, the phase of the RF pulse cycles from 0° to 180° to 0°; in the subsequent 5 acquired cardiac cycles, the phase of the RF pulse cycles from 0° to 270° to 180° to 90° to 0°, achieving two phase cycles in one sequence, thereby encoding B0 information.

[0042] Step S2: The process of constructing the simulation signal dictionary is as follows:

[0043] To perform parameter estimation, a simulation signal dictionary was constructed. The signal entries in this dictionary are all based on the pulse sequence parameters defined in step S1, including all parameters of the sequence, such as the pattern of 5-(3)-5, inversion time, the number and duration of radiofrequency pulses and gradient magnetic fields, etc. A pulseq sequence with the same parameters was written using pypulseq to facilitate subsequent simulations and possible migration of the simulation platform in the future. The dictionary simulation is based on the phase distribution map (PDG) magnetic resonance simulation tool of the open-source MRzeroCore platform, and an improved simulate function is used to generate efficient parallel simulations. The accuracy parameter of the function is set to 0.01 to balance simulation accuracy and simulation efficiency. An optimized simulation signal dictionary was constructed for the T1 and T2 (transverse relaxation time) ranges of the XCAT digital phantom and the scanning parameters used. The ranges and steps of the dictionary parameters T1, T2, B0 offset, and the optional radiofrequency magnetic field B1 factor were carefully selected to ensure accurate characterization and efficient matching of in-vivo signals. The parameter dimensions of the dictionary and its discretized sampling ranges and steps are set as follows: the T1 relaxation time range is from 200 milliseconds to 1250 milliseconds, with a step of 1 millisecond; the B0 field offset range is from -170 Hz to +170 Hz, which is calculated based on the repetition time of the sequence used, corresponding to -π to π in radians, with a step of 10 Hz. In this embodiment, since no encoding modules sensitive to T2 and B1 are introduced in the sequence, to reduce the dictionary size, the T2 relaxation time is fixed at 100 milliseconds and the B1 field scaling factor is fixed at 1.0 (i.e., the ideal B1 field). The dictionary entries are combinations of these parameters (ensuring T2 < T1, which conforms to the inherent physical constraints). During simulation, each parameter combination is regarded as an independent voxel, the proton density (PD) is set to 1, the apparent transverse relaxation time T2 * is set to infinity, and the diffusion coefficient is set to 0. The simulation function is called in a batch processing manner. Depending on the different accuracy parameters, the maximum number of voxel parameter combinations processed in each batch is different. In this embodiment, 484 voxels are processed in each batch. The total signal corresponding to 10 acquisition segments is obtained by simulation for each parameter combination. Subsequently, the average value of the amplitudes of the simulated complex signals within each acquisition segment is extracted to form a signal fingerprint with a length of 10 data points. The signal fingerprints generated by all parameter combinations constitute the simulation signal dictionary.

[0044] Step S3: Use the image obtained from the XCAT simulation as the real signal to match with the simulation signal dictionary, confirm the best dictionary entry pixel by pixel to jointly estimate the T1 and B0 parameter maps, and perform result analysis. The process is as follows:

[0045] Multi-contrast cardiac images obtained from XCAT simulations are used as ground truth signals, and pixel-by-pixel matching is performed with a signal dictionary. First, the ground truth fingerprints loaded and preprocessed in step S1 and the dictionary signal fingerprints generated in step S2 are normalized using the L2 norm. Then, the inner product of the normalized ground truth fingerprint matrix and the normalized signal dictionary matrix is ​​calculated using matrix multiplication to obtain the similarity score between each ground truth fingerprint and all entries in the dictionary. For each pixel, i.e., each ground truth fingerprint, the dictionary entry with the highest similarity score is selected as the best match. Based on the index of the best-matching dictionary entry, the corresponding T1 value and B0 field offset value are found from a pre-stored list of parameter combinations, thus generating the T1 quantitative map and the B0 quantitative map. The generated parameter maps can be saved as .mat files and other formats for visualization. Furthermore, to evaluate the matching quality, the root mean square error (RMSE) between the normalized ground truth fingerprint of each pixel and its best-matched normalized dictionary signal fingerprint is calculated, and an RMSE error map is generated. The RMSE map can visually reflect the goodness of fit of the match.

[0046] The T1 and B0 quantitative maps generated in this embodiment were observed and found to be the same as the simulated XCAT heart T1 and preset B0 quantitative maps. The digital phantom simulation results show that the method of the present invention can estimate the T1 value from the B0 modulated bSSFP signal under the condition of B0 field inhomogeneity, in terms of algorithm and sequence design principle.

[0047] Example 2

[0048] This embodiment provides a B0-robust method for quantitative T1 imaging of the heart, including:

[0049] Step S1: As Figure 2 As shown, a dual-phase increment B0 robust cardiac T1 quantitative imaging acquisition sequence was constructed, and phantom data was acquired based on the sequence using the following acquisition method:

[0050] A series of physical phantoms were prepared using nickel chloride and agarose at the following concentrations: 0.5 mmol / L nickel chloride and 0.5% agarose; 0.625 mmol / L nickel chloride and 0.3% agarose; 0.75 mmol / L nickel chloride and 1% agarose; 0.9 mmol / L nickel chloride and 1.5% agarose; 1 mmol / L nickel chloride and 2% agarose; 1 mmol / L nickel chloride and 2.5% agarose; 1 mmol / L nickel chloride and 3% agarose; 1 mmol / L nickel chloride hexahydrate and 2% agarose; and 2 mmol / L nickel chloride hexahydrate and 2% agarose. Nine gel phantoms with different T1 and T2 relaxation characteristics were used. After spin-echo sequencing, their T1 and T2 relaxation times at 3.0T were: 1640 ms, 209 ms; 1327 ms, 333 ms; 1155 ms, 118 ms; 1052 ms, 82 ms; 929 ms, 64 ms; 936 ms, 50 ms; 752 ms, 34 ms; 636 ms, 57 ms; and 380 ms, 51 ms. During scanning, the phantoms were placed in a container and subjected to a water bath to optimize imaging conditions. The B0 robust cardiac T1 quantitative imaging sequence described in this invention was implemented on a United Imaging UMR 790 3.0T MRI scanner. The ECG signal was simulated using the virtual physiological signal function with a period of 1000 milliseconds.

[0051] The sequence parameters were set for quantitative imaging of the physical phantom. The readout direction field of view (FOV) was set to 340 mm, the phase encoding direction FOV to 300 mm, the slice thickness to 8 mm, the voxel size to 2.21 × 1.77 × 8 cubic mm, the readout direction resolution to 192, the phase encoding direction resolution to 80%, the repetition time / echo time (TR / TE) to 2.86 ms / 1.45 ms, the acquisition bandwidth to 1400 Hz per pixel, and the bSSFP readout flip angle to 35°. The bSSFP phase cycle increment 1 was set to 180°, and the phase increment 2 to 270°. The adiabatic inversion pulse inversion time 1 was set to 150 ms, and the inversion time 2 to 250 ms. Parallel imaging was used for double acceleration, with 24 phase reference lines and partial phase encoding enabled. Volumetric shimming was used.

[0052] The sequence acquisition was performed using ECG gating for 13 cardiac cycles, following a 5-(3)-5 acquisition scheme: 5 cardiac cycles were acquired first, followed by a pause, a 3-cycle rest, and then another 5 cardiac cycles. After an R wave in the ECG, and following a certain trigger delay, data acquisition was performed during mid-diastole while the subject held their breath. In the first of the five acquired cardiac cycles, a 180° adiabatic inversion pulse was applied to invert the longitudinal magnetization vector; this was not applied in the other four cardiac cycles. The waiting time was calculated using the set inversion time. After the waiting time, the sequence was readout using bSSFP for data acquisition. A pre-pulse with 5 inverted angles was used to enter steady state, during which no data was acquired. Then, a radiofrequency pulse with a fixed inversion angle was used for data acquisition, resulting in a total of 71 lines. Within one TR cycle, the slice selection gradient was applied synchronously with the radiofrequency pulse to select the excitation slice, followed by the application of the phase encoding gradient and the readout gradient. To ensure complete balance of the gradient magnetic field within a TR cycle, i.e., the total gradient moments on all three logic axes are zero, the layer selection gradient applies a pre-converging gradient before the RF pulse is applied and a reconverging gradient after the RF pulse is applied; the readout gradient applies a pre-converging gradient before data acquisition and a reconverging gradient after data acquisition; the phase encoding gradient applies before acquisition and a reconverging gradient of equal magnitude but opposite polarity after acquisition. In the next TR cycle, the phase of the RF pulse increases by a phase increment, for example, phase increment 1 is 180° and phase increment 2 is 270°. In the first 5 acquired cardiac cycles, the phase of the RF pulse cycles from 0° to 180° to 0°; in the subsequent 5 acquired cardiac cycles, the phase of the RF pulse cycles from 0° to 270° to 180° to 90° to 0°, achieving two phase cycles in one sequence, thereby encoding B0 information.

[0053] MATLAB was used to reconstruct the acquired k-space raw data. After reading the raw data and protocol parameters, each layer was processed independently. Based on the actual number of phase-encoded rows and readout points in the protocol parameters, the raw k-space data was padded with zeros to match the actual acquisition resolution. Based on the timestamp information and the preset reversal time, the actual reversal time and corresponding index for each acquisition time were calculated. The image order was rearranged according to the actual reversal time, and the R-wave interval of the cardiac cycle was calculated based on the timestamps of adjacent center k-spaces. The ESPIRiT (parallel imaging using eigenvector maps) method was used for coil sensitivity estimation, and the GRAPPA (GeneRalized Autocalibrating Partially Parallel Acquisition) algorithm was used for image reconstruction, ultimately yielding the following results:Figure 4 Ten multi-contrast complex images with different inversion times as shown.

[0054] Step S2: The process of constructing the simulation signal dictionary is as follows:

[0055] For parameter estimation, a multi-dimensional simulation signal dictionary is constructed. The signal entries of this dictionary are all based on the pulse sequence parameters defined in Step S1. A pulseq sequence with the same parameters is written using pypulseq to facilitate subsequent simulations and potential migration to a simulation platform in the future. The dictionary simulation is based on the phase distribution graph (PDG) magnetic resonance simulation tool of the open-source MRzeroCore platform. An improved simulate function is used to generate efficient parallel simulations. The accuracy parameter of the function is set to 1e-2 to balance simulation accuracy and simulation efficiency. An optimized simulation signal dictionary is constructed for the T1 and T2 ranges of the physical phantom and the scanning parameters used. The ranges and steps of the dictionary parameters T1, T2, B0 offset, and the optional B1 factor are carefully selected to ensure accurate characterization and efficient matching of in-vivo signals. The parameter dimensions of the dictionary and its discretized sampling ranges and steps are set as follows: The T1 relaxation time range is from 300 milliseconds to 2000 milliseconds, with a step of 5 milliseconds; the B0 field offset range is from -175 Hz to +175 Hz. This range is calculated based on the repetition time of the sequence used and corresponds to -π to π in radians, with a step of 5 Hz. In this embodiment, since no encoding modules sensitive to T2 and B1 are introduced in the sequence, to reduce the dictionary size, the T2 relaxation time is fixed at 200 milliseconds, and the B1 field scaling factor is fixed at 1.0 (i.e., an ideal B1 field). The dictionary entries are combinations of these parameters (ensuring T2 < T1, which conforms to the inherent physical constraints). During simulation, each parameter combination is regarded as an independent voxel, the proton density (PD) is set to 1, T2* is set to infinity, and the diffusion coefficient is set to 0. The simulation function is called in a batch processing manner. Depending on the different accuracy parameters, the maximum number of voxel parameter combinations processed in each batch is different. In this embodiment, 484 voxels are processed in each batch. The total signal corresponding to 10 acquisition segments is obtained by simulation for each parameter combination. Subsequently, the average value of the amplitudes of the simulated complex signals within each acquisition segment is extracted to form a signal fingerprint with a length of 10 data points. The signal fingerprints generated by all parameter combinations constitute the original signal dictionary.

[0056] Step S3: Match the real signal of the collected physical phantom with the simulation signal dictionary, confirm the best dictionary entry pixel by pixel to jointly estimate the T1 and B0 parameter maps, and perform result analysis. The process is as follows:

[0057] The acquired and reconstructed multi-contrast image data of the physical phantom is used as the real signal, and it is matched pixel by pixel with the signal dictionary. First, the real signal fingerprint loaded and preprocessed in step S1 and the dictionary signal fingerprint generated in step S2 are normalized using the L2 norm. Then, the inner product of the normalized real signal fingerprint matrix and the normalized signal dictionary matrix is ​​calculated by matrix multiplication to obtain the similarity score between each real signal fingerprint and all entries in the dictionary. For each pixel, i.e., each real signal fingerprint, the dictionary entry with the highest similarity score is selected as the best match. Based on the index of the best matching dictionary entry, the corresponding T1 value and B0 field offset value are found from the pre-stored parameter combination list, thereby generating the T1 quantitative map and the B0 quantitative map, as shown below. Figure 4 As shown, the high and low values ​​on the left represent the T1 values, and the high and low values ​​on the right represent the B0 values. The generated parameter map can be saved as a .mat file and other formats for visualization. Furthermore, to evaluate the matching quality, the root mean square error (RMSE) between the normalized true signal fingerprint of each pixel and its best-matched normalized dictionary signal fingerprint was calculated, and an RMSE error map was generated. The RMSE map can intuitively reflect the goodness of fit of the match. Qualitative observation shows that the physical phantom T1 quantitative map generated by the method of this invention is of good quality. In each phantom region with different T1 values, the T1 signal is uniformly distributed, and the boundaries between different sample regions are clearly distinguishable. For regions known to easily produce bSSFP black band artifacts under specific B0 field offset conditions, the T1 quantitative map generated in this embodiment effectively suppresses artifacts in such regions, and the consistency of T1 values ​​is high. Combined with the generated B0 quantitative map, it can be confirmed that there is a B0 field non-uniformity distribution inside or around the phantom, introduced by the experimental settings or inherent to the scanner system. Even in regions where the B0 field shift is significant, the T1 quantitative map generated in this embodiment maintains good stability and numerical reliability. Its measurement results show good consistency with those in regions where the B0 field influence is smaller. The results of the physical phantom quantitative imaging experiment intuitively demonstrate that the method of this invention can accurately quantify samples with different T1 values ​​under conditions of B0 field inhomogeneity, and effectively reduce the interference of bSSFP artifacts on T1 measurements. This preliminarily verifies the B0 robustness and quantitative accuracy of the method.

[0058] Example 3

[0059] This embodiment provides a B0-robust quantitative cardiac T1 imaging method, including the following steps:

[0060] Step S1: As Figure 2 As shown, a dual-phase increment B0 robust cardiac T1 quantitative imaging acquisition sequence was constructed, and cardiac data from volunteers were acquired based on the sequence.

[0061] The following data collection methods were used:

[0062] The B0 robust cardiac T1 quantitative imaging sequence described in this invention was performed on a United Imaging UMR 790 3.0T MRI scanner for a recruited adult volunteer.

[0063] The volunteers were positioned supine with their heads first. A body coil combined with a spinal coil was used for scanning. The scanning employed a "inhale-exhale-hold breath" breathing training pattern to ensure that the rhythm and depth of each breath were as consistent as possible, thereby maintaining the relative stability of the diaphragm position. An ECG gating system was also connected to acquire the volunteers' ECG signals.

[0064] The sequence parameters were set for in vivo quantitative cardiac imaging. The readout orientation field of view (FOV) was set to 340 mm, the phase encoding orientation FOV to 300 mm, the slice thickness to 8 mm, the voxel size to 2.21 × 1.77 × 8 cubic mm, the readout orientation resolution to 192, the phase encoding orientation resolution to 80%, the repetition time / echo time (TR / TE) to 2.86 ms / 1.45 ms, the acquisition bandwidth to 1400 Hz per pixel, and the bSSFP readout flip angle to 35°. The bSSFP phase cycle increment 1 was set to 180°, and the phase increment 2 to 270°. The adiabatic inversion pulse inversion time 1 was set to 150 ms, and the inversion time 2 to 250 ms. Parallel imaging was used for double acceleration, with 24 phase reference lines and partial phase encoding enabled. The shimming mode was "Heart - Automatic Shimming". Standard short-axis cardiac images (including basal, intermediate, and apical slices) were selected for scanning, with an spacing of 120%.

[0065] The sequence acquisition was performed using ECG gating for 13 cardiac cycles, following a 5-(3)-5 acquisition scheme: first, 5 cardiac cycles were acquired, then acquisition was paused for 3 cardiac cycles, and then 5 cardiac cycles were acquired again. After one R wave in the ECG, and after a certain trigger delay, data acquisition was performed during mid-diastole while the subject was holding their breath. In the first cardiac cycle of the five acquired cardiac cycles, a 180° adiabatic inversion pulse was applied to invert the longitudinal magnetization vector, while no pulse was applied in the other 4 cardiac cycles. The waiting time was calculated by setting the inversion time. After the waiting time, the sequence was readout using bSSFP for data acquisition. A pre-pulse with 5 inverted angles was used to enter steady state, during which no data was acquired. Then, a radiofrequency pulse with a fixed inversion angle was used for data acquisition, and a total of 71 lines were acquired. In one TR cycle, the slice selection gradient was applied synchronously with the radiofrequency pulse to select the excitation slice, followed by the application of the phase encoding gradient and the readout gradient. To ensure complete balance of the gradient magnetic field within a TR cycle, i.e., the total gradient moments on all three logic axes are zero, the layer selection gradient applies a pre-converging gradient before the RF pulse is applied and a reconverging gradient after the RF pulse is applied; the readout gradient applies a pre-converging gradient before data acquisition and a reconverging gradient after data acquisition; the phase encoding gradient applies before acquisition and a reconverging gradient of equal magnitude but opposite polarity after acquisition. In the next TR cycle, the phase of the RF pulse increases by a phase increment, for example, phase increment 1 is 180° and phase increment 2 is 270°. In the first 5 acquired cardiac cycles, the phase of the RF pulse cycles from 0° to 180° to 0°; in the subsequent 5 acquired cardiac cycles, the phase of the RF pulse cycles from 0° to 270° to 180° to 90° to 0°, achieving two phase cycles in one sequence, thereby encoding B0 information.

[0066] MATLAB was used to reconstruct the acquired k-space raw data. After reading the raw data and protocol parameters, each layer was processed independently. Based on the actual number of phase-encoded rows and readout points in the protocol parameters, the raw k-space data was padded with zeros to match the actual acquisition resolution. Based on the timestamp information and the preset reversal time, the actual reversal time and corresponding index for each acquisition time were calculated. The image order was rearranged according to the actual reversal time, and the R-wave interval of the cardiac cycle was calculated based on the timestamps of adjacent center k-spaces. The ESPIRiT (parallel imaging using eigenvector maps) method was used for coil sensitivity estimation, and the GRAPPA (GeneRalized Autocalibrating Partially Parallel Acquisition) algorithm was used for image reconstruction, ultimately yielding the following results:Figure 5 Ten multi-contrast complex images with different inversion times as shown.

[0067] Step S2: The process of constructing the simulation signal dictionary is as follows:

[0068] For parameter estimation, a multi-dimensional simulation signal dictionary is constructed. The signal entries of this dictionary are all based on the pulse sequence parameters defined in Step S1. A pulseq sequence with the same parameters is written using pypulseq, which facilitates subsequent simulations and possible migrations to subsequent simulation platforms. The dictionary simulation is based on the phase distribution map (PDG) magnetic resonance simulation tool of the open-source MRzeroCore platform. An improved simulate function is used to generate efficient parallel simulations. The accuracy parameter of the function is set to 1e-2 to balance simulation accuracy and simulation efficiency. An optimized simulation signal dictionary is constructed according to the physiological characteristics of in-vivo cardiac tissue (such as the expected myocardial and blood pool T1 / T2 ranges) and the scanning parameters used. The ranges and steps of the dictionary parameters T1, T2, B0 offset, and the optional B1 factor are carefully selected to ensure accurate characterization and efficient matching of in-vivo signals. The parameter dimensions of the dictionary and its discretized sampling ranges and steps are set as follows: The T1 relaxation time range is from 300 milliseconds to 2000 milliseconds, with a step of 5 milliseconds; the B0 field offset range is from -175 Hz to +175 Hz. This range is calculated based on the repetition time of the sequence used and corresponds to -π to π in radians, with a step of 5 Hz. In this embodiment, since no encoding modules sensitive to T2 and B1 are introduced in the sequence, to reduce the dictionary size, the T2 relaxation time is fixed at 200 milliseconds, and the B1 field scaling factor is fixed at 1.0 (i.e., an ideal B1 field). The dictionary entries are combinations of these parameters (ensuring T2 < T1, which conforms to the inherent physical constraints). During simulation, each parameter combination is regarded as an independent voxel, the proton density (PD) is set to 1, T2* is set to infinity, and the diffusion coefficient is set to 0. The simulation function is called in a batch processing manner. Depending on the different accuracy parameters, the maximum number of voxel parameter combinations processed in each batch is different. In this embodiment, 484 voxels are processed in each batch. The total signal corresponding to 10 acquisition segments is obtained by simulation for each parameter combination. Subsequently, the average value of the amplitudes of the simulated complex signals within each acquisition segment is extracted to form a signal fingerprint 10 data points long. The signal fingerprints generated by all parameter combinations constitute the original signal dictionary.

[0069] Step S3: As Figure 3 shown, the real cardiac magnetic resonance signals collected are matched with the simulation signal dictionary, and the best dictionary entry is confirmed pixel by pixel to jointly estimate the T1 and B₀ parameter maps and perform result analysis. The process is as follows:

[0070] The acquired and reconstructed multi-contrast cardiac images of healthy volunteers were used as the ground truth signals, and each pixel was matched with a signal dictionary. First, the ground truth fingerprints loaded and preprocessed in step S1 and the dictionary signal fingerprints generated in step S2 were normalized using the L2 norm. Then, the inner product of the normalized ground truth fingerprint matrix and the normalized signal dictionary matrix was calculated using matrix multiplication to obtain the similarity score between each ground truth fingerprint and all entries in the dictionary. For each pixel, i.e., each ground truth fingerprint, the dictionary entry with the highest similarity score was selected as the best match. Based on the index of the best-matched dictionary entry, the corresponding T1 value and B0 field offset value were retrieved from a pre-stored list of parameter combinations, thereby generating a T1 quantitative map and a B0 quantitative map for each acquisition layer, as shown below. Figure 6 As shown. The generated parameter map can be saved as a .mat file and other formats for visualization. Furthermore, to evaluate the matching quality, the root mean square error (RMSE) between the normalized true signal fingerprint of each pixel and its best-matched normalized dictionary signal fingerprint was calculated, and an RMSE error map was generated. The RMSE map can intuitively reflect the goodness of fit of the match. Qualitative observation shows that the in vivo T1 quantitative map image generated by the method of this invention has good quality, uniform myocardial signal distribution, clear boundary between myocardium and blood pool, and effective suppression of common bSSFP black band artifacts in the myocardial region. Combined with the generated B0 quantitative map, the physiological B0 field non-uniform distribution in the cardiac region can be observed, and the T1 quantitative map of this invention still maintains good stability and reliability in these regions. Figure 7 As shown, the T1 quantification map generated by the method of this invention, compared with the most widely used MOLLI, yields more uniform T1 quantification results in the region with the most severe black band artifacts, effectively suppressing black band artifacts. These results demonstrate the feasibility of applying the method of this invention to cardiac T1 quantification in healthy volunteers and show its potential to provide robust B0 T1 quantification in complex in vivo environments.

[0071] Example 4

[0072] This embodiment provides a B0-robust quantitative cardiac T1 imaging method, including the following steps:

[0073] Design a dual-phase increment cardiac T1 quantitative imaging acquisition sequence, and acquire raw data based on the dual-phase increment cardiac T1 quantitative imaging acquisition sequence;

[0074] As a specific implementation method, the dual-phase increment cardiac T1 quantitative imaging acquisition sequence adopts an ECG-gated method. In the first cardiac cycle of the first round of five cardiac cycles and the first cardiac cycle of the second round of five cardiac cycles, an adiabatic inversion pulse is applied to invert the longitudinal magnetization vector. The adiabatic inversion pulse is not applied in the other four cardiac cycles of the current round. After the acquisition of the first round of five cardiac cycles is completed, the acquisition of signals for the second round of five cardiac cycles is paused for three cardiac cycles.

[0075] As one specific implementation method, after the gating trigger of each cardiac cycle, signal acquisition is performed after a waiting period. After entering a steady state using pre-pulses with five increasing flip angles, data acquisition is performed using radio frequency pulses with a fixed flip angle.

[0076] In one specific implementation, the waiting time is calculated based on a preset reversal time. The waiting time for the first five cardiac cycles is less than that for the second five cardiac cycles. Signal acquisition is performed while the subject is holding their breath.

[0077] As a specific implementation method, during signal acquisition, the layer selection gradient is applied synchronously with the radio frequency pulse, followed by the application of the phase encoding gradient and the readout gradient. The layer selection gradient applies a pre-converging phase gradient before the application of the radio frequency pulse and a reconverging phase gradient after the application of the radio frequency pulse. The readout gradient applies a pre-converging phase gradient before signal acquisition and a reconverging phase gradient after data acquisition is completed. The phase encoding gradient is applied before signal acquisition and a reconverging phase gradient of equal magnitude and opposite polarity is applied after signal acquisition is completed.

[0078] As one specific implementation method, the phase increment of the radiofrequency pulses in the first five cardiac cycles and the second five cardiac cycles is set to different values.

[0079] Specifically, this embodiment improves the classic MOLLI acquisition strategy and enhances robustness to B0 field inhomogeneity through dual-phase increment. Its core sampling strategy follows a 5-(3)-5 pattern: First, the dual-phase increment B0 robust cardiac T1 quantitative imaging acquisition sequence adopts an ECG-gated method, starting with an ECG-gated trigger and applying a 180-degree adiabatic inversion pulse. In the five consecutive cardiac cycles after the inversion, after a certain delay time following the gating trigger of each cardiac cycle, a balanced steady-state free precession (bSSFP) signal readout module is executed. Within one TR cycle, the gradient magnetic fields applied in the three logical axes of slice selection, phase encoding, and readout gradient are completely balanced, i.e., the gradient moment is zero. Due to its completely balanced gradient characteristic, the phase accumulation caused by the imaging gradient is completely refocused at the end of each TR. This allows the signal to come from the coherent superposition of two paths: free induction attenuation and spin echo, thereby generating a very high signal intensity, especially for blood and myocardium with a high T2 / T1 ratio, and the acquisition speed is very fast. bSSFP is commonly used in cardiac imaging due to its high signal-to-noise ratio and acquisition efficiency. However, its signal is highly sensitive to inhomogeneities in the B0 field, easily producing dark bands in the image known as banding artifacts, which severely affect the accuracy of subsequent quantitative parameters. After completing the first five cardiac cycles of acquisition, the sequence data acquisition is paused, waiting for three cardiac cycles to allow for magnetization recovery. After the waiting period, a second 180-degree adiabatic inversion pulse is applied, with a longer inversion time than the first. After the inversion, a second signal acquisition is performed over the next five consecutive cardiac cycles, with the bSSFP signal readout module executed in each cardiac cycle.

[0080] The key improvement of the dual-phase-increment B0 robust cardiac T1 quantitative imaging acquisition sequence lies in the design of different radiofrequency pulse phase increments for the two acquisition blocks, based on the aforementioned 5-(3)-5 structure, to overcome the B0 sensitivity of bSSFP and encode B0 information. This is significantly different from the fixed 180-degree phase increment scheme commonly used in conventional bSSFP sequences. Specifically, in the first five cardiac cycles acquisition block, a first type of phase increment is set, such as the traditional 180-degree phase increment; in the second five cardiac cycles acquisition block, a second type of phase increment is similarly set, such as one that differs from the first phase increment by 90 degrees.

[0081] The purpose and effect of the above design is that the steady-state signal amplitude and phase of bSSFP are periodic curves of the cumulative phase caused by B0 offset, and different phase increments correspond to the translation of the amplitude curve in the B0 direction. By using different phase increments in the two acquisition blocks, for the same physical location, its bSSFP signal will be placed at different points on this periodic curve in the two acquisition blocks. This means that the signals in the two acquisition blocks will exhibit different amplitude and phase responses to the same B0 offset. This design achieves a dual purpose: first, it reduces the influence of black band artifacts, because even if the B0 value of a pixel causes it to fall into the signal dark band in one acquisition block, it is likely to be in the signal strong region in another acquisition block, thus complementarily reflecting organizational information; second, compared with sequences such as MOLLI that only use inverted pulse encoding of T1 information, this design enhances the B0 discrimination capability of the sequence, because the signal sequences of the two acquired stages differentially encode B0 information, and subsequent dictionary matching can use this difference to more accurately and robustly resolve and separate the T1 relaxation effect and the B0 offset effect. Ultimately, this sequence design, which incorporates a specific bSSFP phase modulation strategy, significantly reduces the error introduced by bSSFP black band artifacts in the generated T1 quantification map and improves robustness to B0 field inhomogeneities, while also enabling the estimation of the B0 quantification map.

[0082] During cardiac magnetic resonance imaging (MRI) data acquisition, the data was acquired while the patient held their breath to minimize the influence of respiratory movements. ECG-gated acquisition was performed, with acquisition triggered during mid-diastole when the heart was relatively still to reduce cardiac interference. A series of raw k-space signal data corresponding to different inversion times and containing different phase increment codes were acquired. Acquisition was accelerated through parallel imaging and partial phase coding, keeping the acquisition window for each cardiac cycle within 200 milliseconds.

[0083] A simulation signal dictionary was constructed based on a magnetic resonance physical model.

[0084] As a specific implementation method, the process of constructing a simulation signal dictionary based on a magnetic resonance physical model includes:

[0085] The parameter combination design is determined, and signal simulation is performed based on the parameter combination design, the magnetic resonance physical model and the phase distribution map magnetic resonance simulation tool to obtain the set of simulation signals under different parameter combinations; a simulation signal dictionary is constructed based on the set of simulation signals; wherein, the parameter combination design includes parameter composition and parameter range, and the parameter composition includes longitudinal relaxation time, transverse relaxation time and B0 field offset.

[0086] Specifically, the simulation signal dictionary is a set of simulated signals generated by computer simulation using a phase distribution map (PDG) based on a magnetic resonance physics model. The dictionary contains a large number of pre-calculated expected signal evolution curves corresponding to the B0 robust cardiac T1 quantitative imaging acquisition sequence with dual-phase increments designed in step S1. The amplitude of the complex signal output by the PDG is extracted, corresponding to the expected signal amplitude at each acquisition time point in the sequence, as the dictionary signal. Each dictionary entry is associated with a unique set of physiological and physical parameters, which mainly define the simulated tissue characteristics and physical environment, including longitudinal relaxation time T1, transverse relaxation time T2, and B0 field offset. Preferably, to further improve accuracy, the parameter set can be extended to include a B1 field scaling factor to describe the inhomogeneity of the B1 field, as well as other parameters that may affect signal evolution. The dictionary construction needs to ensure that the parameter values ​​cover the expected T1 and T2 value ranges of cardiac tissue under physiological and pathological conditions, the B0 field offset value range covers the range of field offsets that may occur in the target imaging area, and (if included) the expected range of B1 field inhomogeneities. The sampling density of each parameter needs to be sufficiently fine to ensure accurate parameter estimation. To balance dictionary generation efficiency, a variable density parameter sampling strategy can be adopted, setting a small step size around the expected tissue T1 value.

[0087] The original data is matched with the simulated signal dictionary to obtain the optimal dictionary entry corresponding to the original data. Based on the optimal dictionary entry, the T1 quantitative map and the B0 quantitative map are obtained.

[0088] As a specific implementation method, the process of obtaining the optimal dictionary entry corresponding to the original data includes:

[0089] Image reconstruction is performed on the original data to obtain multi-contrast images corresponding to each cardiac cycle. The amplitude of the complex signal of each pixel in the multi-contrast images is extracted as the real signal. The inner product of the real signal and each dictionary signal in the simulated signal dictionary is calculated for each pixel. The dictionary entry that maximizes the inner product of the corresponding real signal and dictionary signal is found for each pixel as the best matching entry. The inner product is calculated after normalizing all dictionary signals in the real signal and simulated signal dictionary.

[0090] Specifically, the acquired k-space raw data is reconstructed to obtain 10 multi-contrast images, corresponding to ten cardiac cycles in the acquired data sequence. The amplitude of the complex signal of each pixel is extracted as the true signal in dictionary matching, and compared one by one with each dictionary signal entry in the signal dictionary constructed in step S2. The comparison method is to normalize the true signal and the dictionary signal, and then calculate the inner product. The dictionary entry that maximizes the inner product of the normalized true signal and the dictionary signal for each pixel is found, i.e., the best matching entry. This best matching dictionary entry is generated by a specific set of parameters (T1, T2, B0 offset, and possible B1 scaling factor, etc.), and therefore this set of parameters is determined as the estimated physiological and physical parameter values ​​of the pixel. Finally, by repeating this matching process for all pixels in the image, a high-precision cardiac T1 quantitative map and the corresponding B0 quantitative map robust to B0 field inhomogeneity can be obtained simultaneously. Other parameters included in the dictionary, such as T2 and B1, can also be used to simultaneously generate corresponding T2 quantitative maps and B1 quantitative maps through matching. Dictionary matching methods utilize the multi-parameter information encoded in sequence design. By performing pattern matching with an accurate physical model containing the interactions of these parameters, they achieve synchronous and joint estimation of multiple parameters, thereby naturally solving the coupling problem between parameters and providing richer quantitative information.

[0091] This embodiment provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0092] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0093] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A B0-robust quantitative imaging method for cardiac T1, characterized in that, Includes the following steps: Design a dual-phase increment cardiac T1 quantitative imaging acquisition sequence, and acquire raw data based on the dual-phase increment cardiac T1 quantitative imaging acquisition sequence; The dual-phase incremental cardiac T1 quantitative imaging acquisition sequence uses an ECG-gated method. In the first cardiac cycle of the first round of five cardiac cycles and the first cardiac cycle of the second round of five cardiac cycles, an adiabatic inversion pulse is applied to invert the longitudinal magnetization vector. No adiabatic inversion pulse is applied in the other four cardiac cycles of the current round. After the acquisition of the first round of five cardiac cycles is completed, the acquisition of the signal for the second round of five cardiac cycles is paused for three cardiac cycles. After each cardiac cycle gating trigger, signal acquisition is performed after a waiting period. After entering steady state using pre-pulses with five increasing flip angles, data acquisition is performed using radio frequency pulses with a fixed flip angle. The waiting time is calculated based on the preset reversal time. The waiting time for the first five cardiac cycles is less than that for the second five cardiac cycles. The signal acquisition is performed while the subject is holding their breath. The phase increment of the radiofrequency pulses in the first five cardiac cycles and the second five cardiac cycles were set to different values; A simulation signal dictionary was constructed based on a magnetic resonance physical model. The original data is matched with the simulation signal dictionary to obtain the optimal dictionary entry corresponding to the original data, and the T1 quantitative map and B0 quantitative map are obtained based on the optimal dictionary entry.

2. The B0-robust quantitative cardiac T1 imaging method according to claim 1, characterized in that, During signal acquisition, the layer selection gradient is applied synchronously with the radio frequency pulse, followed by the phase encoding gradient and the readout gradient. The layer selection gradient applies a pre-coalescing gradient before the radio frequency pulse is applied and a re-coalescing gradient after the radio frequency pulse is applied. The readout gradient applies a pre-coalescing gradient before signal acquisition and a re-coalescing gradient after data acquisition is completed. The phase encoding gradient is applied before signal acquisition and a re-coalescing gradient of equal magnitude and opposite polarity is applied after signal acquisition is completed.

3. The B0-robust quantitative cardiac T1 imaging method according to claim 1, characterized in that, The process of constructing a simulation signal dictionary based on a magnetic resonance physical model includes: The parameter combination design is determined, and signal simulation is performed based on the parameter combination design, the magnetic resonance physical model, and the phase distribution map magnetic resonance simulation tool to obtain the set of simulation signals under different parameter combinations; a simulation signal dictionary is constructed based on the set of simulation signals; wherein, the parameter combination design includes parameter composition and parameter range, and the parameter composition includes longitudinal relaxation time, transverse relaxation time, and B0 field offset.

4. The B0-robust quantitative cardiac T1 imaging method according to claim 1, characterized in that, The process of obtaining the optimal dictionary entry corresponding to the original data includes: Image reconstruction is performed on the original data to obtain multi-contrast images corresponding to each cardiac cycle. The amplitude of the complex signal of each pixel in the multi-contrast images is extracted as the real signal. The inner product of the real signal of each pixel with each dictionary signal in the simulation signal dictionary is calculated. For each pixel, the dictionary entry that maximizes the inner product of the corresponding real signal and the dictionary signal is found as the optimal dictionary entry. The inner product calculation is performed after normalizing the real signal and all dictionary signals in the simulation signal dictionary.

5. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the method according to any one of claims 1-4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-4.

Images