Method and device for magnetic particle imaging based on multi-source noise monitoring and collaborative compensation

Abstract

The application belongs to the field of magnetic particle imaging, and particularly relates to a magnetic particle imaging method and device based on multi-source noise monitoring and collaborative compensation, aiming to solve the problem that magnetic particle signals are easily interfered by multi-source noise. The method comprises the following steps: during excitation or gradient field operation, a main receiving signal is collected, and multi-path noise monitoring signals are synchronously collected through at least two noise monitoring coils; the multi-path noise monitoring signals are input into a background noise prediction model to obtain predicted background noise, wherein the background noise prediction model is trained based on signals collected by the main receiving coil and the noise monitoring coil under the no-load state of a magnetic particle imaging device; the predicted background noise is eliminated from the main receiving signal to obtain a purified magnetic particle response signal; and a reconstructed image is generated based on the purified magnetic particle response signal. The application can obtain a high signal-to-noise ratio MPI response signal online, and improve the imaging speed and long-term stability without changing the existing scanning architecture.

Description

Technical Field

[0001] This application belongs to the field of magnetic particle imaging, specifically relating to a magnetic particle imaging method and apparatus based on multi-source noise monitoring and collaborative compensation. Background Technology

[0002] Magnetic particle imaging (MPI) is an emerging functional imaging technique using tracers. It utilizes the nonlinear magnetization response of superparamagnetic iron oxide nanoparticles (SPIONs) in an applied alternating magnetic field to achieve highly sensitive, high-resolution imaging of their spatial distribution. It shows great potential in biomedical fields such as angiography, cell tracking, and tumor-targeted imaging, and is particularly suitable for monitoring dynamic processes.

[0003] However, the particle signals acquired by the MPI receiver link in actual operation are susceptible to background noise interference and drift. The background noise sources are complex, typically a mixture of multiple noise sources, primarily including excitation magnetic field leakage and mutual inductance coupling, power amplifier ripple and its harmonic leakage, transient / inrush currents introduced by gradient switching, electromagnetic interference from power frequency and radio environments, and amplitude and phase drift caused by temperature drift in the coils and front-end electronics. This background noise includes both narrowband components (such as rectifier ripple, power frequency, and harmonics) and broadband / transient components (such as gradient switching spikes), and varies with the operating point, scan trajectory, coil attitude, and temperature. It is particularly prominent in harmonic detection and multi-frequency multiplexing scenarios, often causing the particle response to be submerged and inducing baseline drift. These time-varying, multi-source noises severely reduce the signal-to-noise ratio of the original signal, thus affecting the final imaging quality.

[0004] To suppress the aforementioned noise, existing technologies typically employ the following methods. The first is passive hardware methods, such as adding complex electromagnetic shielding layers, designing differentially balanced receiving coils, or using notch / bandpass filters. However, these methods are highly dependent on specific noise frequencies and scenarios, struggle to handle broadband, time-varying noise and temperature drift, and increase system size and cost. Over-filtering may also weaken useful particle signals. The second is procedural background subtraction, which involves inserting a particle-free empty scan into the imaging sequence to acquire pure background signals, which are then subtracted from the particle-containing data. However, this method interrupts the imaging process, significantly reducing imaging speed, making it unsuitable for scenarios requiring continuous and rapid monitoring. Furthermore, when temperature drift exists, the background acquired before and after the scan is inconsistent, resulting in poor subtraction performance. The third is pure image post-processing methods, which suppress artifacts through algorithms during image reconstruction. However, these methods only address the symptoms, failing to improve the signal-to-noise ratio of the original signal at its source. Moreover, the algorithms are often strongly correlated with specific tasks or data distributions, lacking versatility and robustness.

[0005] In summary, existing technologies still have significant shortcomings in terms of linearity, versatility, and robustness to time-varying and temperature drift, especially in scenarios involving long-term continuous monitoring, rapid scanning, and multi-frequency harmonic detection, where it is difficult to balance imaging speed, signal-to-noise ratio, and system stability. Summary of the Invention

[0006] To address the aforementioned problems in the prior art, namely the susceptibility of magnetic particle signals to multi-source noise interference, this application provides a magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation, comprising:

[0007] During excitation or gradient field operation, a main receiving signal containing magnetic particle response signal and background noise is acquired through the main receiving coil, and multi-directional background noise is simultaneously acquired through at least two noise monitoring coils to obtain multi-channel noise monitoring signals.

[0008] The background noise prediction model is obtained by inputting multiple noise monitoring signals into it. The background noise prediction model is trained based on the signals collected by the main receiving coil and the noise monitoring coil of the magnetic particle imaging device under no-load conditions.

[0009] The predicted background noise is removed from the main received signal to obtain the purified magnetic particle response signal;

[0010] A reconstructed image is generated based on the purified magnetic particle response signal.

[0011] As a preferred implementation, the training process of the background noise prediction model includes:

[0012] In the no-load state of the magnetic particle imaging equipment, the no-load signal of the main receiving coil and the noise signal of the noise monitoring coil are simultaneously acquired and a data pair is formed.

[0013] The training set consists of multiple data pairs, with the empty signal as the target ground truth. The noise signal is trained by optimizing the loss function until the loss function converges, resulting in a trained background noise prediction model. The feature extraction part of the background noise prediction model consists of a parallel convolutional neural network feature extraction branch and an attention mechanism model feature extraction branch.

[0014] As a preferred embodiment, the method further includes:

[0015] The loss function is a piecewise function. Different loss calculation methods are used based on the relationship between the absolute value of the difference between the predicted value of the background noise and the no-load signal and the preset threshold.

[0016] When the absolute value of the difference is less than the preset threshold, the loss function is calculated using the quadratic loss form;

[0017] When the absolute value of the difference is greater than or equal to the preset threshold, the loss function is calculated using the linear loss form.

[0018] In a preferred embodiment, the noise monitoring coils are respectively arranged at compensation positions, which include at least one of the following: an external environment position at a preset distance away from the magnetic particle imaging device, the housing of the magnetic particle imaging device, the displacement stage, and the delivery port.

[0019] As a preferred embodiment, the method further includes:

[0020] During the operation of the magnetic particle imaging equipment, acquire temperature information synchronized with multiple noise monitoring signals;

[0021] Temperature information and multiple noise monitoring signals are input into the background noise prediction model to obtain the predicted background noise.

[0022] As a preferred embodiment, the method further includes:

[0023] Temperature drift compensation is performed on multiple noise monitoring signals based on temperature information.

[0024] As a preferred embodiment, after obtaining the multiple noise monitoring signals, the method further includes:

[0025] Time delay alignment and amplitude-phase registration are performed on multiple noise monitoring signals.

[0026] On the other hand, this application proposes a magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation, used to perform the above-mentioned magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation, including:

[0027] The main receiving coil is used to acquire the main receiving signal, which includes magnetic particle response signals and background noise.

[0028] At least two noise monitoring coils are used to acquire multiple noise monitoring signals;

[0029] The synchronous acquisition module is used to synchronously control the main receiving coil and at least two noise monitoring coils to acquire signals during excitation or gradient field operation.

[0030] The background noise prediction module is used to input multiple noise monitoring signals into the background noise prediction model to obtain the predicted background noise. The background noise prediction model is trained based on the signals collected by the main receiving coil and the noise monitoring coil of the magnetic particle imaging device under no-load conditions.

[0031] The collaborative compensation module is used to eliminate the predicted background noise from the main received signal and obtain the purified magnetic particle response signal.

[0032] The image reconstruction module is used to generate a reconstructed image based on the purified magnetic particle response signal.

[0033] In a preferred embodiment, the apparatus further includes:

[0034] At least one temperature sensing module is used to acquire temperature information of the main receiving coil during operation of the magnetic particle imaging device.

[0035] In a preferred embodiment, the synchronous acquisition module is also used to synchronously control the main receiving coil and at least one temperature sensing module to acquire signals during excitation or gradient field operation.

[0036] The beneficial effects of this application are:

[0037] (1) By acquiring external and internal electromagnetic interference signals in the magnetic particle imaging system and simultaneously inputting them into the deep learning model along with the signal received by the interfered magnetic particle imaging device, the interference signals can be accurately modeled and eliminated, thereby obtaining high-quality, interference-free magnetic particle signals and images, solving the problem of existing technologies relying on hardware shielding.

[0038] (2) By capturing background noise using a noise monitoring coil while acquiring the main signal, parallel processing of noise monitoring and signal acquisition is achieved. The entire prediction and cancellation process is carried out online without interrupting the scan or making mechanical switching, which fundamentally eliminates the time overhead caused by empty background acquisition and greatly improves imaging efficiency.

[0039] (3) This application uses at least two noise monitoring coils for monitoring, which can more comprehensively capture noise characteristics from different sources and different paths, forming a rich noise reference information library. By processing these multi-channel noise monitoring signals to generate accurate prediction values, the complex and time-varying noise in the main signal can be dynamically and adaptively tracked and canceled, thereby obtaining a purer magnetic particle response signal and greatly improving the quality of the magnetic particle response signal. Attached Figure Description

[0040] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0041] Figure 1 This is a flowchart of a magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation provided in one embodiment of this application;

[0042] Figure 2 This is an example diagram of a compensation position provided in one embodiment of this application;

[0043] Figure 3 This is a structural diagram of a magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation provided in one embodiment of this application;

[0044] Figure 4 This is an internal integration framework diagram of a collaborative compensation module provided in one embodiment of this application;

[0045] Figure 5 This is a schematic diagram of the structure of a computer system used to implement the methods, apparatus, and electronic devices of this application. Detailed Implementation

[0046] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0047] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] To address the issues of image quality degradation caused by multi-source, time-varying background and temperature drift in the MPI receiving link, low imaging speed due to the empty background process, and the insufficient versatility and robustness of post-processing artifact removal methods alone, this application provides a magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation. During excitation or gradient field operation, a main receiving coil acquires a main receiving signal containing magnetic particle response signals and background noise, while at least two noise monitoring coils simultaneously acquire background noise, resulting in multiple noise monitoring signals. These multiple noise monitoring signals are input into a background noise prediction model to obtain predicted background noise. The background noise prediction model is trained based on signals acquired by the main receiving coil and noise monitoring coils under no-load conditions of the magnetic particle imaging device. The predicted background noise is then removed from the main receiving signal to obtain a purified magnetic particle response signal. Based on the magnetic particle response signal, a reconstructed image is generated. This method enables online acquisition of high signal-to-noise ratio MPI response signals, improving imaging speed and long-term stability without altering the existing scanning architecture.

[0049] To more clearly explain the magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation in this application, the following will be combined with... Figure 1 The steps in the embodiments of this application are described in detail.

[0050] The magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation in the first embodiment of this application includes steps S10-S40, each of which is described in detail below:

[0051] Step S10: During excitation or gradient field operation, a main receiving signal containing magnetic particle response signal and background noise is acquired through the main receiving coil, and background noise from multiple directions is acquired synchronously through at least two noise monitoring coils to obtain multi-channel noise monitoring signals.

[0052] It's important to note that the excitation magnetic field is a high-frequency, high-intensity alternating magnetic field in the MPI system. Its function is to drive the rapid flipping of the magnetic moments of magnetic nanoparticles (MNPs). Only when the excitation magnetic field is active will the MNPs generate a detectable response signal. In other words, the excitation magnetic field is the source of the MPI signal. The gradient field creates a point or line within the imaging region where the magnetic field is zero (i.e., a "field-free point," FFP). Only MNPs located near the FFP can respond to the excitation magnetic field and flip; MNPs in other locations are "saturated" by the gradient field and therefore do not generate a signal. By moving the FFP, the MPI system can scan the entire region, achieving spatial encoding. The gradient field itself also changes when it moves or switches.

[0053] Optionally, during operation in the excitation magnetic field or gradient field, signals and noise are generated simultaneously. At this time, on the one hand, the main receiving signal is acquired through the main receiving coil, which includes the magnetic particle response signal and the background noise generated by the MPI system itself when performing the imaging task; on the other hand, the background noise from multiple directions is acquired through the noise monitoring coil to generate multiple noise monitoring signals.

[0054] In this embodiment, at least two noise monitoring coils are used to collect background noise in a distributed, multi-directional manner. The noise monitoring coils are respectively arranged at compensation positions, which include at least one of the following: an external environmental location at a preset distance from the magnetic particle imaging device, the outer casing of the magnetic particle imaging device, the displacement stage, and the delivery port.

[0055] As an example, please see Figure 2 The noise monitoring coil is placed at 10 compensation locations, namely: external environment location 201, magnetic particle imaging equipment housing 202, displacement stage 203, delivery port 204, near the drive coil 205, near the gradient coil 206, power amplifier / power node 207, key point of transmission line 208, outside of shield 209, and reference ground / rack 210.

[0056] In this embodiment, the preset distance is 1 meter, that is, the external environment position 201 is 1 meter away from the magnetic particle imaging device; the displacement stage 203 is used to deliver the object to be examined, such as a phantom or mouse model; the area near the drive coil 205 and the area near the gradient coil 206 are both within a straight line distance of 1 meter centered on the main receiving coil.

[0057] It should be noted that, in order to clearly illustrate the placement of the noise monitoring coil, Figure 2 As an abstract device, only the compensation position of the coil arrangement is provided, omitting the receiving coil selection coil, etc.

[0058] Understandably, noise monitoring coils are strategically and distributedly positioned at key locations both inside and outside the MPI system to actively sense and capture background noise from different sources. Different compensation locations can characterize different noise sources. For example, a coil positioned at the external environment location 201, 1 meter away from the equipment, is used to capture far-field environmental electromagnetic noise such as power frequency interference and radio broadcasts; coils positioned on the equipment housing 202, and near the drive coil 205 and gradient coil 206 respectively, are used to monitor strong coupling and leakage magnetic fields generated by the switching of the drive field and gradient field. Coils positioned on the displacement stage 203 carrying the object under test, and at the delivery port 204, are used to monitor electromagnetic interference related to mechanical motion; coils positioned at the power amplifier or power supply node 207 are used to monitor power amplifier ripple and its harmonics; coils positioned at key nodes of the signal transmission line 208, outside the shield 209, and at the reference ground or rack 210 are used to comprehensively capture conducted and radiated noise.

[0059] In other embodiments, the compensation positions for arranging the noise monitoring coils can be determined according to the actual environment, and the number of compensation positions N≥2.

[0060] In this embodiment, the synchronous acquisition of the main received signal and multiple noise monitoring signals is achieved through a unified clock and phase locking.

[0061] In some embodiments, after acquiring multiple noise monitoring signals, delay estimation and alignment are also performed on the monitoring channels.

[0062] As an example, by reading the main received signal and multiple noise monitoring signals, the timestamps and amplitude ranges of different monitoring channels are unified to achieve alignment, and alignment results including time interval, number of sampling points and amplitude deviation index are generated. The difference in timestamps before and after alignment is used as the time delay estimate.

[0063] In some embodiments, during the operation of the magnetic particle imaging device, temperature information synchronized with multiple noise monitoring signals is acquired, and temperature drift compensation is performed on the multiple noise monitoring signals based on the temperature information.

[0064] As an example, by reading the temperature in real time, it can be determined whether the ambient temperature during operation is within a stable range. If the real-time temperature exceeds the stable range, it indicates that the temperature has drifted slowly or fluctuated briefly, which triggers water cooling to avoid decreased sensitivity, baseline shift, or changes in harmonic components caused by heat accumulation.

[0065] The stable range can be, for example, room temperature ± 5°.

[0066] It should be noted that by estimating and aligning time delays, the minute time delays introduced by differences in coil positions, cable lengths, and electronic paths can be corrected, ensuring that all noise signals are precisely aligned in the time domain. By registering amplitude and phase, the amplitude and phase differences between channels can be corrected. By using real-time temperature information for temperature drift compensation, the amplitude and phase drift of the signal can be compensated, thereby improving the accuracy of background prediction.

[0067] In some embodiments, the time delay and amplitude phase can be updated online periodically according to the data acquisition cycle to adapt to time variations and temperature drift. The update cycle varies depending on the acquisition device or acquisition method, and the specific update cycle is determined based on the actual acquisition process.

[0068] Step S20: Input the multi-channel noise monitoring signals into the background noise prediction model to obtain the predicted background noise. The background noise prediction model is trained based on the signals collected by the main receiving coil and the noise monitoring coil of the magnetic particle imaging device under no-load conditions.

[0069] Optionally, a background noise prediction model can be obtained first through model training. Then, the collected multi-channel noise monitoring signals can be input into the background noise prediction model. The model can then predict the background noise and output the predicted background noise as the background noise that needs to be eliminated.

[0070] In this embodiment, when the magnetic particle imaging device is in an unloaded state, the unloaded signal of the main receiving coil and the noise signal of the noise monitoring coil are synchronously collected and formed into data pairs. Multiple data pairs constitute a training set. The unloaded signal is used as the target ground truth. The noise signal is trained by optimizing the loss function until the loss function converges, and the trained background noise prediction model is obtained.

[0071] It should be noted that in the no-load state, the no-load signal acquired by the main receiving coil consists only of the background noise generated by the MPI system itself during imaging, which is entirely composed of internal system noise, coil coupling interference, and environmental electromagnetic interference. The noise signal acquired by the noise monitoring coil includes environmental noise and hardware noise at different locations. The noise monitoring coil samples synchronously with the main receiving coil to ensure that all noise components are perfectly aligned in time. Any pair of aligned no-load and noise signals constitutes a data pair. By repeatedly acquiring data under different environmental conditions, a training dataset covering various noise types and amplitudes can be obtained.

[0072] In this embodiment of the application, the training dataset is divided into a training set and a validation set in an 8:2 ratio.

[0073] Furthermore, in the actual imaging phase involving particle movement, the magnetic particle imaging device operates according to a predetermined scanning sequence under different temperatures and electromagnetic environments, recording the original magnetic particle received signals (including noise) and the noise signals captured by the compensation coil. A multi-round data acquisition approach is adopted, recording five no-load signals and five particle-containing signals each time, forming a test set.

[0074] As one possible implementation, after the dataset is collected, all raw signals are preprocessed: first, normalization is performed to eliminate the impact of amplitude differences on network training; then, downsampling is performed according to the sampling rate and network input requirements to reduce the computational burden; finally, bandpass filtering can be applied to ensure that the signal features are clear and meet the network learning requirements.

[0075] In this embodiment, the core feature extraction part of the model consists of a parallel dual-branch structure. Specifically, the feature extraction part of the background noise prediction model comprises a parallel convolutional neural network feature extraction branch and an attention mechanism model feature extraction branch. The convolutional neural network feature extraction branch contains three cascaded one-dimensional convolutional modules specifically designed to extract local feature patterns from multi-channel noise signals. The first convolutional module sets an appropriate kernel size and number of filters for the one-dimensional convolutional layer, combined with batch normalization and activation functions; for example, the kernel size can be 2×2. The second convolutional module updates the kernel size and filter configuration, further abstracting the feature representation. The third convolutional module connects the second convolutional layer using the kernel size and number of filters, and then performs a transpose operation to fuse with the Transformer branch. The parallel feature projection and position encoding branch first maps the signal dimension of the multi-channel induction coil to the corresponding dimension through a linear projection layer, and then adds sinusoidal position encoding to provide sequence order information for the model. The outputs of the two branches are integrated in the feature fusion layer through element-wise addition, producing a unified representation that combines local features and position information while preserving dimensionality.

[0076] The model utilizes a multi-input structure and attention mechanism to extract and fuse features from different sources. Further, the fused features are input into a Transformer encoder consisting of N identical encoder layers stacked together. Each encoder layer contains two main sub-layers: a multi-head self-attention mechanism and a feedforward neural network. Each sub-layer is followed by residual connections and layer normalization operations. The multi-head self-attention mechanism is configured with n attention heads, each with dimension 'a', comprehensively modeling global dependencies between all time points in the sequence. The feedforward neural network employs a two-linear-layer structure, providing non-linear transformation capabilities. Each encoder layer maintains consistent input and output dimensions. Finally, the output of the Transformer encoder aggregates information along the sequence dimension through a global average pooling layer to produce a feature representation. A linear layer then maps the features back to L dimensions, outputting the noise signal that predicts the master receiving coil.

[0077] It should be noted that the external environment is highly uncertain during data collection, and the signal from the induction coil may contain sudden spikes. The mean squared error loss is sensitive to outliers and may sacrifice overall performance to correct a few large errors. To improve the robustness of the model, the loss function is designed as a piecewise function. Different loss calculation methods are used based on the relationship between the absolute value of the difference between the predicted value of the background noise and the no-load signal and the preset threshold. When the absolute value of the difference is less than the preset threshold, the loss function is calculated using the quadratic loss form; when the absolute value of the difference is greater than or equal to the preset threshold, the loss function is calculated using the linear loss form.

[0078] As an example, the loss function can be represented by the following formula. :

[0079]

[0080] Where N represents the batch size and L represents the sequence length. This represents the true empty signal (i.e., the target true value) obtained from the j-th sequence in the i-th batch. This represents the predicted background noise value for the j-th sequence in the i-th batch, as predicted by the model. This indicates a preset threshold (e.g., a value of 1).

[0081] By training a CNN-Transformer model, it can accurately learn the complex nonlinear mapping from auxiliary channel noise to main channel noise. By optimizing the loss function and continuously adjusting parameters such as the model's learning rate and epochs until the loss function converges, a trained background noise prediction model is obtained. The CNN-Transformer model can completely map the relationship between the compensation receiving coil and the unloaded main receiving coil. The model is encapsulated in .h5 format, and the running parameters are saved in parameters.py.

[0082] Furthermore, the multiple noise monitoring signals are input into the background noise prediction model to accurately estimate the noise components mixed in the current main coil received signal, and output a noise prediction value that strictly matches the main coil signal in the spatial dimension. Because the compensation receiving coil is not sensitive to particle signals, It can be regarded as the sum of the noise from thermal noise, coil noise and external environmental noise on the main coil received signal, and is used as the predicted background noise.

[0083] In some embodiments, during the operation of the magnetic particle imaging device, temperature information synchronized with multiple noise monitoring signals is acquired; the temperature information and multiple noise monitoring signals are input together into the background noise prediction model to obtain the predicted background noise.

[0084] Step S30: Eliminate the predicted background noise from the main received signal to obtain the purified magnetic particle response signal.

[0085] Optionally, the predicted background noise is the sum of noise that needs to be eliminated. Therefore, the main received signal, which is the original signal, is canceled out with the predicted background noise, which is the sum of noise, to obtain the purified magnetic particle response signal.

[0086] As an example, it can be done through the formula In this representation, s(t) represents the purified magnetic particle response signal, and y(t) represents the main received signal. This indicates the predicted background noise.

[0087] The pure magnetic particle signal s(t) is predicted by compensating for background noise in the main received signal y(t). This provides reliable raw data support for high-precision reconstruction in subsequent imaging algorithms.

[0088] This application's method eliminates noise by real-time measurement of multiple noise monitoring signals and obtaining predicted background noise. It dynamically senses interference, suppressing multi-source and time-varying background noise and compensating for temperature drift online without mechanical switching or empty background acquisition. This significantly improves signal-to-noise ratio, imaging speed, and long-term stability. Furthermore, this method is not dependent on a single noise type and can handle various complex situations such as thermal noise, coil coupling interference, and environmental electromagnetic interference. The deep network captures the statistical laws of noise through sensing coils, exhibiting good generalization ability and operating under different MPI system parameters and magnetic field strengths.

[0089] Step S40: Generate a reconstructed image based on the purified magnetic particle response signal.

[0090] Optionally, there are two types of magnetic particle imaging methods. One is direct imaging using time-domain signals, such as the X-space method. After removing background noise using the method of this application, the problem of easily generating stripe artifacts can be avoided. The other is the system matrix method for imaging using frequency-domain signals. After removing background noise using the method of this application, the original sample distribution image is reconstructed based on the pure magnetic particle response signal using the matrix to obtain the reconstructed image.

[0091] The method in this application replaces some hardware isolation with an algorithm combined with a sensing coil, achieving a low-cost, lightweight design and providing possibilities for portable MPI or fast imaging applications.

[0092] The magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation according to the second embodiment of this application is used to perform the above-described magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation, including:

[0093] The main receiving coil is used to acquire the main receiving signal, which includes magnetic particle response signals and background noise.

[0094] At least two noise monitoring coils are used to acquire multiple noise monitoring signals;

[0095] The synchronous acquisition module is used to synchronously control the main receiving coil and at least two noise monitoring coils to acquire signals during excitation or gradient field operation.

[0096] The background noise prediction module is used to input multiple noise monitoring signals into the background noise prediction model to obtain the predicted background noise. The background noise prediction model is trained based on the signals collected by the main receiving coil and the noise monitoring coil of the magnetic particle imaging device under no-load conditions.

[0097] The collaborative compensation module is used to eliminate the predicted background noise from the main received signal and obtain the purified magnetic particle response signal.

[0098] The image reconstruction module is used to generate a reconstructed image based on the purified magnetic particle response signal.

[0099] In some embodiments, the device further includes at least one temperature sensing module for acquiring temperature information of the main receiving coil during operation of the magnetic particle imaging device.

[0100] In the case where the device includes a temperature sensing module, the synchronous acquisition module is also used to synchronously control the main receiving coil and at least one temperature sensing module to acquire signals during excitation or gradient field operation.

[0101] As one possible implementation method, please refer to Figure 3 The magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation includes a main receiving coil 110 and at least two noise monitoring coils. Figure 3 Not shown in the figure, but can be referred to in the method embodiments. Figure 2 The system includes a synchronous acquisition module 120, a unified clock and phase locking module 140, a temperature sensing module 150, a gradient coil and excitation coil current detection module 160, a collaborative compensation module, and an image reconstruction module.

[0102] The device further includes a drive / excitation coil 011, a gradient / field selection coil 012, a power amplifier and power supply 013, a test object bearing and displacement mechanism 014, and a signal processing and system control unit 015. 011-015 constitute a field generation and scanning subsystem, which is used to generate field selection (FFP / FFL), drive field and gradient field, and to perform scanning control and trigger synchronization.

[0103] During the operation of the field generation and scanning subsystem's excitation or gradient field, the synchronous acquisition module 120, under the constraint of a unified clock 140, simultaneously performs multi-channel synchronous sampling of the output signals of the main receiving coil 110 and all noise monitoring coils, ensuring that all acquired signals are strictly aligned in time.

[0104] The collaborative compensation module receives the main received signal and multiple noise monitoring signals from the synchronous acquisition module 120, as well as status data such as temperature from the temperature sensing module 150. For example... Figure 4 As shown, the collaborative compensation module 170 integrates multiple functional units, including a time delay estimation and alignment unit 171, an amplitude and phase registration unit 172, a temperature drift compensation unit 173, a background prediction unit 174, a phase cancellation synthesis unit 175, and an online update unit 176. The collaborative compensation module 170 can be implemented in hardware (such as FPGA, DSP) or software (running in the signal processing and system control unit 015).

[0105] The collaborative compensation module first preprocesses the acquired multi-channel noise monitoring signals. Specifically, the time delay estimation and alignment unit 171 is responsible for correcting the minute time delays introduced by differences in coil positions, cable lengths, and electronic pathways, ensuring that all noise signals are accurately aligned in the time domain. The amplitude and phase registration unit 172 is responsible for correcting the amplitude and phase differences between channels. In addition, the temperature drift compensation unit 173 uses real-time acquired temperature data to compensate for the amplitude and phase drift of the signal, thereby improving the accuracy of background prediction.

[0106] Furthermore, the collaborative compensation module 170 takes the processed multi-channel noise monitoring signals (and optional temperature / operating condition data) as input through the background prediction unit 174, and estimates the predicted background noise mixed in the main receiving coil 110 in real time and dynamically under the current system state through a multi-input single-output background noise prediction model. The destructive synthesis unit 175 converts the predicted background noise value output by the background prediction unit 174 into the predicted background noise value. Subtract or cancel from the original main received signal y(t), i.e. The difference s(t) is the purified magnetic particle response signal with a significantly improved signal-to-noise ratio.

[0107] Furthermore, to adapt to slow changes in system state (such as longer-term temperature drift) or changes in environmental noise, the online update unit 176 can periodically (e.g., by period T) update the registration parameters of the preprocessing stage and the weights or coefficients of the background noise prediction model based on newly acquired data. This gives the device strong adaptability and long-term stability.

[0108] Finally, the purified magnetic particle signal s(t) is output to the image reconstruction module for subsequent image reconstruction, thereby obtaining an MPI image with fewer artifacts and higher quality.

[0109] The aforementioned device enables active, real-time monitoring and compensation of background noise in MPI signals without interrupting the imaging process for background acquisition. It can suppress multi-source, time-varying background noise online and in real time, and compensate for temperature drift. Thus, without sacrificing imaging speed, it significantly improves the signal-to-noise ratio and system stability, making it particularly suitable for applications requiring long-term continuous monitoring.

[0110] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the device described above can be found in the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0111] It should be noted that the magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the modules or steps in the embodiments of this application can be further decomposed or combined. For example, the modules in the above embodiments can be merged into one module, or further divided into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of this application are only for distinguishing the various modules or steps and are not considered as an improper limitation of this application.

[0112] A device according to a third embodiment of this application includes:

[0113] At least one processor;

[0114] and a memory communicatively connected to at least one of the processors;

[0115] The memory stores instructions that can be executed by the processor to implement the above-described magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation.

[0116] A computer-readable storage medium according to a fourth embodiment of this application stores computer instructions that are executed by the computer to implement the above-described magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation.

[0117] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes and related descriptions of the electronic devices, computer-readable storage media, and computer program products described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0118] The following is for reference. Figure 5 It shows a schematic diagram of the structure of a computer system for implementing the embodiments of the apparatus, method, and electronic equipment of this application. Figure 5 The server shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0119] like Figure 5 As shown, the computer system includes a Central Processing Unit (CPU) 501, which can perform various appropriate actions and processes based on programs stored in Read Only Memory (ROM) 502 or programs loaded from storage section 508 into Random Access Memory (RAM) 503. The RAM 503 also stores various programs and data required for system operation. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An Input / Output (I / O) interface 505 is also connected to the bus 504.

[0120] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. Removable media 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0121] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs the functions defined in the methods of this application. It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on a computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0122] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0123] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0124] The terms “first”, “second”, etc., are used to distinguish similar objects, not to describe or indicate a specific order or sequence.

[0125] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / device.

[0126] The technical solutions of this application have been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of this application is obviously not limited to these specific embodiments. Without departing from the principles of this application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of this application.

Claims

1. A magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation, characterized in that, include: During excitation or gradient field operation, a main receiving signal containing magnetic particle response signal and background noise is acquired through the main receiving coil, and multi-directional background noise is simultaneously acquired through at least two noise monitoring coils to obtain multi-channel noise monitoring signals. The multi-channel noise monitoring signals are input into the background noise prediction model to obtain the predicted background noise. The background noise prediction model is trained based on the signals collected by the main receiving coil and the noise monitoring coil of the magnetic particle imaging device under no-load conditions. The predicted background noise is removed from the main received signal to obtain the purified magnetic particle response signal; A reconstructed image is generated based on the purified magnetic particle response signal. The training process of the background noise prediction model includes: In the no-load state of the magnetic particle imaging device, the no-load signal of the main receiving coil and the noise signal of the noise monitoring coil are synchronously acquired and a data pair is formed. The training set consists of multiple data pairs, with the empty signal as the target ground truth. The noise signal is trained by optimizing the loss function until the loss function converges, resulting in a trained background noise prediction model. The feature extraction part of the background noise prediction model consists of a parallel convolutional neural network feature extraction branch and an attention mechanism model feature extraction branch.

2. The magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation according to claim 1, characterized in that, The method further includes: The loss function is a piecewise function, and different loss calculation methods are adopted according to the relationship between the absolute value of the difference between the predicted value of the background noise and the empty signal and the preset threshold. When the absolute value of the difference is less than the preset threshold, the loss function is calculated using a quadratic loss form; When the absolute value of the difference is greater than or equal to the preset threshold, the loss function is calculated using a linear loss form.

3. The magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation according to claim 1, characterized in that, The noise monitoring coils are respectively arranged at compensation positions, and the compensation positions include at least one of the following: an external environment position at a preset distance away from the magnetic particle imaging device, the outer shell of the magnetic particle imaging device, the displacement stage, and the delivery port.

4. The magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation according to claim 1, characterized in that, The method further includes: During the operation of the magnetic particle imaging device, temperature information synchronized with the multi-channel noise monitoring signal is acquired; The temperature information and the multi-channel noise monitoring signal are input together into the background noise prediction model to obtain the predicted background noise.

5. The magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation according to claim 4, characterized in that, The method further includes: Temperature drift compensation is performed on the multi-channel noise monitoring signals based on the temperature information.

6. The magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation according to claim 4, characterized in that, After obtaining the multiple noise monitoring signals, the method further includes: The multi-channel noise monitoring signals are subjected to time delay alignment and amplitude-phase registration processing.

7. A magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation, used to execute the magnetic particle imaging method based on multi-source noise monitoring and collaborative compensation as described in any one of claims 1-6, characterized in that, include: The main receiving coil is used to acquire the main receiving signal, which includes magnetic particle response signals and background noise. At least two noise monitoring coils are used to acquire multiple noise monitoring signals; The synchronous acquisition module is used to synchronously control the main receiving coil and the at least two noise monitoring coils to acquire signals during excitation or gradient field operation. The background noise prediction module is used to input the multi-channel noise monitoring signals into the background noise prediction model to obtain the predicted background noise. The background noise prediction model is trained based on the signals collected by the main receiving coil and the noise monitoring coil of the magnetic particle imaging device under no-load conditions. The collaborative compensation module is used to eliminate the predicted background noise from the main received signal and obtain the purified magnetic particle response signal. The image reconstruction module is used to generate a reconstructed image based on the purified magnetic particle response signal.

8. The magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation according to claim 7, characterized in that, The device further includes: At least one temperature sensing module is used to acquire temperature information of the main receiving coil during operation of the magnetic particle imaging device.

9. The magnetic particle imaging device based on multi-source noise monitoring and collaborative compensation according to claim 8, characterized in that, The synchronous acquisition module is also used to synchronously control the main receiving coil and the at least one temperature sensing module to acquire signals during excitation or gradient field operation.

Images