Method and equipment for quickly measuring relaxation time of low-field NMR (nuclear magnetic resonance)

Abstract

The invention discloses a low-field NMR relaxation time rapid measurement method, and the method comprises the steps: designing an IR-bSSFP pulse sequence which comprises a 180-degree reversal pulse and an excitation pulse sequence, and enabling the flip angle of a radio frequency excitation pulse in the excitation pulse sequence to be a pseudo-random flip angle; the radio-frequency phases of the adjacent radio-frequency excitation pulses are linearly changed by a fixed phase increment; based on the IR-bSSFP pulse sequence, obtaining a component MRF signal sequence of a measured object, and simulating a magnetization vector evolution process to obtain a dictionary matrix; matching the component MRF signal with a dictionary matrix pattern in combination with a smooth constraint term to obtain corresponding longitudinal relaxation time and transverse relaxation time; the invention also discloses low-field NMR relaxation time rapid measurement equipment, which comprises a pulse sequence construction module, a dictionary generation module and a mode matching module. Multi-parameter information can be obtained through single scanning, the measurement time is remarkably shortened, the experiment complexity is reduced, and the precision and repeatability of a measurement result are improved.

Description

Technical Field

[0001] This invention belongs to the field of nuclear magnetic resonance technology, specifically relating to a method for rapid measurement of low-field NMR relaxation time, and also to a device for rapid measurement of low-field NMR relaxation time. Background Technology

[0002] Low-field nuclear magnetic resonance (NMR) relaxation time measurement, as a powerful and non-destructive analytical technique, has unique advantages in distinguishing different states of matter (such as solids, bound fluids, and free fluids) and monitoring physicochemical processes related to molecular dynamics. It is widely used in many fields, including petroleum, chemical, food, agriculture, medicine, and materials. In the field of low-field NMR technology, transverse relaxation time (… The measurement of longitudinal relaxation time is usually based on signals acquired from CPMG pulse sequences and obtained through inversion algorithms; The measurement and inversion calculation are generally performed using inversion recovery pulse sequences (IR). Combining IR-CPMG pulse sequences can realize the longitudinal relaxation time. With lateral relaxation time The combined measurement allows for the simultaneous acquisition of sample data in a single experiment. Relevant spectral distribution information.

[0003] However, measuring the relaxation time of low-field NMR has certain limitations. The measured NMR signal intensity is proportional to the square of the magnetic field strength, resulting in extremely weak signals obtained under low-field conditions. To achieve the signal-to-noise ratio required for inversion analysis, multiple signal accumulations are typically necessary, leading to excessively long data acquisition times and severely impacting experimental efficiency. This limitation becomes a significant constraint, especially for applications involving rapid dynamic processes or requiring high-throughput measurements.

[0004] Magnetic resonance fingerprinting (MRF) is a quantitative magnetic resonance technique. It involves applying a series of pseudo-randomly varying radio frequency pulses (such as the flip angle) sequentially during a single scan. Repeat time (etc.), so that the tested sample generates time-series measured signals with unique response characteristics under different excitation conditions. Then, a pattern recognition method is used to match the signals with a signal dictionary to achieve longitudinal relaxation time (…). ) and lateral relaxation time ( This technique allows for the rapid simultaneous measurement of parameters such as [parameters not specified in the original text]. While commonly used in MRI, it has not yet been introduced into low-field NMR. Applying magnetic resonance fingerprinting (MRF) technology to low-field nuclear magnetic resonance (NMR) systems to achieve , Rapid and synchronous measurement and matching of parameters is a promising but challenging research direction. Compared with high-field (such as 1.5T / 3T commonly used in clinical MRI), low-field (usually <1T, especially the 0.1T-0.5T range of permanent magnet systems) NMR / MRI systems have fundamental differences in hardware, physical basis, and signal characteristics, which brings a series of unique difficulties and unsolved problems to the application of magnetic resonance fingerprinting (MRF).

[0005] Magnetic Resonance Fingerprint Recognition (MRF) for Flip Angle Extremely sensitive to non-uniform radio frequency fields in low-field NMR. Harmony and static magnetic field The field will cause the actual flip angle Deviations from preset values ​​cause signal evolution to deviate from the theoretical model, resulting in systematic quantization errors. In nuclear magnetic resonance systems, the Larmor precession frequency of protons... From static magnetic field The strength is determined by the following relationship: ,in The gyromagnetic ratio. When the applied radio frequency pulse frequency... With the precession frequency of Lamo When they are in agreement, the magnetic resonance condition is met. At this point, a duration of... radio frequency magnetic field The pulse will cause the macroscopic magnetization vector to flip by an angle. Theoretically, the relationship between the flip angle and the radio frequency magnetic field is as follows: , in, The gyromagnetic ratio (also known as the cyclotron ratio) is determined in magnetic resonance fingerprint recognition (MRF) by precisely controlling the radio frequency magnetic field. amplitude or pulse duration To produce a series of changing flip angles This allows for the encoding of characteristic information from different organizations.

[0006] However, in practical systems, the theoretical value of the relationship between the flip angle and the radio frequency magnetic field is affected by the radio frequency field. Inhomogeneity and static magnetic field Effects of non-uniformity: 1) In low field or when using surface coils, the radio frequency magnetic field The uneven distribution in space leads to an increase in the actual radio frequency magnetic field strength at a certain location in space. Deviation from the system's preset radio frequency magnetic field strength The value of makes the actual flip angle at that position... Deviating from expectations. 2) If the static magnetic field Non-uniform, field shift exists at a certain point in space This leads to a change in the local resonant frequency. When using a fixed radio frequency pulse frequency When excited, frequency detuning occurs. At this point, the spin system is subjected to an effective field. (Radio frequency magnetic field) and The vector sum dominates, the excitation efficiency decreases, and the magnetization vector surrounds the effective field. Directional precession. This occurs when a radio frequency magnetic field is simultaneously present. Inhomogeneity and static magnetic field Inhomogeneity (denoted as static magnetic field inhomogeneity) In the case of a target flip angle The pulse, the effective flip angle actually achieved at a certain point in space. It can be approximated as: , This represents linear scaling caused by non-uniformity in the radio frequency field. The term represents the attenuation caused by the off-resonance effect resulting from the inhomogeneity of the static magnetic field.

[0007] In low-field permanent magnet systems, static magnetic field inhomogeneity The effects of low field conditions are often more pronounced. Therefore, how to generate an accurate and controllable flip angle under low field conditions is a core issue in ensuring the quantitative accuracy of magnetic resonance fingerprinting. Summary of the Invention

[0008] The purpose of this invention is to address the aforementioned problems in the prior art by providing a method for rapid measurement of low-field NMR relaxation time, and also to provide a device for rapid measurement of low-field NMR relaxation time.

[0009] The above-mentioned objectives of the present invention are achieved by the following technical means: A rapid method for measuring relaxation time in low-field NMR includes the following steps: Step 1: Design the IR-bSSFP pulse sequence. The IR-bSSFP pulse sequence consists of a 180° inversion pulse and an excitation pulse sequence. The excitation pulse sequence includes a series of RF excitation pulses, each with a corresponding flip angle. The pseudo-random flip angle; the repetition time between adjacent RF excitation pulses. The settings are based on a random generation method using Perlin noise; the RF phase change rule for adjacent RF excitation pulses is as follows: , For time point number, , They are time point numbers respectively Corresponding RF phase and time point number Corresponding radio frequency phase, The modulo operator, This is the phase increment; Step 2: Based on the Bloch equation model and the IR-bSSFP pulse sequence set in Step 1, simulate the magnetization vector evolution process and calculate the dictionary matrix. ; Step 3: Obtain the component MRF signal sequence of the object under test according to the IR-bSSFP pulse sequence set in Step 1; Step 4: Based on the acquired component MRF signals and dictionary matrix Pattern matching and recognition are performed to obtain the corresponding longitudinal relaxation time of the tested object. The value of the horizontal relaxation time The value of .

[0010] As described above, the flip angle The value of changes periodically, and in each cycle of the flip angle change, the th The flip angle corresponding to each radio frequency excitation pulse The calculation formula is as follows: , in, To represent multiplication, The index within a single rotation angle change period is denoted as the index within the period. , Indicates the serial number within the period Corresponding flip angle; time point number The sequence number of the time point of the radio frequency excitation pulse in the excitation pulse sequence, and the sequence number of the same time point. The radio frequency excitation pulse corresponds to a flip angle. and a repeating time Time point number and the serial number within the period The following relationship must be satisfied: , in, The number of time points covered by the flip angle change cycle, and the time point number. The corresponding flip angle is denoted as Then there is ; rand(5) represents a uniformly distributed random perturbation value generated in the interval [−5°, +5°].

[0011] As mentioned above, both the 180° inversion pulse and the RF excitation pulse use BIR-4 adiabatic pulses.

[0012] As described above, the BIR-4 adiabatic pulse meets the following requirements: , , , in, Indicates amplitude modulation function, , Represents the frequency modulation function. Represents the phase modulation function, time , satisfy: , , The duration of the pulse. This represents the peak amplitude of the radio frequency magnetic field. and It is a dimensionless constant. Indicates the time point sequence number The corresponding initial radio frequency phase value of the radio frequency excitation pulse. It is the phase offset, and its value is... ; It is the maximum frequency offset.

[0013] As described above, step 2 specifically includes the following steps: Step 2.1: Discretize the Bloch equation. The discretized Bloch equation is as follows: , , in, Indicates the time point sequence number The corresponding three-dimensional components of the magnetization vector, Indicates the time point sequence number The corresponding magnetization vector before the application of the radio frequency excitation pulse. Indicates the time point sequence number The corresponding magnetization vector after the application of the radio frequency excitation pulse, Indicates the flip angle For rotation angle Axis rotation matrix, defined with static magnetic field The direction that is the same as the direction is axial direction, with The mutually perpendicular directions in the orthogonal transverse plane are respectively shaft and axis; It is a positive phase rotation matrix. This is a reverse phase rotation matrix, and the two are inverse matrices of each other; and These represent the phase rotation matrices before and after the application of the radio frequency (RF) excitation pulse, respectively. They are used to introduce phase modulation of the RF pulse during each excitation process, thereby adjusting the RF phase... With phase increment Variation within the range of 0 to 2π; intermediate quantity , Represents a diagonal matrix, intermediate quantity intermediate quantity , Indicates the time point sequence number The corresponding repetition time; The three-dimensional components of the initial magnetization vector are set as This corresponds to the initial state after the 180° reversal pulse in step 1; To balance the magnetization; Step 2.2: Based on the longitudinal relaxation time The preset value range and horizontal relaxation time The preset value range is set, and multiple pairs of longitudinal relaxation times are configured. With lateral relaxation time The combination of these is denoted as the relaxation time combination. All relaxation time combinations The set of parameters forms the initial relaxation time parameter space. A simulation mesh is established on the initial relaxation time parameter space, and each grid point in the simulation mesh corresponds to a relaxation time combination. ; Step 2.3: At each grid point of the simulation grid, calculate the response process of the magnetization vector under the action of the IR-bSSFP pulse sequence designed in step 1 according to the discrete Bloch equation, and obtain the corresponding complex signal sequence, which is denoted as the evolved MRF signal sequence. Step 2.4: Combine all relaxation times The corresponding evolved MRF signal sequences are stored row by row, and the longitudinal relaxation time is removed. The value is less than the transverse relaxation time. The combination of values ​​forms a dictionary matrix. .

[0014] As described above, step 4 includes the following process: Step 4.1: Perform amplitude normalization on the component MRF signal sequence to obtain the normalized component MRF signal sequence. For dictionary matrix The amplitude of each evolving MRF signal sequence is normalized to obtain the normalized evolving MRF signal sequence. ; Step 4.2: Calculate each normalized evolved MRF signal sequence. The amplitude of the complex signal is compared with the normalized component MRF signal sequence. The inner product between the amplitudes of complex signals is used as a similarity index: , in, Indicates similarity, serving as the matching strength; Indicates the first Time point number in a normalized evolutionary MRF signal sequence The signal value; Indicates the time point number in the normalized component MRF signal sequence. The signal value; Indicates modulo; Step 4.3: In all normalized evolved MRF signal sequences Find similarity The index is the sequence number of the largest normalized evolved MRF signal. And based on the dictionary matrix The sequence number and relaxation time combination of the evolved MRF signal sequence during generation Find the corresponding relationship and locate the index. The corresponding relaxation time combination is denoted as Thus, the longitudinal relaxation time of the measured object is equal to the longitudinal relaxation time. The transverse relaxation time is equal to the transverse relaxation time. .

[0015] As described above, step 4 includes the following process: Step 4.1: Normalize the component MRF signal sequence and the evolved MRF signal sequence of each row. The normalization formula is as follows: , , in, This represents the normalized component MRF signal sequence. This represents the normalized evolutionary MRF signal sequence. Represents the first element in the dictionary matrix. The evolution of MRF signal sequences, For a normalized dictionary matrix, Indicates the sequence number corresponding to the relaxation time combination; The component MRF signal sequence; Normalized component MRF signal sequence The real and imaginary parts are concatenated to form a real vector, which is used to normalize the evolved MRF signal sequence. The real and imaginary parts of the vector are also concatenated to form a real vector: , , in, and These represent the real and imaginary parts of the signal, respectively. Let be the real vector formed by concatenating the real and imaginary parts of the normalized evolved MRF signal sequence, denoted as the concatenated real vector. ; The dictionary matrix representing the concatenation of real and imaginary parts consists of all concatenated real vectors. Arranged in rows This represents a real vector concatenated from the real and imaginary parts of the normalized component MRF signal sequence. Step 4.2: Set the objective function as follows: , in, For the weight vector, It is a smoothing constraint term. It is a regularization factor. Let be the graph Laplacian matrix constructed from the relaxation time parameter space corresponding to the dictionary matrix. All corresponding relaxation time combinations A set; Indicates minimization; for transpose, for transpose, for Transpose of; Represents the square of the L2 norm; Step 4.3: Solve the objective function to obtain the final weight vector. Non-zero weight elements That is, the weight of each component; then, based on the dictionary matrix... Evolutionary MRF signal sequence and relaxation time combination The correspondence, reverse lookup of the weight vector All non-zero weight elements Corresponding dictionary matrix relaxation time combination This allows us to determine the longitudinal relaxation time of each component of the tested object. The value of the horizontal relaxation time The value of .

[0016] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement steps 1, 2 and 4 of any one of the methods for rapid measurement of low-field NMR relaxation time.

[0017] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements steps 1, 2, and 4 of any one of the methods for rapid measurement of low-field NMR relaxation time.

[0018] A computer program product includes a computer program that, when executed by a processor, implements steps 1, 2, and 4 of any one of the methods for rapid measurement of low-field NMR relaxation time.

[0019] Compared with the prior art, the present invention has the following advantages: This invention proposes a method for rapid measurement of relaxation time in low-field NMR based on magnetic resonance fingerprinting (MRF). Compared with traditional multi-sequence scanning methods, this invention can obtain multi-parameter information in a single scan, significantly shortening the measurement time, reducing experimental complexity, and improving the accuracy and repeatability of the measurement results.

[0020] By using BIR-4 adiabatic pulses and a designed pulse sequence, the problem of flip angle deviation in low-field systems can be effectively overcome. Adiabatic pulses are a special type of RF pulse that can effectively overcome flip angle deviation in low-field systems. In non-uniform spatial conditions, the magnetization vector is uniformly excited, refocused, or reversed. Unlike traditional radio frequency pulses, adiabatic pulses do not follow the theoretical relationship between the flip angle and the radio frequency magnetic field. Instead, the flip angle of an adiabatic pulse depends on the radio frequency field. How to change its amplitude and modulation frequency (or RF phase) during the pulse. By appropriately adjusting the modulation function, only when the RF field... When the amplitude of the modulation envelope exceeds a certain threshold, it can be used in different radio frequency fields. The same flip angle is generated when the adiabatic pulse is excited. This characteristic gives the adiabatic pulse an effect on the radio frequency field. Excellent robustness to change, effectively resisting The influence of field inhomogeneity.

[0021] The pulse sequence based on BIR-4 adiabatic pulses used in this invention improves the accuracy of flip angle control and the reliability of signal acquisition.

[0022] At low fields, the signal-to-noise ratio (SNR) of an NMR signal is related to the 7 / 4 power of the field strength. The inherent SNR of low-field systems is significantly lower than that of high-field systems. Magnetic resonance fingerprinting (MRF) relies on acquiring dynamic signal changes at hundreds or even thousands of time points to generate a "fingerprint." A low SNR directly increases random noise in the fingerprint signal, making the extracted "fingerprint" blurry and severely degrading the parameters. This invention addresses the impact of low signal-to-noise ratio (SNR) on MRF signal matching by introducing a smoothing constraint term. Low SNR typically leads to high-frequency noise and random fluctuations in the signal, while the smoothing constraint term reduces the interference of high-frequency noise on the matching results through smoothing the signal. Specifically, the smoothing constraint term ensures that the matched signal has a consistent quantization accuracy and matching reliability. The horizontal axis represents the lateral relaxation time. The variation in the two-dimensional parameter plane with the vertical axis is continuous and stable, thereby improving the accuracy and reliability of matching. Through this method, even under low signal-to-noise ratio (SNR) conditions, the algorithm can still identify the correct signal features, avoiding incorrect matching caused by noise. Attached Figure Description

[0023] Figure 1A This is a flowchart illustrating Embodiment 1 of the present invention; Figure 1B This is a flowchart illustrating Embodiment 2 of the present invention; The process includes: 1. Constructing an IR-bSSFP pulse sequence; 2. Simulating and obtaining a dictionary matrix; 3. Obtaining the component MRF signal sequence of the test object based on the set IR-bSSFP pulse sequence; 4. Pattern recognition and parameter matching, and reversely searching for the longitudinal relaxation time corresponding to the test object from the dictionary matrix. The value of the horizontal relaxation time The value; Indicates the longitudinal relaxation time. Indicates the lateral relaxation time, ( ) is the th element in the dictionary matrix. The relaxation time combination corresponding to each evolved MRF signal sequence.

[0024] Figure 2 The IR-bSSFP pulse sequence used in this invention;

[0025] Indicates the time point sequence number The corresponding flip angle, Indicates the time point sequence number The corresponding repetition time, in the embodiments of the present invention .

[0026] Figure 3A This is a sequence of flip angles used to define the flip angles. The pattern of how the value changes over time; Figure 3B A repeating time series, used to define the repetition time. The pattern of how the value changes over time.

[0027] Figure 4A The inner product result of matching the MRF signal sequence of the components corresponding to a 0.2 mM / L MnCl2 solution with the dictionary matrix; Figure 4B According to Figure 4A A comparison of the obtained matched evolutionary MRF signal sequence (green line) and the measured component MRF signal sequence (blue line).

[0028] Figure 5A The inner product result of matching the component MRF signals with the dictionary matrix for a MnCl2 solution with a concentration of 0.47 mM / L; Figure 5B According to Figure 5A A comparison of the obtained matched evolutionary MRF signal (green line) and the measured component MRF signal (blue line).

[0029] Figure 6 The weight matrix is ​​a Gaussian distribution.

[0030] Figure 7A Two-dimensional weighted plots of each component of a multi-component simulated MRF signal sequence; Figure 7B The two-dimensional weight map of each component is obtained by matching the non-negative least squares (NNLS) algorithm.

[0031] Figure 8A According to Figure 7A , Figure 7B The actual component MRF signals obtained from the multi-component simulation correspond to the actual Weights and Matching Weight projection comparison chart; Figure 8B According to Figure 7A , Figure 7B The actual component MRF signals obtained from the multi-component simulation correspond to the actual Weights and Matching Weighted projection comparison chart. Detailed Implementation

[0032] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to implementation examples. It should be understood that the implementation examples described herein are for illustration and explanation only and are not intended to limit the present invention.

[0033] Example 1: A rapid method for measuring relaxation time in low-field NMR, such as Figure 1A As shown, it includes four steps: 1. Design of IR-bSSFP pulse sequence to achieve signal excitation; 2. Pre-simulation of the signal dictionary to calculate the dictionary matrix. 3. Based on the set IR-bSSFP pulse sequence, obtain the corresponding MRF signal sequence of the test object including single components (denoted as component MRF signal sequence); 4. Pattern recognition and parameter matching to obtain the corresponding longitudinal relaxation time of the test object. The value of the horizontal relaxation time The value is as follows: The specific process is as follows: Step 1: Design the IR-bSSFP pulse sequence: The IR-bSSFP pulse sequence includes a preparation phase and an acquisition phase. The preparation phase includes a 180° inversion pulse, and the acquisition phase includes an excitation pulse sequence. The excitation pulse sequence refers to a series of radio frequency excitation pulses applied consecutively after the 180° inversion pulse, used to acquire the MRF signal; the flip angle corresponding to each radio frequency excitation pulse... The pseudo-random flip angle; the repetition time between adjacent RF excitation pulses. The settings are based on a random generation method using Perlin noise; the RF phase of adjacent RF excitation pulses is set with a fixed phase increment. Linear change.

[0034] This invention employs a pseudo-random flip angle and repetition time. The signal is excited and acquired using an inverted recovery equilibrium steady-state free precession (IR-bSSFP) pulse sequence. The IR-bSSFP pulse sequence is as follows: Figure 2 As shown.

[0035] In the specific implementation, the IR-bSSFP pulse sequence first applies a 180° inversion pulse to reverse the magnetization vector to the negative direction. Axial direction, positive The axial direction is defined as relative to the static magnetic field. The directions are consistent. Then, a series of radio frequency excitation pulses are applied sequentially, each pulse at a time point having an independently set flip angle. and repetition time Flip angle Values ​​and repetition times The values ​​change according to the preset flip angle sequence and repeating time sequence, respectively. The radio frequency phase of the radio frequency excitation pulse in the excitation pulse sequence is set using a phase cycle method. The radio frequency phase of adjacent radio frequency excitation pulses is set with a fixed phase increment. Linear variation causes the radio frequency phase to cycle within the range of 0 to 2π, i.e. , For time point number, , They are time point numbers respectively Corresponding RF phase and time point number Corresponding radio frequency phase, This is the modulo operator. In this embodiment, the phase increment... The magnitude is set to 90°, and the initial RF phase is 0°. After each RF excitation, a signal acquisition operation is performed to record the magnetization vector at that time.

[0036] The specific methods for generating the flip-angle sequence and the repeating time series are as follows: The flip angle sequence is a periodic variation sequence. In this embodiment, each flip angle variation cycle contains 600 time points. A complete flip angle variation cycle includes four stages in sequence: a first sinusoidal modulation segment, a first interval segment, a second sinusoidal modulation segment, and a second interval segment, as follows: Figure 3A As shown.

[0037] To meet the requirements of the inversion recovery equilibrium steady-state free precession (IR-bSSFP) pulse sequence, the flip angle is... Using a pseudo-random method, the angle is flipped. The value of changes periodically, and in each cycle of the flip angle change, the th The flip angle corresponding to each radio frequency excitation pulse The calculation formula is as follows: , in, To represent multiplication, The index within a single rotation angle change period is denoted as the index within the period. , Indicates the serial number within the period Corresponding flip angle; time point number The sequence number of the time point of the radio frequency excitation pulse in the excitation pulse sequence, and the sequence number of the same time point. The radio frequency excitation pulse corresponds to a flip angle. and a repeating time Time point number and the serial number within the period The following relationship must be satisfied: , in, For the modulo operator, make , In this embodiment, the number of time points covered by the flip angle change cycle is... Time point number The corresponding flip angle is denoted as Then there is ; rand(5) represents a uniformly distributed random perturbation value generated in the interval [−5°, +5°].

[0038] This embodiment selects a flip angle sequence of 1000 time points, arranged in a periodic structure, where each flip angle change cycle includes 600 time points. Since 1000 time points are not divisible by the 600-time-point flip angle change cycle, the IR-bSSFP pulse sequence contains one complete cycle, i.e., from... arrive and an incomplete cycle (from arrive The total number of time points is not directly related to the number of time points included in a flip angle change cycle. In this embodiment, although there are only 600 time points in a flip angle change cycle, 1000 time points are used in order to make the signal evolution sufficient.

[0039] The repeated time series employs a Perlin noise-based random generation model to obtain smooth, continuously varying time series with repeating time intervals, such as... Figure 3B As shown. Specifically, the generation process of repeating time series can be summarized as follows: Input preset parameters into the Perlin noise random generation model. The preset parameters include sequence length (the sequence length is equal to the number of RF excitation pulses in the IR-bSSFP pulse sequence), noise fundamental frequency, noise amplitude, number of fractal superposition layers, and amplitude attenuation coefficient to obtain a set of noise sequences with continuous variation characteristics. Then, the noise sequence is constrained to the preset repetition time range [10, 15] ms through linear mapping to form the final repetition time sequence.

[0040] In this embodiment, the number of time points in the repeated time series is 1000, which is consistent with the flip angle sequence to ensure the synchronization of parameter updates during signal evolution.

[0041] Step 2: Based on the Bloch equation model and the IR-bSSFP pulse sequence set in Step 1, simulate the magnetization vector evolution process and calculate the dictionary matrix. This allows for the establishment of a "fingerprint database": Based on the Bloch equation model, simulations were conducted on different tissues (corresponding to different longitudinal relaxation times). Lateral relaxation time The magnetization vector evolution process of the combination of IR-bSSFP pulses under the action of the set IR-bSSFP pulse sequence is used to obtain the complex signal sequence corresponding to each tissue, which is denoted as the evolved MRF signal sequence.

[0042] The continuous form of the Bloch equations can be expressed as: , in, The magnetization vector. Indicates the magnetization vector in the transverse direction The component along the axial direction is denoted as the transverse component. ; Indicates the magnetization vector in the transverse direction The component along the axial direction is denoted as the transverse component. ; Indicates the magnetization vector in the longitudinal direction Components along the axial direction; The axial direction is defined as relative to the static magnetic field. A direction that is in the same direction. shaft and The axis is located at and Directions perpendicular to each other in a transverse plane orthogonal to the axis. The time-varying external magnetic field vector acting on the spin system includes the longitudinal static magnetic field component, denoted as the static magnetic field. And the radio frequency magnetic field component located in the transverse plane, denoted as the radio frequency magnetic field. It may further include frequency detuning terms caused by inhomogeneities in the static magnetic field. The gyroscope ratio, and These represent the longitudinal relaxation time and the transverse relaxation time, respectively. To balance the magnetization intensity.

[0043] Step 2.1: To adapt to the IR-bSSFP pulse sequence designed in Step 1 of this invention, the Bloch equation is discretized in this embodiment. The magnetization vector undergoes two stages of change at each time point: (1) Instantaneous reversal caused by radio frequency excitation pulse; (2) Repetition time The period of relaxation and recovery.

[0044] Its discrete form can be expressed as follows: , , in, Indicates the time point sequence number The corresponding three-dimensional components of the magnetization vector, Indicates the time point sequence number The corresponding three-dimensional components of the magnetization vector, Indicates the time point sequence number The corresponding magnetization vector before the application of the radio frequency excitation pulse. Indicates the time point sequence number The corresponding magnetization vector after the application of the radio frequency excitation pulse, Indicates the flip angle For rotation angle Axis rotation matrix; Belongs to the phase cyclic matrix, It is a positive phase rotation matrix. This is a reverse phase rotation matrix, and the two are inverse matrices of each other; and These represent the phase rotation matrices before and after the application of the radio frequency (RF) excitation pulse, respectively. They are used to introduce phase modulation of the RF pulse during each excitation process, thereby adjusting the RF phase... With phase increment It varies within the range of 0 to 2π. (Intermediate quantity) , Represents a diagonal matrix, intermediate quantity intermediate quantity , Indicates the time point sequence number The corresponding repetition time.

[0045] In this embodiment, the three-dimensional components of the initial magnetization vector are set as follows: This corresponds to the initial state after the 180° reversal pulse in step 1.

[0046] The magnetization vector is sampled immediately after each RF flip, acquiring the transverse component. The transverse component... and Composition of complex signals And arrange them in chronological order to form a signal sequence under a single set of parameters.

[0047] Step 2.2: Based on the longitudinal relaxation time The preset value range and horizontal relaxation time The preset value range is set, and multiple pairs of longitudinal relaxation times are configured. With lateral relaxation time The combination of these is denoted as the relaxation time combination. All relaxation time combinations The set of parameters forms the initial relaxation time parameter space. A simulation mesh is established on the initial relaxation time parameter space, and each grid point in the simulation mesh corresponds to a relaxation time combination. In this embodiment, the longitudinal relaxation time... The preset value range is 100~5000 ms, with a step size of 20 ms in the 100~2000 ms range and a step size of 300 ms in the 2000~5000 ms range; lateral relaxation time The preset value range is 20~3000ms, with a step size of 5ms for the 20~100ms range, 10ms for the 100~200ms range, and 200ms for the 200~3000ms range. This is achieved through all relaxation time combinations. It can generate a complete initial relaxation time parameter space.

[0048] Step 2.3: At each grid point of the simulation grid, calculate the response process of the magnetization vector under the action of the IR-bSSFP pulse sequence designed in Step 1 according to the Discrete Bloch equation, and obtain the corresponding complex signal sequence, denoted as the evolved MRF signal sequence. During the simulation, the radio frequency phase cycling effect is considered, causing the radio frequency phase to increment within the range of 0 to 2π. Periodic changes.

[0049] Step 2.4: Combine all relaxation times The corresponding evolved MRF signal sequences are stored row by row, and the longitudinal relaxation time is removed. The value is less than the transverse relaxation time. The combination of values ​​forms a dictionary matrix with dimensions (3334×1000). dictionary matrix Each row in the table corresponds to a set of relaxation times. Evolution of MRF signal sequences under certain conditions. Dictionary matrix. This will be used in the subsequent matching and identification process to enable rapid measurement of tissue relaxation time parameters.

[0050] Step 3: Perform radio frequency excitation according to the IR-bSSFP pulse sequence set in Step 1, and immediately perform signal acquisition after each radio frequency excitation to obtain the component MRF signal sequence of the analyte, including single components: In this embodiment, a magnetic resonance experimental setup was used to acquire signals from the samples. The equipment used was a 6 MHz low-field nuclear magnetic resonance spectrometer system. The radio frequency (RF) coil used in the experiment was a single-turn RF coil with a diameter of 3 cm. The samples were prepared using 0.2 mmol / L and 0.47 mmol / L MnCl2 solutions. During the experiment, the samples were loaded into individual sample tubes and placed at the center of the RF coil to ensure magnetic field homogeneity. After the system was tuned, the IR-bSSFP pulse sequence designed in step 1 was loaded via the control console for signal excitation and acquisition.

[0051] Throughout the entire IR-bSSFP pulse sequence, the 180° inversion pulse and the subsequent RF excitation pulse both employ BIR-4 adiabatic pulses to achieve control over the flip angle. Precise control is achieved. As an adiabatic pulse, BIR-4 can achieve precise and stable excitation even in highly non-uniform radio frequency fields. The amplitude modulation function of the BIR-4 adiabatic pulse at each time point... Frequency modulation function With phase modulation function They are defined as follows: , , , in, Represents the frequency modulation function, time , satisfy: , , The duration of the pulse. This represents the peak amplitude of the radio frequency magnetic field. and It is a dimensionless constant used to determine the degree to which the pulse satisfies the adiabatic condition. Indicates the time point sequence number In this embodiment, corresponding to the initial radio frequency phase value of the radio frequency excitation pulse, According to the phase increment set in step 1 The size is taken as 90°. , , And so on. It is important to note that the phase in BIR-4 adiabatic pulse phase modulation is fundamentally different from the phase in the RF phase cycle of the RF excitation pulse. The phase in BIR-4 adiabatic pulse phase modulation describes the instantaneous phase of the adiabatic pulse that changes continuously with time within the pulse, controlling the trajectory of the magnetization vector in the rotating coordinate system. In contrast, the phase in the RF phase cycle of the RF excitation pulse describes the RF phase difference between RF excitation pulses at different time points. Both act on the internal structure of the RF excitation pulse and the inter-pulse phase relationship, respectively, and are independent of each other in terms of time scale and physical mechanism.

[0052] It is the phase offset, and its value is... Time point number Corresponding to each flip angle That is, as in this embodiment Each flip corner All consist of a duration of The modulation function described above generates the amplitude and phase changes during the duration of the radio frequency pulse. Implemented internally, its overall effect determines the net flip angle of the magnetization vector after the pulse ends. Among these, the phase shift... Used to regulate the radio frequency pulse in the first The target flip angle generated at each time point It should be noted that the modulation function should be considered as a whole throughout the entire duration. The cumulative effect of the internal magnetization vector ultimately determines the effect of the corresponding flip angle. This is the maximum frequency offset. Using the modulation function described above, at the end of each RF pulse, the magnetization vector will undergo a corresponding flip angle.

[0053] In this embodiment, the pulse duration Take 100us, Take 1823 Hz as a dimensionless constant. =10, dimensionless constant It is 1.47.

[0054] Sequence number at each time point After the corresponding magnetization vector flips, the corresponding free induction decay (FID) signal is immediately acquired, and the sampling point with the largest FID signal amplitude is taken as the MRF signal input. The time point numbers are then set. The corresponding MRF signal inputs are arranged to obtain the component MRF signal sequence. In this embodiment, the sampling bandwidth is 500kHz, the sampling time is 64µs, and all signal acquisitions are completed under the same system calibration conditions.

[0055] Step 4: Pattern recognition and parameter matching.

[0056] The component MRF signal sequences actually acquired in step 3 are compared with the dictionary matrix generated in step 2. Pattern recognition was performed on the evolved MRF signal sequence to determine the longitudinal relaxation time of the sample. The value of the horizontal relaxation time The value of .

[0057] Step 4.1: Before pattern recognition, the amplitude normalization process is performed on the acquired component MRF signal sequences and the evolved MRF signal sequences. Let the dictionary matrix from Step 3 be used. for: , Each row Indicates the th in the dictionary matrix The evolution of MRF signal sequences, For the first Time point number in an evolving MRF signal sequence The signal value, The index represents the sequence number of the relaxation time combination, and the dictionary matrix represents the sequence number of the relaxation time combination. The sequence number of the evolved MRF signal sequence. In this embodiment, the total number of relaxation time combinations in the dictionary matrix is ​​[value missing]. .

[0058] The acquired component MRF signal sequences are denoted as: , Indicates the time point number in the component MRF signal sequence The signal values. To eliminate the impact of signal amplitude differences on matching accuracy, amplitude normalization was performed on the signals in the evolved MRF signal sequence and the acquired component MRF signal sequence, respectively: , , in, This represents the normalized evolutionary MRF signal sequence. This represents the normalized component MRF signal sequence. The L2 norm of the signal is represented by: , , in, In this embodiment, the number of radio frequency excitation pulses in the entire IR-bSSFP pulse sequence is [number]. ; Indicates modulo; Step 4.2: This embodiment uses the inner product method as the pattern recognition method. Specifically, it calculates the inner product method for each normalized evolutionary MRF signal sequence. The amplitude of the signal is compared with the normalized component MRF signal sequence. The inner product between the amplitudes of the signals is used as a similarity index:

[0059] , in, Indicates similarity, characterizing the strength of the match; Indicates the first Time point number in a normalized evolutionary MRF signal sequence The signal value; Indicates the time point number in the normalized component MRF signal sequence. The signal value.

[0060] Step 4.3: In all normalized evolved MRF signal sequences In the process, find the match strength (i.e., similarity) The largest index (i.e., the sequence number of the normalized evolved MRF signal sequence) is denoted as index. Therefore, based on the dictionary matrix The sequence number and relaxation time combination of the evolved MRF signal sequence during generation The correspondence, reverse lookup dictionary matrix Find the index The corresponding relaxation time combination is denoted as Thus, the longitudinal relaxation time of the measured object is equal to the longitudinal relaxation time. The transverse relaxation time of the tested object is equal to the transverse relaxation time. .

[0061] In this embodiment, the normalized component MRF signal sequence and the normalized evolution MRF signal sequence corresponding to the 0.2 mM / L MnCl2 solution are... The result of the matching inner product is as follows Figure 4A As shown; Normalized component MRF signal sequence and normalized evolution MRF signal sequence corresponding to 0.47 mM / L MnCl2 solution The result of the matching inner product is as follows Figure 5A As shown; according to Figure 4A The comparison graph of the resulting matched normalized evolved MRF signal sequence (green line) and the normalized component MRF signal sequence (blue line) is shown below. Figure 4B As shown; according to Figure 5A The comparison graph of the resulting matched normalized evolved MRF signal sequence (green line) and the normalized component MRF signal sequence (blue line) is shown below. Figure 5B As shown; The experimental results show that the method of the present invention can measure NMR relaxation time relatively accurately under certain signal-to-noise ratio conditions.

[0062] Example 2 A rapid method for measuring relaxation time in low-field NMR, such as Figure 1B As shown, it includes four steps: 1. IR-bSSFP pulse sequence design; 2. Pre-simulation of the signal dictionary to calculate the dictionary matrix. 3. Based on the set IR-bSSFP pulse sequence, obtain the MRF signal (denoted as component MRF signal sequence) of the test object including multiple components. In this embodiment, the corresponding component MRF signal is generated by simulating the test object including multiple components using a Gaussian weight matrix; 4. Pattern recognition and parameter matching are performed to obtain the corresponding longitudinal relaxation time of the test object. The value of the horizontal relaxation time The value is as follows: The specific process is as follows: Step 1, IR-bSSFP pulse sequence design: This invention employs a pseudo-random flip angle and repetition time. The signal is excited and acquired using an inverted recovery equilibrium steady-state free precession (IR-bSSFP) pulse sequence. The IR-bSSFP pulse sequence is as follows: Figure 2 As shown.

[0063] The IR-bSSFP pulse sequence includes a preparation phase and an acquisition phase. The preparation phase includes a 180° inversion pulse, and the acquisition phase includes an excitation pulse sequence. The excitation pulse sequence refers to a series of radio frequency excitation pulses applied consecutively after the 180° inversion pulse, used to acquire the MRF signal; the flip angle corresponding to each radio frequency excitation pulse... The pseudo-random flip angle; the repetition time between adjacent RF excitation pulses. The settings are based on a random generation method using Perlin noise; the RF phase of adjacent RF excitation pulses is set with a fixed phase increment. Linear change.

[0064] In the specific implementation, the IR-bSSFP pulse sequence first applies a 180° inversion pulse to reverse the magnetization vector to the negative direction. Axial direction. Subsequently, a series of radio frequency excitation pulses are applied sequentially, each pulse at a time point having an independently set flip angle. and repetition time Flip angle Values ​​and repetition times The values ​​change according to the preset flip angle sequence and repeating time sequence, respectively. The radio frequency phase of the radio frequency excitation pulse in the excitation pulse sequence is set using a phase cycle method. The radio frequency phase of adjacent radio frequency excitation pulses is set with a fixed phase increment. Linear variation causes the radio frequency phase to cycle within the range of 0 to 2π, i.e. , For time point number, , They are time point numbers respectively Corresponding RF phase and time point number Corresponding radio frequency phase, This is the modulo operator. In this embodiment, the phase increment... The magnitude is set to 90°, and the initial RF phase is 0°. After each RF excitation, a signal acquisition operation is performed to record the magnetization vector at that time.

[0065] The specific methods for generating flip-angle sequences and repeating time series are as follows: The flip angle sequence is a periodic variation sequence. In this embodiment, each flip angle variation cycle contains 600 time points. A complete flip angle variation cycle includes four stages in sequence: a first sinusoidal modulation segment, a first interval segment, a second sinusoidal modulation segment, and a second interval segment, as follows: Figure 3A As shown.

[0066] Flip angle The value of changes periodically, and in each cycle of the flip angle change, the th The flip angle corresponding to each radio frequency excitation pulse The calculation formula is as follows: , in, To represent multiplication, The index within a single rotation angle change period is denoted as the index within the period. , Indicates the serial number within the period Corresponding flip angle; time point number The sequence number of the time point of the radio frequency excitation pulse in the excitation pulse sequence, and the sequence number of the same time point. There is a corresponding flip angle and a repeating time Time point number and the serial number within the period The following relationship must be satisfied: , in, The modulo operator is used to make... , In this embodiment, the number of time points covered by the flip angle change cycle is... Time point number The corresponding flip angle is denoted as Then there is ; Here, rand(5) represents the uniformly distributed random perturbation value generated in the interval [−5°,+5°].

[0067] This embodiment selects a flip angle sequence comprising 1000 time points, arranged in a periodic structure, where each flip angle change cycle includes 600 time points. Since 1000 time points are not divisible by the 600-time-point flip angle change cycle, the IR-bSSFP pulse sequence contains one complete cycle, i.e., from... arrive And an incomplete cycle, namely from arrive The total number of time points is not directly related to the number of time points included in a flip angle change cycle. In this embodiment, although there are only 600 time points in a flip angle change cycle, 1000 time points are used in order to make the signal evolution sufficient.

[0068] Repeating time series are generated using a random generation model based on Perlin noise to obtain smooth repeating time series with continuously varying time intervals, such as... Figure 3B As shown.

[0069] Specifically, the process of generating repeating time series can be summarized as follows: Input preset parameters into the Perlin noise random generation model. The preset parameters include sequence length (the sequence length is equal to the number of RF excitation pulses in the IR-bSSFP pulse sequence), noise fundamental frequency, noise amplitude, number of fractal superposition layers, and amplitude attenuation coefficient to obtain a set of noise sequences with continuous variation characteristics. Then, the noise sequence is constrained to the preset repetition time range [10, 15] ms through linear mapping to form the final repetition time sequence.

[0070] In this embodiment, the number of time points in the repeated time series is 1000, which is consistent with the flip angle sequence to ensure the synchronization of parameter updates during signal evolution.

[0071] Step 2: Based on the Bloch equation model and the IR-bSSFP pulse sequence set in Step 1, simulate the magnetization vector evolution process and calculate the dictionary matrix. This allows for the establishment of a "fingerprint database": Based on the Bloch equation model, simulations were conducted on different tissues (corresponding to different longitudinal relaxation times). Lateral relaxation time The magnetization vector evolution process of the combination of IR-bSSFP pulses under the action of the set IR-bSSFP pulse sequence is used to obtain the complex signal sequence corresponding to each tissue, which is denoted as the evolved MRF signal sequence.

[0072] The continuous form of the Bloch equations can be expressed as: , in, The magnetization vector. Indicates the magnetization vector in the transverse direction The component along the axial direction is denoted as the transverse component. ; Indicates the magnetization vector in the transverse direction The component along the axial direction is denoted as the transverse component. ; Indicates the magnetization vector in the longitudinal direction Components along the axial direction; The axial direction is defined as relative to the static magnetic field. A direction that is in the same direction. shaft and The axis is located at and Directions perpendicular to each other in a transverse plane orthogonal to the axis. The time-varying external magnetic field vector acting on the spin system includes the longitudinal static magnetic field component, denoted as the static magnetic field. And the radio frequency magnetic field component located in the transverse plane, denoted as the radio frequency magnetic field. It may further include frequency detuning terms caused by static magnetic field inhomogeneity; The gyroscope ratio, and These represent the longitudinal relaxation time and the transverse relaxation time, respectively. To balance the magnetization intensity.

[0073] Step 2.1: To adapt to the IR-bSSFP pulse sequence designed in Step 1 of this invention, the Bloch equation is discretized in this embodiment. The magnetization vector undergoes two stages of change at each time point: (1) Instantaneous reversal caused by radio frequency excitation pulse; (2) Repetition time The period of relaxation and recovery.

[0074] Its discrete form can be expressed as follows: , , in, Indicates the time point sequence number The corresponding three-dimensional components of the magnetization vector, Indicates the time point sequence number The corresponding three-dimensional components of the magnetization vector, Indicates the time point sequence number The corresponding magnetization vector before the application of the radio frequency excitation pulse. Indicates the time point sequence number The corresponding magnetization vector after the application of the radio frequency excitation pulse, Indicates the flip angle For rotation angle Axis rotation matrix; Belongs to the phase cyclic matrix, It is a positive phase rotation matrix. This is a reverse phase rotation matrix, and the two are inverse matrices of each other; and These represent the phase rotation matrices before and after the application of the radio frequency (RF) excitation pulse, respectively. They are used to introduce phase modulation of the RF pulse during each excitation process, thereby adjusting the RF phase... With phase increment Variation within the range of 0 to 2π; intermediate quantity , Represents a diagonal matrix, intermediate quantity intermediate quantity , Indicates the time point sequence number The corresponding repetition time.

[0075] In this embodiment, the three-dimensional components of the initial magnetization vector are set as follows: This corresponds to the initial state after the 180° reversal pulse in step 1.

[0076] The magnetization vector is sampled immediately after each RF flip, acquiring the transverse component. The transverse component... and Composition of complex signals And arrange them in chronological order to form a signal sequence under a single set of parameters.

[0077] Step 2.2: Based on the longitudinal relaxation time The preset value range and horizontal relaxation time The preset value range is set, and multiple pairs of longitudinal relaxation times are configured. With lateral relaxation time The combination of these is denoted as the relaxation time combination. All relaxation time combinations The set of parameters forms the initial relaxation time parameter space. A simulation mesh is established on the initial relaxation time parameter space, and each grid point in the simulation mesh corresponds to a relaxation time combination. In this embodiment, the longitudinal relaxation time... and lateral relaxation time The values ​​range from 10ms to 10s. To precisely cover the entire relaxation time range, and All values ​​are calculated using a logarithmic scale, with each logarithmic relaxation time divided into 80 points (i.e., an 80×80 grid). The step size increases with the logarithmic relaxation time. Within a smaller relaxation time range, the step size is smaller; within a larger relaxation time range, the step size is larger. This is achieved through all relaxation time combinations. It can generate a complete initial relaxation time parameter space.

[0078] Step 2.3: At each grid point of the simulation grid, calculate the response process of the magnetization vector under the action of the IR-bSSFP pulse sequence designed in Step 1 according to the Discrete Bloch equation, and obtain the corresponding complex signal sequence, denoted as the evolved MRF signal sequence. During the simulation, the phase cycling effect is considered, causing the RF phase to increment by phase increments within the range of 0~2π. Periodic changes.

[0079] Step 2.4: Combine all relaxation times The corresponding evolved MRF signal sequences are stored row by row, and the longitudinal relaxation time is removed. The value is less than the transverse relaxation time. The combination of values ​​forms a dictionary matrix with dimensions (3240×1000). dictionary matrix Each row in the table corresponds to a set of relaxation times. Evolutionary MRF signal sequences under certain conditions, dictionary matrix The Middle Relaxation time combination corresponding to the line evolution MRF signal sequence This is denoted as the relaxation time combination. dictionary matrix All corresponding relaxation time combinations The set of elements constitutes the relaxation time parameter space corresponding to the dictionary matrix. (Dictionary matrix) This will be used in the subsequent matching and identification process to enable rapid measurement of tissue relaxation time parameters.

[0080] Step 3: Based on the IR-bSSFP pulse sequence set in Step 1, obtain the component MRF signal sequence of the analyte, which includes multiple components.

[0081] In this embodiment, a Gaussian distribution weight matrix is ​​used for simulation to generate a component MRF signal sequence of the multi-component object under test.

[0082] This step generates a Gaussian-distributed weight matrix, such as... Figure 6 As shown, the Gaussian distribution weight matrix and dictionary matrix are used. Weighted summation is performed to generate the final component MRF signal sequence.

[0083] First, a Gaussian weight matrix is ​​constructed that corresponds to the simulation mesh in step 3. The dimensions of the Gaussian weight matrix are the same as those of the simulation mesh, which is 80×80, representing the longitudinal relaxation time. and lateral relaxation time Each point is assigned to 80 points, and each element in the Gaussian weight matrix represents a dictionary matrix. The corresponding relaxation time combination The weight values ​​are determined by the Gaussian distribution. In this embodiment, the Gaussian weight matrix is ​​generated using a logarithmic Gaussian function, with the following formula: , in, These are the elements of a Gaussian distributed weight matrix, representing the corresponding combination of relaxation times. The weights; , These are the mean parameters of the Gaussian distribution for the longitudinal relaxation time and the mean parameters of the Gaussian distribution for the transverse relaxation time, respectively. They are used only as the central parameters of the Gaussian weighting function to construct the prior weight distribution of the analog signal, and are not obtained by statistical averaging of dictionary or signal data. and These are the standard deviations of the longitudinal relaxation time and the transverse relaxation time, respectively.

[0084] In this embodiment, the standard deviation of the longitudinal relaxation time and the standard deviation of the lateral relaxation time Both are 0.1, while the mean parameter of the Gaussian distribution of the longitudinal relaxation time is... With the mean parameter of the Gaussian distribution of the transverse relaxation time It is not a fixed preset, but is generated randomly. Specifically, because ,and Therefore, the corresponding logarithmic range is and To ensure that the Gaussian peak appears randomly throughout the entire logarithmic scale space, this embodiment specifies the mean parameter of the Gaussian distribution of the longitudinal relaxation time. and the mean parameter of the Gaussian distribution of the transverse relaxation time Use the following random generation method: , , That is, the mean parameter of the Gaussian distribution of the longitudinal relaxation time. and the mean parameter of the Gaussian distribution of the transverse relaxation time longitudinal relaxation time and lateral relaxation time Uniform sampling within the logarithmic range ensures that the Gaussian center has complete randomness.

[0085] After generating the corresponding Gaussian distributed weight matrix, the Gaussian distributed weight matrix is ​​normalized to ensure that the weight values ​​in the Gaussian distributed weight matrix are standardized to a certain range. Specifically, matrix normalization is used to scale the maximum value of each row of the Gaussian distributed weight matrix to 1 to ensure that the relative contribution of the weights is consistent.

[0086] After the above processing, the resulting Gaussian distribution weight matrix and dictionary matrix are... Each evolved MRF signal sequence is multiplied element-wise and accumulated to obtain a preliminary simulated MRF signal sequence. Specifically, for each valid grid point (one grid point corresponds to one relaxation time combination)... The corresponding evolved MRF signal sequence is multiplied by the weights at the corresponding positions in the Gaussian weight matrix, and all multiplications are summed to obtain the preliminary simulated MRF signal sequence. This process can be represented as:

[0087] in, This represents a preliminary simulated MRF signal sequence with a dimension of 1×1000; It is the weight matrix of the Gaussian distribution. Line 1 Column elements, and These represent the longitudinal relaxation time and the transverse relaxation time corresponding to the grid points in the simulation mesh, respectively. It is the combination of relaxation times in the dictionary matrix. The value of the evolved MRF signal sequence. Corresponding longitudinal relaxation time. The value is less than the transverse relaxation time. When the value is, .

[0088] Through this weighted accumulation process, the preliminary simulation of the MRF signal sequence can accurately reflect each relaxation time combination. It contributes to the final signal and can better mimic the magnetic resonance signal response of different tissues.

[0089] To simulate the effects of noise in actual measurements, this embodiment adds Gaussian noise with a signal-to-noise ratio (SNR) of 50 dB to the initial simulated MRF signal sequence, thereby obtaining the final component MRF signal sequence: , in, The component MRF signal sequence, It is Gaussian noise generated according to a preset signal-to-noise ratio level.

[0090] Step 4: Pattern recognition and parameter matching.

[0091] This step involves pattern recognition and parameter matching between the simulated component MRF signal sequences and the evolved MRF signal sequences. Specifically, the generated component MRF signal sequences and the evolved MRF signal sequences are first preprocessed uniformly. Then, a pattern recognition algorithm is used to match the preprocessed signals, and the longitudinal relaxation time corresponding to the measured object is determined based on the matching results. Values ​​and lateral relaxation time The value of .

[0092] Step 4.1: Normalize the component MRF signal sequence and the evolved MRF signal sequence of each row to ensure that the amplitudes of all signals in the same signal sequence are consistent, so that they have the same characteristics during subsequent matching. The normalization formula is: , , in, This represents the normalized component MRF signal sequence. This represents the normalized evolutionary MRF signal sequence. Represents the first element in the dictionary matrix. The evolution of MRF signal sequences, For a normalized dictionary matrix, Indicates the sequence number corresponding to the relaxation time combination, representing the relaxation time combination. Corresponding dictionary matrix The sequence number of the evolved MRF signal sequence, i.e., the evolved MRF signal sequence. Corresponding relaxation time combination , ; Since the MRF signal is a complex signal, the normalized component MRF signal sequence needs to be processed during the matching process. The real and imaginary parts are concatenated to form a real vector, which is used to normalize the evolved MRF signal sequence. The real and imaginary parts of the vector are also concatenated into a single real vector. The concatenation formula is as follows: , .

[0093] in, and These represent the real and imaginary parts of the signal, respectively. Let be the real vector formed by concatenating the real and imaginary parts of the normalized evolved MRF signal sequence, denoted as the concatenated real vector. ; The dictionary matrix representing the concatenation of real and imaginary parts consists of all concatenated real vectors. Arranged in rows This represents a real vector formed by concatenating the real and imaginary parts of the normalized component MRF signal sequence.

[0094] Step 4.2: Set the objective function: After preprocessing in step 4.1, a pattern recognition algorithm is used to match the component MRF signal sequence with the evolutionary MRF signal sequence. This algorithm optimizes the fit of the evolutionary MRF signal sequence to the analog signal by minimizing the objective function, thereby obtaining the corresponding relaxation time parameters.

[0095] The basic objective function of pattern recognition is expressed as: , in, For the weight vector, Let be the weight vector to be solved, representing the weight magnitude of each evolved MRF signal sequence. As a weighted element; This represents the total number of relaxation time combinations in the dictionary matrix, in this embodiment... , for transpose, for Transpose of; This represents the square of the L2 norm.

[0096] In MRF signal analysis, the combination of adjacent relaxation times The corresponding MRF signal sequence is usually continuous. To suppress oscillations in the solution during the solution process and improve the stability of parameter estimation, a graph Laplacian matrix regularization term, i.e., a smoothing constraint term, is introduced into the objective function. The graph Laplacian matrix regularization term penalizes the combination of adjacent relaxation times. The changes in the weights of the corresponding MRF signal sequences ensure that the solution remains continuous and smooth in the relaxation time parameter space corresponding to the dictionary matrix, thereby improving the physical rationality and numerical stability of the solution. At this point, the objective function is expressed as: , in, For the weight vector, It is a smoothing constraint term. It is a regularization factor. The graph Laplacian matrix is ​​constructed from the relaxation time parameter space corresponding to the dictionary matrix. It is used to characterize the adjacency relationships and local smoothing constraints between evolving MRF signal sequences. The relaxation time parameter space corresponding to the dictionary matrix is ​​the dictionary matrix. All corresponding relaxation time combinations A set; Indicates minimization; for The transpose of .

[0097] Specifically, firstly, based on the dictionary matrix Various evolutionary MRF signal sequences and relaxation time combinations A one-to-one correspondence is established between the weight vectors. Each weight element Mapped to longitudinal relaxation time The horizontal axis represents the lateral relaxation time. This is a two-dimensional parameter plane with the vertical axis. Then, for each weight element... On this two-dimensional parameter plane, the Euclidean distance is calculated to determine the relationship with the weight elements. The most recent weight elements are considered as neighboring points in the parameter space.

[0098] remember , Weight vector The two weight elements in the equation, and their corresponding relaxation time combinations are denoted as follows: and Then the relaxation time combination and Euclidean distance in a two-dimensional parametric plane is defined as: , After constructing the adjacency relationship in the above manner, the graph Laplacian matrix Smoothing constraints are applied to adjacent weight elements in the two-dimensional parametric plane, thereby ensuring that the resulting weight distribution remains continuous and smooth in the two-dimensional parametric plane.

[0099] Step 4.3: Solve for the objective function and perform a reverse lookup of the dictionary matrix. The longitudinal relaxation time corresponding to each component was obtained. The value of the horizontal relaxation time The values ​​and weights of each component: In this embodiment, the nonnegative least squares (NNLS) algorithm is used to solve the above objective function.

[0100] In the process of minimizing the objective function, the final weight vector is obtained through multiple optimization iterations using the nonnegative least squares (NNLS) method. (i.e., the weights of each evolving MRF signal sequence), non-zero weight elements That is, the weight of each component; then, based on the dictionary matrix... Evolutionary MRF signal sequence and relaxation time combination The correspondence, reverse lookup of the weight vector All non-zero weight elements Corresponding dictionary matrix relaxation time combination This allows us to determine the longitudinal relaxation time of each component of the tested object. With lateral relaxation time As shown in Figures 7 and 8, the relaxation time combinations obtained from multiple components... The result resembles multiple consecutive peaks.

[0101] according to Figure 7A Two-dimensional weighted plot of multi-component MRF signal sequences (obtained from the Gaussian weight matrix in step 3). Weight projection and Figure 7B Two-dimensional weighted map of the component MRF signal sequences of the multi-component group (obtained from step 4) Weight projection comparison chart, such as Figure 8A As shown; according to Figure 7A Two-dimensional weighted plot of multi-component MRF signal sequences (obtained from the Gaussian weight matrix in step 3). Weight projection and Figure 7B Weight vector of component MRF signal sequence in multi-component (Obtained from step 4) Weight projection comparison chart, such as Figure 8B As shown; The experimental results show that, under certain signal-to-noise ratio conditions, the method of the present invention can accurately measure the NMR relaxation time and its proportion of multiple components.

[0102] Example 3 A device for rapid measurement of low-field NMR relaxation time, used to implement the rapid measurement method for low-field NMR relaxation time in Examples 1 and 2, includes: A pulse sequence construction module is used to implement step 1 in Examples 1 and 2; A dictionary generation module is used to implement step 2 in Examples 1 and 2; The pattern matching module is used to implement step 4 in Embodiment 1 and Embodiment 2.

[0103] Example 4 A computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement steps 1, 2, and 4 of Embodiments 1 and 2 described above.

[0104] Example 5 A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements steps 1, 2, and 4 of Embodiments 1 and 2 described above.

[0105] Example 6 A computer program product includes a computer program that, when executed by a processor, implements steps 1, 2, and 4 of embodiments 1 and 2 described above.

[0106] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for rapid measurement of relaxation time in low-field NMR, characterized in that, Includes the following steps: Step 1: Design the IR-bSSFP pulse sequence. The IR-bSSFP pulse sequence consists of a 180° inversion pulse and an excitation pulse sequence. The excitation pulse sequence includes a series of RF excitation pulses, each with a corresponding flip angle. The pseudo-random flip angle; the repetition time between adjacent RF excitation pulses. The settings are based on a random generation method using Perlin noise; the RF phase change rule for adjacent RF excitation pulses is as follows: , For time point number, , They are time point numbers respectively Corresponding RF phase and time point number Corresponding radio frequency phase, The modulo operator, This is the phase increment; Step 2: Based on the Bloch equation model and the IR-bSSFP pulse sequence set in Step 1, simulate the magnetization vector evolution process and calculate the dictionary matrix. ; Step 3: Obtain the component MRF signal sequence of the object under test according to the IR-bSSFP pulse sequence set in Step 1; Step 4: Based on the acquired component MRF signals and dictionary matrix Pattern matching and recognition are performed to obtain the corresponding longitudinal relaxation time of the tested object. The value of the horizontal relaxation time The value of .

2. The method for rapid measurement of low-field NMR relaxation time according to claim 1, characterized in that, The flip angle The value of changes periodically, and in each cycle of the flip angle change, the th The flip angle corresponding to each radio frequency excitation pulse The calculation formula is as follows: ， in, To represent multiplication, The index within a single rotation angle change period is denoted as the index within the period. , Indicates the serial number within the period Corresponding flip angle; time point number The sequence number of the time point of the radio frequency excitation pulse in the excitation pulse sequence, and the sequence number of the same time point. The radio frequency excitation pulse corresponds to a flip angle. and a repeating time Time point number and the serial number within the period The following relationship must be satisfied: ， in, The number of time points covered by the flip angle change cycle, and the time point number. The corresponding flip angle is denoted as Then there is ; rand(5) represents a uniformly distributed random perturbation value generated in the interval [−5°, +5°].

3. The method for rapid measurement of low-field NMR relaxation time according to claim 1, characterized in that, Both the 180° reversal pulse and the radio frequency excitation pulse use BIR-4 adiabatic pulses.

4. The method for rapid measurement of low-field NMR relaxation time according to claim 3, characterized in that, The BIR-4 adiabatic pulse must meet the following requirements: ， ， ， in, Indicates amplitude modulation function, , Represents the frequency modulation function. Represents the phase modulation function, time , satisfy: , , The duration of the pulse. This represents the peak amplitude of the radio frequency magnetic field. and It is a dimensionless constant. Indicates the time point sequence number The corresponding initial radio frequency phase value of the radio frequency excitation pulse. It is the phase offset, and its value is... ; It is the maximum frequency offset.

5. The method for rapid measurement of low-field NMR relaxation time according to claim 1, characterized in that, Step 2 specifically includes the following steps: Step 2.1: Discretize the Bloch equation. The discretized Bloch equation is as follows: ， ， in, Indicates the time point sequence number The corresponding three-dimensional components of the magnetization vector, Indicates the time point sequence number The corresponding magnetization vector before the application of the radio frequency excitation pulse. Indicates the time point sequence number The corresponding magnetization vector after the application of the radio frequency excitation pulse, Indicates the flip angle For rotation angle Axis rotation matrix, defined with static magnetic field The direction that is the same as the direction is axial direction, with The mutually perpendicular directions in the orthogonal transverse plane are respectively shaft and axis; It is a positive phase rotation matrix. This is a reverse phase rotation matrix, and the two are inverse matrices of each other; and These represent the phase rotation matrices before and after the application of the radio frequency (RF) excitation pulse, respectively. They are used to introduce phase modulation of the RF pulse during each excitation process, thereby adjusting the RF phase... With phase increment Variation within the range of 0 to 2π; intermediate quantity , Represents a diagonal matrix, intermediate quantity intermediate quantity , Indicates the time point sequence number The corresponding repetition time; The three-dimensional components of the initial magnetization vector are set as This corresponds to the initial state after the 180° reversal pulse in step 1; To balance the magnetization; Step 2.2: Based on the longitudinal relaxation time The preset value range and horizontal relaxation time The preset value range is set, and multiple pairs of longitudinal relaxation times are configured. With lateral relaxation time The combination of these is denoted as the relaxation time combination. All relaxation time combinations The set of parameters forms the initial relaxation time parameter space. A simulation mesh is established on the initial relaxation time parameter space, and each grid point in the simulation mesh corresponds to a relaxation time combination. ; Step 2.3: At each grid point of the simulation grid, calculate the response process of the magnetization vector under the action of the IR-bSSFP pulse sequence designed in step 1 according to the discrete Bloch equation, and obtain the corresponding complex signal sequence, which is denoted as the evolved MRF signal sequence. Step 2.4: Combine all relaxation times The corresponding evolved MRF signal sequences are stored row by row, and the longitudinal relaxation time is removed. The value is less than the transverse relaxation time. The combination of values ​​forms a dictionary matrix. .

6. The method for rapid measurement of low-field NMR relaxation time according to claim 1, characterized in that, Step 4 includes the following process: Step 4.1: Perform amplitude normalization on the component MRF signal sequence to obtain the normalized component MRF signal sequence. For dictionary matrix The amplitude of each evolving MRF signal sequence is normalized to obtain the normalized evolving MRF signal sequence. ; Step 4.2: Calculate each normalized evolved MRF signal sequence. The amplitude of the complex signal is compared with the normalized component MRF signal sequence. The inner product between the amplitudes of complex signals is used as a similarity index: ， in, Indicates similarity, serving as the matching strength; Indicates the first Time point number in a normalized evolutionary MRF signal sequence The signal value; Indicates the time point number in the normalized component MRF signal sequence. The signal value; Indicates modulo; Step 4.3: In all normalized evolved MRF signal sequences Find similarity The index is the sequence number of the largest normalized evolved MRF signal. And based on the dictionary matrix The sequence number and relaxation time combination of the evolved MRF signal sequence during generation Find the corresponding relationship and locate the index. The corresponding relaxation time combination is denoted as Thus, the longitudinal relaxation time of the measured object is equal to the longitudinal relaxation time. The transverse relaxation time is equal to the transverse relaxation time. .

7. The method for rapid measurement of low-field NMR relaxation time according to claim 1, characterized in that, Step 4 includes the following process: Step 4.1: Normalize the component MRF signal sequence and the evolved MRF signal sequence of each row. The normalization formula is as follows: ， ， in, This represents the normalized component MRF signal sequence. This represents the normalized evolutionary MRF signal sequence. Represents the first element in the dictionary matrix. The evolution of MRF signal sequences, For a normalized dictionary matrix, Indicates the sequence number corresponding to the relaxation time combination; The component MRF signal sequence; Normalized component MRF signal sequence The real and imaginary parts are concatenated to form a real vector, which is used to normalize the evolved MRF signal sequence. The real and imaginary parts of the vector are also concatenated to form a real vector: ， ， in, and These represent the real and imaginary parts of the signal, respectively. Let be the real vector formed by concatenating the real and imaginary parts of the normalized evolved MRF signal sequence, denoted as the concatenated real vector. ; The dictionary matrix representing the concatenation of real and imaginary parts consists of all concatenated real vectors. Arranged in rows This represents a real vector concatenated from the real and imaginary parts of the normalized component MRF signal sequence. Step 4.2: Set the objective function as follows: ， in, For the weight vector, It is a smoothing constraint term. It is a regularization factor. Let be the graph Laplacian matrix constructed from the relaxation time parameter space corresponding to the dictionary matrix. All corresponding relaxation time combinations A set; Indicates minimization; for transpose, for transpose, for Transpose of; Represents the square of the L2 norm; Step 4.3: Solve the objective function to obtain the final weight vector. Non-zero weight elements That is, the weight of each component; then, based on the dictionary matrix... Evolutionary MRF signal sequence and relaxation time combination The correspondence, reverse lookup of the weight vector All non-zero weight elements Corresponding dictionary matrix relaxation time combination This allows us to determine the longitudinal relaxation time of each component of the tested object. The value of the horizontal relaxation time The value of .

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements steps 1, 2, and 4 of the method for rapid measurement of low-field NMR relaxation time as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements steps 1, 2, and 4 of the method for rapid measurement of low-field NMR relaxation time as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements steps 1, 2, and 4 of the method for rapid measurement of low-field NMR relaxation time as described in any one of claims 1 to 7.

Images