Ultra-low field magnetic resonance imaging noise suppression method and ultra-low field magnetic resonance imaging system

Abstract

The application relates to an ultra-low-field magnetic resonance imaging noise suppression method and an ultra-low-field magnetic resonance imaging system. The method comprises the following steps: acquiring original magnetic resonance signals of a main imaging signal channel and reference noise signals of a plurality of reference channels; dividing time periods according to a scanning sequence and generating synchronization control signals corresponding to different time periods; in response to a first synchronization control signal, entering a training mode in a non-signal acquisition window of the scanning sequence, wherein the training mode comprises updating weight coefficients of an adaptive filter set based on the reference noise signals and the original magnetic resonance signals; in response to a second synchronization control signal, entering a scanning mode in a signal acquisition window of the scanning sequence, wherein the scanning mode comprises freezing or reducing the update speed of the weight coefficients of the adaptive filter set, and using the updated weight coefficients to perform noise cancellation on the original magnetic resonance signals and output purified magnetic resonance signals. The image signal-to-noise ratio of the ultra-low-field magnetic resonance imaging is improved.

Description

Technical Field

[0001] This application relates to the field of ultra-low field magnetic resonance imaging, and in particular to an ultra-low field magnetic resonance imaging noise suppression method and an ultra-low field magnetic resonance imaging system. Background Technology

[0002] Ultra-low field magnetic resonance imaging (ULF-MRI, field strength ≤0.1T), with its open structure, offers advantages such as high patient comfort, low system manufacturing cost and operating power consumption, and no ionizing radiation. It has shown great application potential in primary healthcare, mobile diagnosis and treatment, and intraoperative monitoring. However, its clinical promotion has long been limited by a fundamental drawback—a significantly low image signal-to-noise ratio (SNR). This is mainly due to the following factors: First, the magnetic resonance signal intensity and the static magnetic field strength... The signal strength is proportional to the square of the magnetic field strength. For example, at a field strength of 0.05T, the signal strength is only about 1 / 900 of that of a 1.5T system, indicating that the signal source itself is extremely weak. Secondly, open structures cannot employ a fully magnetically shielded room like high-field closed MRI systems, making the system highly susceptible to electromagnetic noise from the external environment (such as power frequency interference, broadcast signals, and transients during equipment switching). Furthermore, the strong electromagnetic interference generated by high-power components inside the system (such as gradient amplifiers and RF power amplifiers) during operation can contaminate the already extremely weak MRI signal through conduction, radiation, and coupling, further degrading the imaging quality.

[0003] Therefore, related technologies mainly focus on two paths for improvement: First, optimization of the magnet and mechanical structure (e.g., CN113848520A) is emphasized, reducing costs and improving openness by improving permanent magnet materials, array arrangement, and magnetic circuit design. However, this type of solution only focuses on the static magnetic field generation module and does not address noise suppression in the signal chain, resulting in limited improvement in image quality. Second, improvements are made to local links in the imaging chain (e.g., CN107110930A). For example, improving magnetic field uniformity by adding shimming coils or using traditional filtering and digital averaging in signal processing. While these methods address image quality to some extent, their improvement methods are relatively isolated and one-sided, and performance improvement faces significant bottlenecks under extremely weak signal conditions at very low field levels.

[0004] Currently, there is no effective solution to the problem of low signal-to-noise ratio in ultra-low field magnetic resonance imaging. Summary of the Invention

[0005] Therefore, it is necessary to provide a noise suppression method and system for ultra-low field magnetic resonance imaging that can improve the signal-to-noise ratio of images, in order to address the aforementioned technical problems.

[0006] In a first aspect, this application provides a method for suppressing noise in ultra-low field magnetic resonance imaging, comprising:

[0007] Acquire the raw magnetic resonance signal from the main imaging signal channel, and the reference noise signals from multiple reference channels corresponding to noise sensors;

[0008] The scanning sequence is divided into time periods, and synchronization control signals corresponding to different time periods are generated.

[0009] In response to the first synchronization control signal, during the non-signal acquisition window of the scan sequence, a training mode is entered, the training mode including: updating the weight coefficients of the adaptive filter bank based on the reference noise signal and the original magnetic resonance signal;

[0010] In response to the second synchronization control signal, the scanning mode is entered in the signal acquisition window of the scanning sequence. The scanning mode includes: freezing or reducing the update speed of the weight coefficients of the adaptive filter bank, and using the updated weight coefficients to cancel noise in the original magnetic resonance signal, and outputting a purified magnetic resonance signal.

[0011] In some embodiments, time periods are divided according to the scan sequence, and synchronization control signals corresponding to different time periods are generated, including:

[0012] The protocol timing of the scan sequence is analyzed to determine the timing of radio frequency transmission, gradient switching, and signal acquisition in each pulse sequence;

[0013] Based on the timing sequence of radio frequency transmission, gradient switching, and signal acquisition in each pulse sequence, the signal acquisition blank period, the pre-scanning stage, and the time boundaries between the signal acquisition windows are determined.

[0014] The first synchronization control signal and the second synchronization control signal are generated based on the time boundary between the signal acquisition blank period, the pre-scanning stage, and the signal acquisition window.

[0015] In some embodiments, the adaptive filter bank includes multiple adaptive filters, each constructed based on a normalized least mean square algorithm, and each adaptive filter corresponds to a reference channel; the weight coefficients of the adaptive filter bank are updated based on the reference noise signal and the original magnetic resonance signal, including:

[0016] The weight coefficients of the adaptive filter bank are iteratively updated with the goal of minimizing the power of the output error signal.

[0017] In some embodiments, the iterative update formula for the weight coefficients of the adaptive filter bank is:

[0018] ;

[0019] Where n is the index of the reference channel, w i (n) represents the coefficient vector of the adaptive filter corresponding to the i-th reference channel at time n, r i e(n) is the i-th reference noise signal vector, e(n) is the error signal at time n, μ is the step size factor, and δ is the regularization factor.

[0020] In some embodiments, freezing or reducing the update rate of the weight coefficients of the adaptive filter bank includes:

[0021] When entering the scanning mode, the value of the step size factor is controlled so that the value of the step size factor in the scanning mode is less than the value of the step size factor in the training mode.

[0022] In some embodiments, the method further includes: querying a pre-stored pattern database before executing the current scan sequence; wherein the pattern database stores an initial set of weight coefficients associated with different scan sequence types and scan parameters;

[0023] The initial weight coefficient set obtained from the query is loaded as the initial weight of the adaptive filter.

[0024] Secondly, this application provides an ultra-low field magnetic resonance imaging system, comprising:

[0025] The main magnet is used to generate a low-intensity static magnetic field in the imaging area;

[0026] Gradient coils are used to superimpose spatially encoded gradient magnetic fields onto the static magnetic field;

[0027] The radio frequency module is used to transmit radio frequency pulses to excite nuclear magnetic resonance and receive the induced magnetic resonance signals;

[0028] A power distribution unit is used to supply power to the main magnet, the gradient coil, and the radio frequency module;

[0029] The main imaging signal channel is used to output the raw magnetic resonance signal;

[0030] A noise sensor array, comprising multiple noise sensors, each used to acquire reference noise signals from different noise sources;

[0031] The control module is used to acquire the original magnetic resonance signal and the reference noise signal, and execute the ultra-low field magnetic resonance imaging noise suppression method described in the first aspect above, and output the purified magnetic resonance signal.

[0032] In some embodiments, the control module includes: an adaptive filter bank and a sequence controller; wherein the adaptive filter bank includes a plurality of adaptive filters, each used to estimate the noise propagation path characteristics from each interference source to the main imaging signal channel based on the input reference noise signal;

[0033] The sequence controller is used to divide the time period according to the scanning sequence, generate a first synchronization control signal corresponding to the non-signal acquisition window of the scanning sequence and a second synchronization control signal corresponding to the signal acquisition window of the scanning sequence, and send the synchronization control signal to the adaptive filter bank to control the adaptive filter bank to switch the working mode.

[0034] In some embodiments, the power distribution unit includes:

[0035] The system includes a double-shielded isolation transformer, a hybrid active filter module, a power distribution monitoring module, and a power distribution backplane. The power distribution backplane comprises a multi-layer PCB, and the double-shielded isolation transformer, the hybrid active filter module, and the power distribution monitoring module are respectively disposed on different PCB layers of the power distribution backplane.

[0036] In some embodiments, the noise sensor group includes:

[0037] The first sensor is located outside the gradient coil and is used to collect the time-varying magnetic field noise generated when the gradient current switches.

[0038] The second sensor is located inside the cavity of the main magnet and is used to collect low-frequency magnetic field noise in the environment.

[0039] The third sensor is located near the entrance of the power distribution unit and is used to collect grid coupling and spatial radiation interference noise.

[0040] The aforementioned ultra-low field magnetic resonance imaging (ULMI) noise suppression method and system construct a noise observation and cancellation chain parallel to the main imaging signal chain by distributively acquiring reference noise signals from different noise sources. It estimates the noise propagation path characteristics from each interference source to the main imaging signal channel and matches the magnetic resonance scanning sequence synchronization mechanism. During the non-signal acquisition window of the scanning sequence, it enters the training mode to quickly update the filter weight coefficients. During the signal acquisition window of the scanning sequence, it freezes or reduces the update speed of the weight coefficients. The trained model is used to cancel noise in the original magnetic resonance signal. This improves the image signal-to-noise ratio of ULMI while avoiding the modulation of weak magnetic resonance signals by continuous adaptive filtering and promoting the convergence of the adaptive filter bank. Attached Figure Description

[0041] Figure 1This is a schematic diagram of the structure of an ultra-low field magnetic resonance imaging system in one embodiment;

[0042] Figure 2 This is a flowchart illustrating a method for suppressing noise in ultra-low field magnetic resonance imaging in one embodiment.

[0043] Figure 3 This is a schematic diagram of the structure of an ultra-low field magnetic resonance imaging system in another embodiment;

[0044] Figure 4 This is a schematic diagram of the structure of an ultra-low field magnetic resonance imaging system in another embodiment;

[0045] Figure 5 This is a schematic diagram of the power distribution unit in one embodiment. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0047] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning understood by one of ordinary skill in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these” used in this application do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to these processes, methods, products, or devices. Words such as “connected,” “linked,” and “coupled” used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. Normally, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific order of objects.

[0048] The related technologies generally suffer from the following common problems: Their anti-interference methods are passive and static, primarily relying on physical shielding, isolation, and fixed-parameter filtering—a defensive approach that struggles to adapt to the complex, dynamically changing interference encountered during scanning. The technical solutions are fragmented, lacking a global, collaborative design approach that considers the magnet, gradient, radio frequency, power supply, and receiving circuitry as a complete electromagnetic compatibility system. Each module is optimized independently, failing to systematically block or cancel noise coupling paths. Furthermore, the systems lack intelligence and adaptability; fixed parameter settings prevent dynamic adjustments based on the real-time scanning environment, sequence type, and patient condition. This results in image quality becoming overly dependent on operator experience and ideal environmental conditions, limiting the clinical applicability and stability of ultra-low field MRI equipment.

[0049] To address the technical problems existing in the related technologies, in one embodiment, Figure 1 A schematic diagram of an ultra-low field magnetic resonance imaging system is provided, which includes:

[0050] The main magnet is used to generate a low-intensity static magnetic field in the imaging area. Taking a 0.05T open permanent magnet whole-body MRI system as an example, the main magnet can be an open magnetic array made of SmCo permanent magnet material with a central field strength of 0.05T.

[0051] Gradient coils are used to superimpose spatially encoded gradient magnetic fields onto a static magnetic field. For example, a passive shielded gradient coil can be used with a maximum gradient intensity of 30 mT / m and a switching rate of 100 T / m / s.

[0052] The radio frequency (RF) module is used to transmit RF pulses to excite nuclear magnetic resonance (NMR) and receive the induced magnetic resonance signals. For example, a transceiver head coil is used, with an operating frequency of approximately 2.13 MHz.

[0053] The power distribution unit supplies power to the main magnet, gradient coil, and RF module. In one implementation, the power distribution unit uses a power cabinet, with an input of 220V, 50Hz single-phase AC power and a maximum input current of 25A. Its output can be divided into multiple channels, as follows:

[0054] Output path 1: 150VDC / 10A, for gradient power amplifier.

[0055] Output path 2: 48VDC / 5A, for RF power amplifier.

[0056] Output path 3: AC220V / 5A, for the spectrometer and control unit.

[0057] Output path 4: 15VDC / 2A, for the front-end box and data acquisition card.

[0058] Output path 5: 24VDC / 3A, for other auxiliary equipment.

[0059] The main imaging signal channel is used to output the raw magnetic resonance signal. The main imaging signal channel refers to the channel from the RF receiving coil via the preamplifier, and its output signal is the raw magnetic resonance signal d(t).

[0060] The noise sensor array comprises multiple noise sensors, each used to acquire reference noise signals from different noise sources. These sources include gradient coils generating gradient switching noise, ambient background magnetic field noise within the magnet imaging area, and power distribution unit inlet introducing conducted noise from the power grid. Each sensor serves as an independent reference channel, and its output signal is the reference noise signal r. i (t). The reference noise signal includes gradient switching noise, power grid conducted noise, and background magnetic field noise from the magnet imaging environment.

[0061] In one implementation, the noise sensor array includes:

[0062] The first sensor, located outside the gradient coil, is used to collect the time-varying magnetic field noise generated during gradient current switching. For example, a miniature Rogowski coil (e.g., 2 cm in diameter, 10 turns) can be placed near the outer region of the gradient coil.

[0063] The second sensor, located inside the cavity of the main magnet, is used to collect low-frequency magnetic field noise from the environment. For example, a small fluxgate sensor probe can be placed inside the cavity of the main magnet, close to the imaging area but far from the gradient coil area.

[0064] The third sensor, located near the power distribution unit inlet, is used to collect grid coupling and spatial radiated interference noise. For example, an active electric field probe (e.g., frequency range DC-30MHz) can be placed inside the power distribution unit.

[0065] The control module acquires the raw magnetic resonance signal and the reference noise signal, executes an ultra-low field magnetic resonance imaging noise suppression method, and outputs the purified magnetic resonance signal. The control module can employ a processing unit based on programmable logic (such as an FPGA) or a high-speed digital signal processor (DSP).

[0066] Figure 2 This is a flowchart illustrating the ultra-low field magnetic resonance imaging noise suppression method in this embodiment, which includes the following steps:

[0067] Step S101: Acquire the raw magnetic resonance signal from the main imaging signal channel and the reference noise signal from multiple reference channels corresponding to noise sensors.

[0068] The main imaging signal channel refers to the channel output from the radio frequency receiving coil via the preamplifier, and its output signal is the original magnetic resonance signal d(t).

[0069] Each sensor acts as an independent reference channel, and its output signal is the reference noise signal r. i (t). The reference noise signal includes gradient switching noise, power grid conducted noise, and background magnetic field noise from the magnet imaging environment.

[0070] Step S102: Divide the time period according to the scanning sequence and generate synchronization control signals corresponding to different time periods.

[0071] The time period can be divided into a non-signal acquisition window of the scan sequence and a signal acquisition window of the scan sequence. The non-signal acquisition window of the scan sequence includes a signal acquisition gap (i.e., sequence time) or a pre-scan phase. As one implementation, the timing sequence of radio frequency transmission, gradient switching, and signal acquisition in each pulse sequence can be determined by parsing the protocol timing of the scan sequence; based on the timing sequence of radio frequency transmission, gradient switching, and signal acquisition in each pulse sequence, the time boundaries between the signal acquisition gap, the pre-scan phase, and the signal acquisition window can be determined; and based on the time boundaries between the signal acquisition gap, the pre-scan phase, and the signal acquisition window, a first synchronization control signal and a second synchronization control signal can be generated respectively.

[0072] Step S103: In response to the first synchronization control signal, enter the training mode in the non-signal acquisition window of the scanning sequence. The training mode includes updating the weight coefficients of the adaptive filter bank based on the reference noise signal and the original magnetic resonance signal.

[0073] In this step, the adaptive filter bank includes multiple adaptive filters, each used to estimate the noise propagation path characteristics from each interference source to the main imaging signal channel based on the input reference noise signal. This involves constructing a noise propagation model of the physical and electrical paths traversed by each interference source as it propagates from its reference channel to the main imaging signal channel. The noise propagation path characteristics or noise propagation model involve amplitude attenuation, phase delay, or frequency response, among other factors.

[0074] The adaptive filters are constructed based on the normalized least mean square algorithm, with each adaptive filter corresponding to a reference channel. As one implementation method, the weight coefficients of the adaptive filter bank can be iteratively updated with the objective of minimizing the power of the output error signal (i.e., minimizing residual interference). The iterative update formula is as follows:

[0075] ;

[0076] Where n is the index of the reference channel, w i (n) represents the coefficient vector of the adaptive filter corresponding to the i-th reference channel at time n, r ie(n) is the i-th reference noise signal vector, e(n) is the error signal at time n, μ is the step size factor used to control the convergence speed and stability, and δ is the regularization factor used to prevent the denominator from being zero.

[0077] In this formula, the reference noise signal r for each reference channel is... i (n) is considered an interference source and is processed by an adaptive filter W. i (z) is used to estimate the noise propagation path characteristics from each interference source to the main imaging signal channel. Adaptive filter W i The weight coefficients of (z) are iteratively updated to approximate the transfer function of its true path. When W i After (z) converges, its response to the reference noise signal r i The filtered output y of (n) i (n) represents the best estimate of the noise component caused by this interference source in the main imaging signal channel. This is achieved by subtracting all y values ​​from the original magnetic resonance signal. i (n) can achieve noise cancellation. Compared with traditional fixed filtering, this embodiment can dynamically track path characteristic drift caused by temperature changes, component aging or patient movement.

[0078] In step S104, in response to the second synchronization control signal, the scanning mode is entered in the signal acquisition window of the scanning sequence. The scanning mode includes: freezing or reducing the update speed of the weight coefficients of the adaptive filter bank, and using the updated weight coefficients to cancel noise in the original magnetic resonance signal and output the purified magnetic resonance signal.

[0079] In this step, based on the updated weight coefficients of the adaptive filter bank, the reference noise signal is filtered to synthesize a total interference estimate. This total interference estimate is then subtracted from the original magnetic resonance signal to obtain and output the purified magnetic resonance signal. Specifically, each filter outputs y... i (n) represents the estimated corresponding interference components existing in the main imaging signal channel, and all y i The total interference estimate y(n) is obtained by summing (n), and then subtracted from the original magnetic resonance signal d(t) in real time to obtain the error signal e(n), which is the final purified magnetic resonance signal output.

[0080] In steps S101 to S104 above, a noise observation and cancellation chain parallel to the main imaging signal chain is constructed by distributively acquiring reference noise signals from different noise sources. The noise propagation path characteristics from each interference source to the main imaging signal channel are estimated, and the magnetic resonance scanning sequence synchronization mechanism is matched. The training mode is entered during the non-signal acquisition window of the scanning sequence to quickly update the filter weight coefficients, while the update speed of the weight coefficients is frozen or reduced during the signal acquisition window of the scanning sequence. The trained model is used to cancel noise in the original magnetic resonance signal. This improves the signal-to-noise ratio of ultra-low field magnetic resonance imaging, avoids the modulation of weak magnetic resonance signals by continuous adaptive filtering, and promotes the convergence of the adaptive filter bank.

[0081] In one embodiment, Figure 3 A schematic diagram of another ultra-low field magnetic resonance imaging system is provided. In this system, the control module includes an adaptive filter bank and a sequence controller.

[0082] The adaptive filter bank includes multiple adaptive filters, each used to estimate the noise propagation path characteristics from each interference source to the main imaging signal channel based on the input reference noise signal. In step S104, freezing or reducing the update speed of the weight coefficients of the adaptive filter bank includes: when entering the scanning mode, controlling the value of the step size factor μ so that the value of the step size factor μ in the scanning mode is less than the value of the step size factor μ in the training mode.

[0083] The sequence controller is used to divide the time period according to the scan sequence, generate a first synchronization control signal corresponding to the non-signal acquisition window of the scan sequence and a second synchronization control signal corresponding to the signal acquisition window of the scan sequence, and send the synchronization control signal to the adaptive filter bank to control the adaptive filter bank to switch the working mode.

[0084] In this embodiment, to optimize noise suppression and adapt to the intermittent nature of MRI operation, a sequence synchronization controller is introduced. Its control logic is divided into training mode and scanning mode.

[0085] During the non-signal acquisition window of the scanning sequence, such as the blank period after radio frequency transmission and before signal acquisition, or the pre-scanning stage, the system enters training mode. At this time, the weight coefficient update of the adaptive filter bank is enabled. Since there is no useful MRI signal and the main imaging signal channel is mainly noise, the noise transmission model from each interference source to the main imaging signal channel can be established quickly and accurately.

[0086] During the signal acquisition window of the scanning sequence, i.e., the period when MRI signals are actually acquired, the system enters scanning mode. The update rate of the weight coefficients of the adaptive filter bank is frozen or significantly reduced. At this time, the system uses a pre-trained noise propagation model to perform real-time noise cancellation on the raw MRI signal, avoiding distortion of the weak MRI signal caused by the dynamic adjustment of the weight coefficients.

[0087] Considering that traditional adaptive filtering in the communication field generally operates continuously, while MRI signal acquisition is intermittent and pulse-based, this embodiment utilizes the non-signal acquisition window of the scan sequence for model training, and applies a noise propagation model within the signal acquisition window of the scan sequence. This setup allows the system to match the operating timing of the MRI, resolving the problem that adaptive algorithms may have difficulty converging or modulate the signal when a useful signal is present.

[0088] In one embodiment, Figure 4 A schematic diagram of another ultra-low field magnetic resonance imaging system is provided. Figure 3 In addition to the above, the system also includes a pre-amplifier array and a synchronous ADC array. The noise sensor group amplifies the signals from each sensor by inputting them to the pre-amplifier array, and then, together with the raw magnetic resonance signal from the main imaging signal channel, it is acquired by the synchronous ADC array and input to an adaptive filter bank. The adaptive filter bank contains adaptive filters W1-W1 corresponding to each sensor. n The sequence controller is used to send synchronization control signals to the adaptive filter bank to control the switching of its operating modes.

[0089] In some embodiments, the above-described ultra-low field magnetic resonance imaging noise suppression method further includes: querying a pre-stored pattern database before executing the current scan sequence; wherein the pattern database stores an initial weight coefficient set associated with different scan sequence types and scan parameters; loading the queried initial weight coefficient set as the initial weights of the adaptive filter to further accelerate the convergence of the training pattern.

[0090] In this embodiment, the system accumulates the weight coefficients of adaptive filters under different scan sequences (such as spin echo (SE), gradient echo (GRE), etc.) and different gradient modes, storing them as corresponding initial weight coefficient sets to obtain a pattern database. At the start of a new scan, parameters from similar scenarios can be directly called as initial values ​​for the weight coefficients based on the pattern database, enabling the system to start warmly, reducing training time, and improving scan efficiency and stability between different scans.

[0091] In one embodiment, Figure 5A schematic diagram of a power distribution unit is provided. The power distribution unit includes: a double-shielded isolation transformer, a hybrid active filter module, a power distribution monitoring module, and a power distribution backplane. The power distribution backplane includes a multi-layer PCB, with the double-shielded isolation transformer, hybrid active filter module, and power distribution monitoring module respectively disposed on different PCB layers of the power distribution backplane. Optionally, the power distribution unit also includes protection circuitry. The components of the power distribution unit will be described below.

[0092] The double-shielded isolation transformer includes a primary winding and a secondary winding, with two independent electrostatic shielding layers between them. The inner shielding layer is connected to the input protective ground to absorb and conduct high-frequency common-mode noise from the primary winding side. The outer shielding layer is connected to the secondary ground via a high-voltage capacitor (e.g., 100pF / 2kV), forming a high-impedance barrier against high-frequency interference while preventing ground loops. This configuration improves the attenuation capability of common-mode interference on the grid side.

[0093] The hybrid active filtering module employs a hybrid topology of passive and active filters. First, broadband noise is filtered out by an LC passive filter. Then, the signal enters an adaptive active power frequency notch filter and a π-type filter. The center frequency of this filter can be tracked in real-time based on a phase-locked loop (PLL) circuit to the actual power grid frequency (50Hz / 60Hz) and its main harmonics. The notch depth is dynamically adjustable, up to 70dB, aiming to eliminate the periodic stripe artifacts commonly found in images.

[0094] Double-shielded isolation transformers primarily suppress common-mode interference and low-frequency surges, but their suppression of power grid frequency harmonics (such as 50 / 60Hz and its harmonics) is limited. Hybrid active filter modules, on the other hand, are specifically designed for deep notch filtering of these specific periodic interferences. The integrated design ensures that the purified power from the transformer directly enters the filter module, avoiding secondary contamination in intermediate stages. This allows the active filter to operate on a relatively clean baseline, resulting in more precise and stable locking and elimination of power frequency harmonics. The two components complement each other, creating a cascaded purification effect.

[0095] The power distribution monitoring module provides multiple (e.g., 6) independent outputs, which are respectively supplied to gradient power amplifiers, RF power amplifiers, control computers, analog front-ends, etc.

[0096] In this embodiment, the grid input passes through a protection circuit to a double-shielded isolation transformer. The secondary output of the double-shielded isolation transformer is then processed sequentially by an LC passive filter, an adaptive active power frequency notch filter, and a π-type filter. The purified power enters the power distribution monitoring module and is divided into multiple outputs. Each output includes a solid-state relay (SSR), a power monitoring circuit, and local protection.

[0097] The power distribution unit in this embodiment integrates traditionally distributed isolation, filtering, and power distribution functions into a compact chassis that can be directly installed in the system's main control cabinet. Internally, it uses a multi-layer printed circuit board (PCB) to achieve the shortest possible copper busbar connection between modules, eliminating parasitic parameters and radiating loops introduced by long-distance cables in traditional separate installations.

[0098] In one implementation, the system further includes a main controller, a communication interface module, and a control command interface. The main controller communicates with the power distribution monitoring module, managing and monitoring the power distribution unit via a communication bus and receiving control commands from the system host.

[0099] When the main controller performs sequential power-on / power-off, it controls the power supply of each component in the order of control circuit → analog front end → RF power amplifier (standby) → gradient power amplifier to avoid current surges caused by simultaneous startup.

[0100] When the main controller performs dynamic load management, during the scanning process, it controls the RF power amplifier to quickly switch between high-power transmission and low-noise reception states according to the sequence instructions, maximizing energy efficiency and minimizing its own thermal noise.

[0101] In traditional solutions, the connection from the transformer output to the filter module and then to the power distribution board requires long cables inside the cabinet. These cables themselves act as antennas, receiving and radiating noise. This embodiment uses a multi-layer PCB backplane to tightly interconnect all components in the power distribution unit, limiting the flow path of high-frequency interference current to a centimeter scale, reducing path impedance and radiation efficiency, and enabling the subsequent filter module to function more efficiently.

[0102] Compared to discrete power supply systems, the compact integrated chassis in this embodiment forms a good shield, encapsulating all internal high-interference components (such as transformers and filter inductors) and reducing their spatial radiation to other sensitive circuits within the MRI system (such as preamplifiers). Simultaneously, the power distribution monitoring module can uniformly schedule power behavior (such as timed power-on and dynamic load switching) based on the overall system status (obtained from the sequence controller). This global information exchange and collaborative management from the energy end to the load end optimizes the electromagnetic environment at the system level.

[0103] This embodiment utilizes an active noise suppression system to directly cancel out most gradient noise and power supply interference at the source of signal acquisition. The adaptive power frequency notch filtering technology in the power distribution unit dynamically tracks and eliminates interference caused by power grid frequency drift, which helps eliminate regular zipper-like or ring-shaped stripe artifacts in images, improving diagnostic reliability. Simultaneously, the active anti-interference capability reduces the stringent requirements for electromagnetic shielding in the installation environment, enabling the system to operate stably in ordinary clinics and expanding its application scenarios.

[0104] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0105] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for suppressing noise in ultra-low field magnetic resonance imaging, characterized in that, include: Acquire the raw magnetic resonance signal from the main imaging signal channel, and the reference noise signals from multiple reference channels corresponding to noise sensors; The scanning sequence is divided into time periods, and synchronization control signals corresponding to different time periods are generated. In response to the first synchronization control signal, during the non-signal acquisition window of the scan sequence, a training mode is entered, the training mode including: updating the weight coefficients of the adaptive filter bank based on the reference noise signal and the original magnetic resonance signal; In response to the second synchronization control signal, the scanning mode is entered in the signal acquisition window of the scanning sequence. The scanning mode includes: freezing or reducing the update speed of the weight coefficients of the adaptive filter bank, and using the updated weight coefficients to cancel noise in the original magnetic resonance signal, and outputting a purified magnetic resonance signal.

2. The method for suppressing noise in ultra-low field magnetic resonance imaging according to claim 1, characterized in that, The scan sequence is divided into time periods, and synchronization control signals corresponding to different time periods are generated, including: The protocol timing of the scan sequence is analyzed to determine the timing of radio frequency transmission, gradient switching, and signal acquisition in each pulse sequence; Based on the timing sequence of radio frequency transmission, gradient switching, and signal acquisition in each pulse sequence, the signal acquisition blank period, the pre-scanning stage, and the time boundaries between the signal acquisition windows are determined. The first synchronization control signal and the second synchronization control signal are generated based on the time boundary between the signal acquisition blank period, the pre-scanning stage, and the signal acquisition window.

3. The method for suppressing noise in ultra-low field magnetic resonance imaging according to claim 1, characterized in that, The adaptive filter bank includes multiple adaptive filters, each of which is constructed based on the normalized least mean square algorithm, and each of the adaptive filters corresponds to one of the reference channels. Based on the reference noise signal and the original magnetic resonance signal, the weight coefficients of the adaptive filter bank are updated, including: The weight coefficients of the adaptive filter bank are iteratively updated with the goal of minimizing the power of the output error signal.

4. The method for suppressing noise in ultra-low field magnetic resonance imaging according to claim 3, characterized in that, The iterative update formula for the weight coefficients of the adaptive filter bank is as follows: ； Where n is the index of the reference channel, w i (n) represents the coefficient vector of the adaptive filter corresponding to the i-th reference channel at time n, r i e(n) is the i-th reference noise signal vector, e(n) is the error signal at time n, μ is the step size factor, and δ is the regularization factor.

5. The method for suppressing noise in ultra-low field magnetic resonance imaging according to claim 4, characterized in that, Freezing or reducing the update rate of the weight coefficients of the adaptive filter bank includes: When entering the scanning mode, the value of the step size factor is controlled so that the value of the step size factor in the scanning mode is less than the value of the step size factor in the training mode.

6. The method for suppressing noise in ultra-low field magnetic resonance imaging according to claim 1, characterized in that, The method further includes: Before executing the current scan sequence, a pre-stored pattern database is queried; wherein, the pattern database stores an initial set of weight coefficients associated with different scan sequence types and scan parameters; The initial weight coefficient set obtained from the query is loaded as the initial weight of the adaptive filter.

7. An ultra-low field magnetic resonance imaging system, characterized in that, include: The main magnet is used to generate a low-intensity static magnetic field in the imaging area; Gradient coils are used to superimpose spatially encoded gradient magnetic fields onto the static magnetic field; The radio frequency module is used to transmit radio frequency pulses to excite nuclear magnetic resonance and receive the induced magnetic resonance signals; A power distribution unit is used to supply power to the main magnet, the gradient coil, and the radio frequency module; The main imaging signal channel is used to output the raw magnetic resonance signal; A noise sensor array, comprising multiple noise sensors, each used to acquire reference noise signals from different noise sources; The control module is used to acquire the original magnetic resonance signal and the reference noise signal, and execute the ultra-low field magnetic resonance imaging noise suppression method according to any one of claims 1 to 6, and output the purified magnetic resonance signal.

8. The ultra-low field magnetic resonance imaging system according to claim 7, characterized in that, The control module includes: an adaptive filter bank and a sequence controller; wherein... The adaptive filter bank includes multiple adaptive filters, each used to estimate the noise propagation path characteristics from each interference source to the main imaging signal channel based on the input reference noise signal. The sequence controller is used to divide the time period according to the scanning sequence, generate a first synchronization control signal corresponding to the non-signal acquisition window of the scanning sequence and a second synchronization control signal corresponding to the signal acquisition window of the scanning sequence, and send the synchronization control signal to the adaptive filter bank to control the adaptive filter bank to switch the working mode.

9. The ultra-low field magnetic resonance imaging system according to claim 7, characterized in that, The power distribution unit includes: The system includes a double-shielded isolation transformer, a hybrid active filter module, a power distribution monitoring module, and a power distribution backplane. The power distribution backplane comprises a multi-layer PCB, and the double-shielded isolation transformer, the hybrid active filter module, and the power distribution monitoring module are respectively disposed on different PCB layers of the power distribution backplane.

10. The ultra-low field magnetic resonance imaging system according to claim 9, characterized in that, The noise sensor group includes: The first sensor is located outside the gradient coil and is used to collect the time-varying magnetic field noise generated when the gradient current switches. The second sensor is located inside the cavity of the main magnet and is used to collect low-frequency magnetic field noise in the environment. The third sensor is located near the entrance of the power distribution unit and is used to collect grid coupling and spatial radiation interference noise.

Images