A gradient waveform design method of a silent diffusion weighted magnetic resonance imaging sequence

Abstract

The application discloses a gradient waveform design method of a silent diffusion weighted magnetic resonance imaging sequence. The method firstly obtains a frequency response function of a magnetic resonance system through experiment measurement; then, a noise sound pressure level prediction module is constructed based on the measured frequency response function, and an optimized initial waveform generation module is constructed according to a gradient waveform object to be optimized; subsequently, a multi-constraint optimization problem is established by taking the minimum predicted noise A-weighted sound pressure level corresponding to the gradient waveform as an optimization target, and taking hardware limits, waveform shape, gradient matrix and diffusion weighted b value as constraints; finally, the above optimization problem is solved by using a certain constraint optimization solving algorithm, and a silent sequence gradient waveform is obtained. The gradient waveform optimized by the method can be softened on the basis of maintaining original characteristics, so as to suppress the noise generated due to gradient coil vibration during scanning of the magnetic resonance system.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance engineering technology, and in particular to a gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence. Background Technology

[0002] Magnetic Resonance Imaging (MRI) is widely used in imaging examinations for various populations due to its high resolution, high contrast, and non-invasive, non-ionizing radiation characteristics for soft tissues. However, when the gradient coil, the core component of MRI, is operating, the rapidly switching current pulses generate varying Lorentz forces, causing vibrations in the coil structure and consequently, the air within the MRI cavity. These vibrations are perceived as noise by the patient. Depending on the imaging sequence, the acoustic noise sound pressure level (SPL) of an MRI scan can range from 70 to 110 dBA, potentially posing significant safety risks to patients, especially sensitive groups such as pregnant women and infants.

[0003] Diffusion-weighted magnetic resonance imaging (DMRI) is an imaging sequence widely used in clinical and research settings. This sequence combines a high diffusion gradient encoding phase with a rapid gradient switching phase for acquiring echo signals to quickly fill the K-space in echo-planar imaging (EPI) readout, thus generating greater scanning noise compared to other sequences.

[0004] Currently, various methods exist for suppressing noise generated during magnetic resonance imaging (MRI) scans, including providing passive protection for patients such as earplugs and earmuffs, optimizing gradient coil design, using active noise reduction devices, and optimizing the sequence for muting. CN116306017A proposes a design method for low-noise gradient coils, while CN 104765010 A designs a noise shielding device using a vacuum sandwich layer; both can reduce the sound pressure level of MRI noise at the hardware level. CN 113075603 B proposes a method for reducing sequence noise by using machine learning to design parameters for a planar echo sequence based on a half-cycle sine wave as the basic unit. CN 103995243 B uses spline interpolation to round and soften the gradient waveform of the MRI sequence to suppress the noise spectrum.

[0005] However, the research and development of methods for modifying the hardware of magnetic resonance imaging (MRI) equipment is costly and has low compatibility. In contrast, methods for silencing waveforms and timing sequences can achieve better noise reduction while being deployed on different MRI platforms, demonstrating strong versatility. Currently available methods for designing silent sequence gradients, such as directly replacing trapezoidal waves with sine waves, reduce the utilization rate of the gradient system and may lead to a certain degree of degradation in the quality of acquired images. Spline interpolation methods for softening waveforms have weak controllability and target specificity, and cannot introduce sufficient optimization constraints to ensure that the generated waveform is a controllable optimization that maximizes the preservation of imaging quality. Summary of the Invention

[0006] This invention addresses the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences. Based on the frequency response function of the magnetic resonance system, it proposes a method that incorporates multiple controllable constraints to optimize the solution of minimizing the noise gradient waveform by minimizing the A-weighted sound pressure level corresponding to the gradient waveform.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] In a first aspect, the present invention provides a gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence. The method takes the gradient waveform of the diffusion-weighted magnetic resonance imaging sequence as the optimization object, and based on a pre-constructed noise sound pressure level prediction module, takes minimizing the predicted noise A-weighted sound pressure level corresponding to the gradient waveform as the optimization objective. The method establishes a constrained optimization problem with constraints such as hardware limitations, waveform shape, gradient moment and diffusion weight b value, and solves the constrained optimization problem using a constrained optimization solution algorithm based on the initial waveform to obtain the noise-minimized silent sequence gradient waveform.

[0009] The noise sound pressure level prediction module takes a gradient waveform as its input. The module first transforms the gradient waveform in the time domain to the frequency domain using a Fourier transform. Then, it calls the frequency response function (FRF) pre-determined for the magnetic resonance imaging equipment performing the diffusion-weighted magnetic resonance imaging sequence to convert the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal. Finally, it converts the frequency domain representation of the sound signal into the noise A-weighted sound pressure level.

[0010] As a preferred embodiment of the first aspect mentioned above, the noise sound pressure level prediction module has built-in corresponding frequency response functions for the three axes of x, y, and z. During the optimization process, for each axis in the diffusion-weighted magnetic resonance imaging sequence where a gradient waveform exists, it is necessary to call the frequency response function corresponding to that axis to convert the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal.

[0011] As a preferred embodiment of the first aspect above, in the noise sound pressure level prediction module, for any axis i∈{x,y,z}, the corresponding frequency response function H i (f) Obtained through a noise response measurement experiment. The measurement method is as follows: First, place the acoustic sensor at the center of the cavity of the magnetic resonance imaging device performing the diffusion-weighted magnetic resonance imaging sequence, and then apply a sweep frequency sequence g separately along the axis i. i (t), the time-domain sound signal s recorded by the acoustic sensor during the sweep frequency sequence. i (t), and finally the sound signal s is transformed by Fourier transform. i (t) and the sweep sequence g i (t) are transformed into frequency domain representation, and the sound signal s i The frequency domain representation of S(t) i (f) Divide by the sweep frequency sequence g i The frequency domain representation of G(t) i (f) is the corresponding frequency response function H. i (f).

[0012] As a preferred embodiment of the first aspect above, in the noise sound pressure level prediction module, the frequency domain representation of the input gradient waveform is multiplied by the corresponding frequency response function H. i (f) The frequency domain representation of the subsequently converted sound signal S i (f)=G i (f)·H i (f).

[0013] As a preferred embodiment of the first aspect, in the noise sound pressure level prediction module, the frequency domain representation of the sound signal is first processed by A-weighted filtering to calculate the root mean square value of the spectrum, and then the decibel value is taken to obtain the corresponding noise A-weighted sound pressure level.

[0014] As a preferred embodiment of the first aspect above, each of the initial waveforms is a trapezoidal waveform composed of a rising segment, a falling segment, and a plateau period.

[0015] As a preferred embodiment of the first aspect above, the constrained optimization problem is as follows:

[0016]

[0017] Where: M SPL (g(t)) represents the noise sound pressure level prediction module with g(t) as input, where g(t) is the gradient waveform of a single axis in the diffusion-weighted magnetic resonance imaging sequence, which is the object of optimization; G max and S max These represent the maximum gradient magnitude and maximum gradient switching rate, respectively, which are the performance limitations of the gradient system in a magnetic resonance imaging (MRI) device; g p (t) and gq (t) represent the points in the gradient waveform located at the positive and negative amplitude lobes, respectively; T Diff This represents the duration of the gradient waveform; n = 0 indicates the selection of the zeroth-order gradient moment constraint; γ represents the proton. 1 The gyromagnetic ratio of H; b target The target b-value for the diffusion-weighted imaging sequence.

[0018] In a second aspect, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, enables the gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence as described in any of the solutions of the first aspect above.

[0019] Thirdly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, enables the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences as described in any of the solutions of the first aspect above.

[0020] Fourthly, the present invention provides a computer electronic device, which includes a memory and a processor;

[0021] The memory is used to store computer programs;

[0022] The processor is configured to, when executing the computer program, implement the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences as described in any of the first aspects above.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. To address the design problem of silent sequences in magnetic resonance imaging, this invention proposes a gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences based on the frequency response function of noise generated by the magnetic resonance system. This method introduces multiple controllable constraints, such as hardware limitations, waveform shape, gradient moments, and diffusion weighting b-values, with the goal of minimizing the A-weighted sound pressure level corresponding to the waveform. The method establishes a multi-constraint optimization problem and optimizes the solution to minimize the noise gradient waveform.

[0025] 2. This invention achieves noise suppression by optimizing the gradient waveform of the sequence used during scanning in magnetic resonance imaging equipment. Compared with hardware-level noise reduction methods, it is more convenient and less expensive, and can be used in conjunction with hardware methods to achieve better noise reduction results.

[0026] 3. Compared with existing methods that use spline interpolation or replace sinusoidal gradient waveforms for sequence gradient design, the technical solution of this invention can achieve softening while maintaining the original characteristics of the gradient waveform. It can be used for waveform optimization design of diffusion-weighted segments and echo-planar imaging (EPI) segments in diffusion magnetic resonance imaging sequences, thereby suppressing noise generated by gradient coil vibration during magnetic resonance system scanning and improving patient comfort and safety during the scanning process. Attached Figure Description

[0027] Figure 1 The flowchart shows the gradient waveform design method for the silent diffusion-weighted magnetic resonance imaging sequence provided by this invention.

[0028] Figure 2 The frequency response function of each axial coil is obtained from the calculation in step S1.

[0029] Figure 3 A schematic diagram showing the initial waveform of the generated trapezoidal optimization and the upper and lower bound constraints of the optimization problem determined by it.

[0030] Figure 4 This is a set of exemplary comparison charts of waveforms, gradient moments, and predicted noise spectra before and after optimization in an embodiment of the present invention.

[0031] Figure 5 This is another exemplary comparison chart of waveforms, gradient moments, and predicted noise spectra before and after optimization in an embodiment of the present invention. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are only for explaining the technical solutions of this invention and are not intended to limit the invention.

[0033] This invention provides a gradient waveform design method for silent diffusion-weighted magnetic resonance imaging (DWI) sequences. This method is used to suppress noise in the gradient waveform of DWI sequences, thereby obtaining a noise-minimized silent DWI sequence. The specific steps of this design method are as follows:

[0034] Taking the gradient waveform of the diffusion-weighted magnetic resonance imaging sequence as the optimization object, based on a pre-constructed noise sound pressure level prediction module, the optimization objective is to minimize the predicted noise A-weighted sound pressure level corresponding to the gradient waveform. With constraints of hardware limitations, waveform shape, gradient moment, and diffusion weighting b value, a constrained optimization problem is established. Based on the initial waveform, the constrained optimization problem is solved using a constrained optimization solution algorithm to obtain the noise-minimized silent sequence gradient waveform.

[0035] The aforementioned noise sound pressure level prediction module needs to be pre-built before performing gradient waveform design. This allows the module to be directly called during gradient waveform design to predict the corresponding noise A-weighted sound pressure level based on the designed gradient waveform. The core of this noise sound pressure level prediction module is to determine a frequency response function (FRF) for a magnetic resonance imaging (MRI) device performing diffusion-weighted magnetic resonance imaging (DWI) sequences. This FRF converts the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal. The input to this noise sound pressure level prediction module is the gradient waveform. The module first transforms the gradient waveform from the time domain to the frequency domain using a Fourier transform. Then, it calls the pre-determined frequency response function (FRF) for the MRI device performing DWI sequences to convert the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal. Finally, it converts the frequency domain representation of the sound signal into the noise A-weighted sound pressure level.

[0036] In the embodiments of the present invention, considering the independence of the frequency response characteristics of the three directional coils, the noise sound pressure level prediction module incorporates corresponding frequency response functions for the x, y, and z axes respectively. During the optimization process, for each axis in the diffusion-weighted magnetic resonance imaging sequence where a gradient waveform exists, when designing the gradient waveform for that axis, it is necessary to call the corresponding frequency response function to convert the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal.

[0037] The aforementioned frequency response function can be obtained through noise response measurement experiments using a sweep sequence covering the measurement frequency band. Specifically, in the noise sound pressure level prediction module, for any axis i∈{x,y,z}, the corresponding frequency response function H... i (f) Obtained through noise response measurement experiments. The measurement method is as follows: First, place the acoustic sensor at the center of the cavity of the magnetic resonance imaging device performing the diffusion-weighted magnetic resonance imaging sequence, and then apply the sweep frequency sequence g separately along the axis i. i (t), the time-domain sound signal s recorded by the acoustic sensor when a swept frequency sequence (i.e., a sequence of waveforms with a uniformly distributed spectrum) is applied. i (t), and finally the sound signal s is transformed by Fourier transform. i (t) and the sweep sequence g i (t) are transformed into frequency domain representation, and the sound signal s i The frequency domain representation of S(t) i (f) Divide by the sweep frequency sequence g i The frequency domain representation of G(t) i (f) is the corresponding frequency response function. In the parameters above, t in parentheses represents time t in the time domain, and f represents frequency f in the frequency domain, which will not be elaborated further.

[0038] When the above frequency response function H is obtained i After (f), it can be used to predict the noise spectrum generated by any gradient waveform. The calculation method is to multiply the frequency domain representation of the input gradient waveform by the corresponding frequency response function H. i (f) The frequency domain representation of the subsequently converted sound signal S i (f)=G i (f)·H i (f). Frequency domain representation of the noisy sound signal S i (f) After applying A-weighted filtering, calculate the root mean square value of the predicted spectrum, and then take the decibel value to obtain the predicted noise A-weighted sound pressure level corresponding to the waveform from the frequency domain. The specific calculation method for converting the frequency domain representation of the sound signal into the noise A-weighted sound pressure level is existing technology, and can be found in the algorithms and formulas in the existing technology.

[0039] In addition, an initial waveform needs to be constructed before optimization. According to optimization theory, the initial waveform can theoretically be any waveform, and it will be optimized into the required waveform during subsequent constraint optimization. However, considering that the initial waveform has a certain impact on the efficiency of the optimization algorithm and the found optimal solution, and that it facilitates the definition of the upper and lower bound constraints (waveform shape constraints) of the optimization problem, this initial waveform is constructed. Generally, each waveform in the initial waveform is a trapezoidal waveform composed of a rising / falling segment (ramp) and a plateau period (plateau).

[0040] It is also important to note that diffusion-weighted magnetic resonance imaging (DWI) sequences may require trapezoidal waveforms in different axes. Therefore, the waveform design for this invention needs to be performed independently for each axis. For each axis requiring a trapezoidal waveform, the constraint optimization problem includes an optimization objective and corresponding constraints. The optimization objective is to minimize the A-weighted sound pressure level predicted by the noise sound pressure level prediction module for that axis's gradient waveform input. The constraints can be categorized as constraints imposed by hardware performance limitations, waveform shape constraints ensuring the optimized waveform remains similar to the initial optimization waveform, and gradient moment and diffusion-weighted b-value constraints related to the properties of DWI. Therefore, the entire constraint optimization problem can be expressed as follows:

[0041]

[0042] Where: M SPL(g(t)) represents the noise sound pressure level prediction module with g(t) as input, where g(t) is the gradient waveform of a single axis (i.e., the axis currently being optimized) in the diffusion-weighted magnetic resonance imaging sequence; G max and S max These represent the maximum gradient magnitude and maximum gradient switching rate, respectively, which are the performance limitations of the gradient system in a magnetic resonance imaging (MRI) device; g p (t) and g q (t) represents the points in the gradient waveform located on the positive and negative amplitude lobes, respectively (i.e., any point on the positive and negative amplitude lobe waveforms must satisfy their respective conditions); T Diff This represents the duration of the gradient waveform; n = 0 indicates the selection of the zeroth-order gradient moment constraint; γ represents the proton. 1 The gyromagnetic ratio of H; b target The target b-value for the diffusion-weighted imaging sequence.

[0043] It should be noted that the steps of the gradient waveform design method for the above-mentioned silent diffusion-weighted magnetic resonance imaging sequence can essentially be implemented in the form of a computer program.

[0044] Similarly, based on the same inventive concept, the present invention also provides a computer electronic device corresponding to the gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence provided in the above embodiments, which includes a memory and a processor;

[0045] The memory is used to store computer programs;

[0046] The processor is configured to implement the gradient waveform design method for the silent diffusion-weighted magnetic resonance imaging sequence as described above when executing the computer program.

[0047] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0048] Therefore, based on the same inventive concept, the present invention provides a computer-readable storage medium corresponding to the gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence. The storage medium stores a computer program, which, when executed by a processor, can realize the gradient waveform design method for the silent diffusion-weighted magnetic resonance imaging sequence as described above.

[0049] Therefore, based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, can realize the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences as described above.

[0050] It is understood that the aforementioned storage media may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Furthermore, the storage media may also be various media capable of storing program code, such as USB flash drives, external hard drives, magnetic disks, or optical discs.

[0051] It is understood that the processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0052] It should also be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. In the embodiments provided in this application, the division of steps or modules in the system and method is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules or steps may be combined or integrated together, and a module or step may also be split.

[0053] The gradient waveform design method of the silent diffusion-weighted magnetic resonance imaging sequence in the above embodiments will be applied to a specific example to demonstrate its specific design process and technical effect.

[0054] Example

[0055] like Figure 1 As shown, in this embodiment, the gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence is implemented through the following steps:

[0056] S1. Calculate the frequency response function (FRF) of the magnetic resonance system.

[0057] The frequency response function described in this embodiment was experimentally calculated on a 3T magnetic resonance imaging system (Siemens MAGNETOM Prisma) using professional noise measurement equipment, including a measuring microphone, a microphone preamplifier, and a multi-channel signal analyzer. The measuring microphone was calibrated using an acoustic calibrator before the experiment. The performance indicators and parameters after calibration are as follows: standard sound pressure level 93.8 dB, standard frequency 1000 Hz, sensor sensitivity level (dB) -26.3, sensor sensitivity (mV) 48.4172, reference 0 dB 2*10⁻⁵, engineering unit Pa. During the experiment, the measuring equipment sampled the audio signal at a rate of 48000 Hz. After being connected to the leads, the measuring microphone was placed on the magnetic resonance table at the position corresponding to the patient's head and inserted into the center of the scanning cavity.

[0058] The FRF measurement experiment used a sweep sequence with an amplitude of 5 mT / m and a gradient waveform covering a frequency band of 20 to 5000 Hz. Considering the independence of the frequency response characteristics of the coils in the three directions, the sweep sequence was applied separately along each of the three axes, and the corresponding audio response was recorded for each axis. After removing the background noise of the MRI examination room, the results were calculated using the following formulas. Figure 2 The frequency response function of each axial gradient coil in the diagram.

[0059]

[0060] Among them, G i (f) is the input time-domain gradient waveform, i.e., the frequency sweep sequence g. i The frequency domain representation of (t), H i (f) is the frequency response function, S i (f) is the sound signal s i The frequency domain representation of (t) is given by i = x, y, z, which means that gradient fields are applied to the gradient coils along the three axes x, y, z respectively during the experiment.

[0061] S2. Construct a noise sound pressure level prediction module based on the calculated frequency response function.

[0062] Based on the calculated frequency response function, S i (f)=G i (f)·H i (f) The noise spectrum generated by any gradient waveform can be predicted. After applying A-weighted filtering to the predicted noise spectrum, the root mean square of the spectrum is calculated according to the following formula, and the decibel value is obtained. Then, the A-weighted sound pressure level of the predicted noise corresponding to the waveform can be obtained from the frequency domain:

[0063]

[0064] In the above formula, the calculated spectrum is the root mean square value of the frequency bands f1 to f2 in the single-sided spectrum, expressed in units of sound pressure amplitude Pa. A0 is the A-weighted filtered spectral amplitude corresponding to f1. k It is the A-weighted filtered spectral amplitude corresponding to f2; the following formula calculates the sound pressure level in decibels, p ref This represents the commonly used reference sound pressure level in air, 20 μPa.

[0065] S3. Construct the initial waveform for optimization based on the gradient waveform object to be optimized.

[0066] The gradient waveform object optimized in this embodiment is the diffusion-weighted segment gradient waveform in the Oscillating Gradient Spin Echo (OGSE) sequence, which is a trapezoidal waveform that oscillates periodically and satisfies a specific fundamental frequency and diffusion-weighted b value.

[0067] Based on prior knowledge of OGSE gradient waveforms, the trapezoidal OGSE gradient waveform is defined as a combination of rising / falling segments (ramp) and plateau periods. The plateau duration is calculated according to the input OGSE fundamental frequency (60Hz in this embodiment), and the rising / falling segments are combined with the plateau to generate a lobe corresponding to the main frequency. Then, according to the b-value requirement, it starts to increase from one cycle. If the target b-value is not met, one cycle is added until the target b-value is met. Figure 3 In the middle, r and p are the durations of the rising / falling segment and plateau phase used by the two quarter-period lobes, and r′ and p′ are the durations of the rising / falling segment and plateau phase used by the half-period lobes. The relationship between the two is (p′+r′)=2·(p+r).

[0068] The initial waveform g of the OGSE gradient waveform generated in this embodiment trape (t) such as Figure 3 As shown in the diagram. This initial waveform will define the constraints in the subsequent optimization problem, and therefore it naturally satisfies all constraints.

[0069] S4. With the optimization objective of minimizing the predicted noise A-weighted sound pressure level corresponding to the gradient waveform, and constrained by hardware limitations, waveform shape, gradient moments, and dispersion weighting b, a multi-constraint optimization problem is established.

[0070] In this embodiment, the following formula is selected as the objective function.

[0071] min M SPL (g(t))

[0072] Among them, M SPL(g(t)) represents the noise sound pressure level prediction module with g(t) as input, where g(t) is the gradient waveform of a single axis in the diffusion-weighted magnetic resonance imaging sequence, which is the object of optimization; g(t)∈{g x (t),g y (t),g z (t)}, specifically optimizing which axis M of the sequence. SPL In (g(t)), the frequency response function H of this axis needs to be called. i (f).

[0073] In this embodiment, the constraints can be divided into constraints caused by hardware performance limitations, waveform shape constraints that control the optimized waveform to be similar to the optimized initial waveform, and gradient moment and diffusion weighting b-value constraints related to the properties of diffusion-weighted magnetic resonance imaging.

[0074] In this embodiment, the maximum gradient magnitude G is limited by the performance of the gradient system of the magnetic resonance imaging device. max With maximum gradient switching rate S max The hardware constraints are shown in the following equation, and can be defined by upper and lower bound constraints and linear inequality constraints:

[0075]

[0076] The shape of the OGSE oscillation waveform is constrained as shown in the following equation, which can be defined by upper and lower bound constraints:

[0077] 0 + <g p (t)≤G max

[0078] -G max ≤g q (t)<0 -

[0079] Where g p (t) and g q (t) correspond to the points in the waveform to be optimized located on the positive and negative amplitude lobes, respectively. It should be noted that each point on the positive and negative amplitude lobes must satisfy the above constraints.

[0080] Considering that practical optimization programs generally implement the upper and lower bound constraints and linear inequality constraints in the above hardware constraints and waveform shape constraints in vector form, in this embodiment, the vector representations of the above upper and lower bound constraints and linear inequality constraints can be adjusted to the following forms respectively:

[0081] lb≤g(t)≤ub

[0082] A·g(t)≤b

[0083] In this embodiment, the oscillation of the waveform and the satisfaction of G can be controlled by setting a reasonable upper bound ub and a lower bound lb for the waveform. max Restrictions, settings as follows Figure 3 As shown in the figure; while A·g(t)≤b is used to constrain the first-order derivative of the optimized waveform to not exceed the maximum switching rate S of the hardware device. max Its detailed matrix representation is as follows:

[0084]

[0085] In the formula, n is the length of the optimized waveform. Matrix A is the difference matrix between the preceding and following waveforms; if A·g(t)≤b, then the difference between the preceding and following waveforms g(t) is controlled to be less than S. max .

[0086] Furthermore, considering that practical optimization procedures are generally implemented in vector or matrix form, the gradient moment constraint in this embodiment is shown in the following equation, which can be defined by the following linear equality constraint:

[0087]

[0088] Where T Diff To determine the duration of the diffusion-weighted gradient field, since diffusion weighting requires the zeroth-order momentum of the gradient waveform to be zero, n = 0 must be chosen to set the zeroth-order gradient moment to zero. Under the constraint of setting higher-order gradient moments to zero (Gradient Moment Nulling), such as n = 1, 2..., the sensitivity to motion artifacts can be reduced, achieving higher-order motion compensation. When the gradient moments are other fixed constants, the phase accumulation effect of the gradient field on the proton spin can be kept constant. For example, when the optimization object is the gradient waveform in the Echo Plane Imaging (EPI) stage, the area under the integral of its single-lobe gradient waveform (zeroth-order moment) can be limited to a certain value.

[0089] Similarly, practical optimization procedures are generally implemented in vector or matrix form. In this embodiment, the gradient moment constraint is represented as a matrix:

[0090] Aeq·g(t)=beq

[0091]

[0092] In the formula, n is the length of the optimized waveform. Specifically, the first row of Aeq calculates the zeroth moment, the second row calculates the first moment, and beq is a vector of all zeros.

[0093] For the diffusion-weighted gradient waveform, a b-value (diffusion weighting coefficient) constraint needs to be introduced to ensure that the gradient waveform can undergo normal diffusion weighting. The b-value constraint is shown in the following formula and can be defined by nonlinear equality constraints:

[0094]

[0095] Where γ is a proton 1 The gyromagnetic ratio of H, T Diff For the duration of the diffusion-weighted gradient, b target The target b-value for the diffusion-weighted imaging sequence. The b-value constraint is achieved through a nonlinear equality constraint ceq(x) = 0, that is, ceq is modified according to the definition of the b-value as follows:

[0096]

[0097] It should be noted that the vectorized forms of the various constraints described above in this embodiment are only to meet the computational requirements of the optimization program or related software, and are not necessary limitations on the present invention.

[0098] S5. Using a constrained optimization algorithm, the noise optimization gradient waveform is obtained by solving the above optimization problem.

[0099] In this embodiment, the interior-point method or sequential quadratic programming (SQP) can be used to optimize and solve the following multi-constraint optimization problem:

[0100]

[0101] In this embodiment, if the fundamental frequency of the OGSE waveform to be optimized is 60Hz, the target b value for a single axis is 210s / mm. 2 Maximum gradient magnitude G max The maximum gradient switching rate is 80mT / m. max The waveform is 73 T / m / s. The predicted A-weighted sound pressure level before optimization is 82.09 dBA. The waveform after iterative optimization is as follows: Figure 4 As shown, the predicted A-weighted sound pressure level dropped to 71.42 dBA.

[0102] In this embodiment, if the fundamental frequency of the OGSE waveform to be optimized is 60Hz, the single-axis target b value is 1000s / mm. 2 Maximum gradient magnitude G max The maximum gradient switching rate is 200mT / m. max The sound pressure level is 300 T / m / s. The predicted A-weighted sound pressure level before optimization is 95.46 dBA. The waveform after iterative optimization is as follows: Figure 5 As shown, the predicted A-weighted sound pressure level dropped to 74.27 dBA.

[0103] As can be seen from the above embodiments, the predicted noise spectrum is sufficiently suppressed, the gradient waveform is softened on the basis of the original trapezoid, and the A-weighted sound pressure level of the predicted noise can be reduced by ten to twenty decibels.

[0104] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A method for designing gradient waveforms in a silent diffusion-weighted magnetic resonance imaging sequence, characterized in that: Taking the gradient waveform of the diffusion-weighted magnetic resonance imaging sequence as the optimization object, based on the pre-constructed noise sound pressure level prediction module, the optimization objective is to minimize the predicted noise A-weighted sound pressure level corresponding to the gradient waveform. With constraints of hardware limitations, waveform shape, gradient moment and diffusion weight b value, a constrained optimization problem is established. Based on the initial waveform, the constrained optimization problem is solved using a constrained optimization solution algorithm to obtain the noise-minimized silent sequence gradient waveform. The input to the noise sound pressure level prediction module is a gradient waveform. The module first transforms the gradient waveform in the time domain to the frequency domain through Fourier transform, then calls the frequency response function pre-determined for the magnetic resonance imaging equipment performing the diffusion-weighted magnetic resonance imaging sequence to convert the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal, and finally converts the frequency domain representation of the sound signal into the noise A-weighted sound pressure level.

2. The gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence according to claim 1, characterized in that, In the noise sound pressure level prediction module, corresponding frequency response functions are built-in for the three axes x, y, and z. During the optimization process, for each axis in the diffusion-weighted magnetic resonance imaging sequence where a gradient waveform exists, the frequency response function corresponding to that axis needs to be called to convert the frequency domain representation of the gradient waveform into the frequency domain representation of the sound signal.

3. The gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence according to claim 2, characterized in that, In the noise sound pressure level prediction module, for any axis i∈{x,y,z}, the corresponding frequency response function H i (f) Obtained through a noise response measurement experiment. The measurement method is as follows: first, the acoustic sensor is placed at the center of the cavity of the magnetic resonance imaging device executing the imaging sequence, and then a sweep frequency sequence g is applied separately along the axial direction i. i (t), the time-domain sound signal s recorded by the acoustic sensor during the sweep frequency sequence. i (t), and finally the sound signal s is transformed by Fourier transform. i (t) and the sweep sequence g i (t) are transformed into frequency domain representation, and the sound signal s i The frequency domain representation of S(t) i (f) Divide by the sweep frequency sequence g i The frequency domain representation of G(t) i (f) is the corresponding frequency response function H. i (f).

4. The gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence according to claim 1, characterized in that, In the noise sound pressure level prediction module, the frequency domain representation of the input gradient waveform is multiplied by the corresponding frequency response function H. i (f) The frequency domain representation of the subsequently converted sound signal S i (f)=G i (f)·H i (f).

5. The gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence according to claim 1, characterized in that, In the noise sound pressure level prediction module, the frequency domain representation of the sound signal is first processed by A-weighted filtering, and then the root mean square value of the spectrum is calculated. Finally, the decibel value is taken to obtain the corresponding noise A-weighted sound pressure level.

6. The gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence according to claim 1, characterized in that, In the initial waveform, each waveform is a trapezoidal waveform composed of a rising segment, a falling segment, and a plateau period.

7. The gradient waveform design method for a silent diffusion-weighted magnetic resonance imaging sequence according to claim 1, characterized in that, The constrained optimization problem is as follows: Where: M SPL (g(t)) represents the noise sound pressure level prediction module with g(t) as input, where g(t) is the gradient waveform of a single axis in the diffusion-weighted magnetic resonance imaging sequence, and g(t)∈{g x (t),g y (t),g z (t)};G max and S max These represent the maximum gradient magnitude and maximum gradient switching rate, respectively, which are the performance limitations of the gradient system of the magnetic resonance imaging device; g p (t) and g q (t) represents the points in the gradient waveform located at the positive and negative amplitude lobes, respectively; T Diff This represents the duration of the gradient waveform; n = 0 indicates the selection of the zeroth-order gradient moment constraint; γ represents the proton. 1 The gyromagnetic ratio of H; b target The target b-value for the diffusion-weighted imaging sequence.

8. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it can implement the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences as described in any one of claims 1 to 7.

10. A computer electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to, when executing the computer program, implement the gradient waveform design method for silent diffusion-weighted magnetic resonance imaging sequences as described in any one of claims 1 to 7.

Images