Method and system for calculating the propagation and scattering processes of low-frequency electromagnetic waves in the brain

Abstract

The application discloses a calculation method and system for the transmission and scattering process of low-frequency electromagnetic waves in the brain, and relates to the technical field of brain detection. The method simplifies the brain tissue into a homogeneous isotropic spherical model, which significantly reduces the calculation complexity and lays a foundation for efficient solution. Then, an electromagnetic wave propagation and scattering physical model based on Maxwell's equations is constructed to accurately describe the motion law of electromagnetic waves in brain tissue. The physical model is deeply integrated into a deep learning algorithm to efficiently solve the three-dimensional electromagnetic field distribution in the brain. Finally, the propagation path, energy distribution and scattering field characteristics of electromagnetic waves are analyzed to realize the positioning and qualitative diagnosis of brain edema, tumors and other lesions, and provide key support for brain tissue health assessment. The application overcomes the efficiency bottleneck of traditional numerical methods in brain electromagnetic calculation, improves the timeliness and accuracy of brain disease diagnosis, and provides reliable technical support for bedside rapid diagnosis of acute conditions such as stroke.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of brain detection, and particularly relates to a calculation method and system for conduction and scattering process of low-frequency electromagnetic waves in the brain. BACKGROUND

[0002] With the increasing demand for early screening of brain diseases, non-invasive detection technology based on electromagnetic waves has become a key development direction in the field of neurology. Low-frequency electromagnetic waves have unique advantages in the diagnosis of brain edema, brain tumors and other diseases due to their deep penetration ability and biological tissue safety.

[0003] The prior art mainly relies on traditional numerical calculation methods to simulate the behavior of electromagnetic waves in brain tissue, such as the finite element method. This method faces serious challenges in clinical applications: due to the complexity of the brain boundary conditions, the electromagnetic calculation of the whole brain requires processing a large number of grid elements, resulting in a long simulation time of several hours or even several days for a single simulation, which cannot meet the real-time diagnosis requirements of clinical applications; to improve the calculation speed, the brain model is simplified too much, which causes large accuracy errors and significantly reduces the reliability of disease detection; the existing methods mainly focus on the calculation of electromagnetic wave propagation paths, and the quantitative analysis of scattering field characteristics is seriously insufficient, making it difficult to identify early small lesions; the more prominent problem is that these methods often rely on high-performance computing clusters, which seriously hinders the popularization of the technology in primary medical care.

[0004] Although artificial intelligence methods have shown potential in accelerating calculations, directly applying general neural network models will violate the basic physical laws of electromagnetic fields and lack the explainability required for medical diagnosis.

[0005] The current technology has not effectively solved the threefold contradiction between accuracy, efficiency and explainability in brain electromagnetic calculation, especially the research on the correlation mechanism of scattering field characteristics and lesion properties is blank. SUMMARY

[0006] The main purpose of the present application is to provide a calculation method and system for the conduction and scattering process of low-frequency electromagnetic waves in the brain, to solve the problem of low accuracy of traditional low-frequency electromagnetic waves for brain detection in the prior art.

[0007] In order to achieve the above purpose, the present application provides the following technical solutions:

[0008] A calculation method for the conduction and scattering process of low-frequency electromagnetic waves in the brain, the calculation method is applied to low-frequency electromagnetic waves incident on and propagating in a target brain tissue, and the calculation method comprises:

[0009] Step S1, obtaining the dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and calculating the unified electromagnetic property parameters of the whole brain by weighted average;

[0010] Step S2, a uniform sphere model matching the diameter of the human cranial cavity anatomical size is constructed, and the boundary condition of the uniform sphere model from inside to outside is defined as an open electromagnetic boundary of tissue to air;

[0011] Step S3, the Maxwell differential control equation is constructed with the uniform sphere model as the calculation domain and the whole brain unified electromagnetic characteristic parameter;

[0012] Step S4, the external electromagnetic wave is input into the differential control equation in different incident parameter combinations through dynamic simulation, and a plurality of intracranial electromagnetic field distribution data sets are obtained;

[0013] Step S5, all intracranial electromagnetic field distribution data sets are discretized into three-dimensional grid data sets, and all three-dimensional grid data sets are divided into sample sets and verification sets;

[0014] Step S6, all sample sets are input into a CNN / RNN hybrid deep neural network for training to establish a mapping relationship from incident parameters to intracranial three-dimensional electromagnetic field distribution, and a trained neural network model is obtained;

[0015] Step S7, all verification sets are input into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship.

[0016] As a further improvement of the present application, in step S7, all verification sets are input into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship, and then, comprising:

[0017] Step S10, the electric field attenuation rate and distribution broadening degree in the electromagnetic response prediction data are extracted and compared with a preset healthy brain tissue benchmark database respectively;

[0018] Step S20, if the comparison result is abnormal increase of dielectric parameters and concentric circle diffusion of electric field, brain edema is determined;

[0019] Step S30, if the comparison result is the appearance of local high absorption area accompanied by loss of scattering directionality, brain tumor is determined.

[0020] As a further improvement of the present application, in step S1, the dielectric constant and magnetic permeability parameters of the target brain tissue at a preset frequency band are obtained, and the whole brain unified electromagnetic characteristic parameter is obtained by weighted average calculation, comprising:

[0021] Step S11, the target brain tissue is partitioned based on a preset voxel size, and the dielectric constant and magnetic permeability parameters of each partition at a preset frequency band are collected;

[0022] Step S12, the whole brain unified dielectric constant and whole brain unified magnetic permeability parameters are obtained by weighted average of formula (1):

[0023] (1);

[0024] wherein, is a whole brain uniform permittivity, is a first partition voxel of the target brain tissue, is a permittivity of the first partition voxel, is a whole brain uniform permeability parameter, is a first partition voxel permeability parameter;

[0025] Step S13, integrating the whole brain uniform permittivity and the whole brain uniform permeability parameter of the same partition voxel into a whole brain uniform electromagnetic characteristic parameter.

[0026] As a further improvement of the present application, step S3, constructing a Maxwell differential control equation with the uniform sphere model as a calculation domain and the whole brain uniform electromagnetic characteristic parameter, comprising:

[0027] Step S31, constructing the Maxwell differential control equation by formula (2):

[0028] (2);

[0029] wherein, is a vector differential operator, is a curl, is an electric field intensity vector of the calculation domain, is an imaginary unit, is a permeability parameter of the calculation domain, is a magnetic field intensity vector of the calculation domain, is a time-harmonic oscillation frequency of the calculation domain, is a permittivity of the calculation domain, is an excitation source current density;

[0030] Step S32, defining a constraint condition of the Maxwell differential control equation according to formula (3) to satisfy the energy dissipation characteristics of the external electromagnetic wave at the far field boundary:

[0031] (3);

[0032] wherein, is a propagation distance of the excitation source, is a scattering field of the excitation source, is a wave number of the external electromagnetic wave.

[0033] As a further improvement of the present application, step S4, the external electromagnetic wave is input into the differential control equation with different incident parameter combinations through dynamic simulation to obtain several brain electromagnetic field distribution data sets, including:

[0034] Step S41, several multi-dimensional parameter matrices are constructed based on several incident parameter combinations of the external electromagnetic wave;

[0035] Step S42, the differential control equation is converted into a discrete algebraic equation set through the finite difference time domain method;

[0036] Step S43, each multi-dimensional parameter matrix is input into the discrete algebraic equation set respectively, and a brain electromagnetic field distribution data set is obtained by solving.

[0037] As a further improvement of the present application, step S5, all brain electromagnetic field distribution data sets are discretized into three-dimensional grid data sets, and all three-dimensional grid data sets are divided into a sample set and a validation set, including:

[0038] Step S51, a right-handed coordinate system with an origin located at the center of the uniform sphere model is defined based on the uniform sphere model, and the Z-axis of the right-handed coordinate system is aligned with the anatomical sagittal axis of the target brain tissue;

[0039] Step S52, the right-handed coordinate system is divided based on the preset voxel size to form several three-dimensional grids;

[0040] Step S53, the current brain electromagnetic field distribution data set is mapped into the three-dimensional grid of the right-handed coordinate system through the trilinear interpolation method;

[0041] Step S54, the zero-value filling is performed on the infinite three-dimensional grid, and the voxel data set with physical labels is obtained based on all three-dimensional grids, that is, a three-dimensional grid data set;

[0042] Step S55, the current voxel data set is divided into a sample set and a validation set.

[0043] As a further improvement of the present application, step S6, all sample sets are input into the CNN / RNN hybrid deep neural network for training to establish the mapping relationship from the incident parameters to the brain three-dimensional electromagnetic field distribution, and a trained neural network model is obtained, including:

[0044] Step S61, all brain electromagnetic field distribution data sets are organized into multi-channel three-dimensional tensors according to the spatial dimensions of the calculation domain;

[0045] Step S62, all multi-dimensional parameter matrices are organized into time-varying feature matrices that are time-sequentially aligned with all multi-channel three-dimensional tensors;

[0046] Step S63, the number of input channels of the CNN branch of the CNN / RNN hybrid deep neural network is defined to be consistent with the dimension of all multi-channel three-dimensional tensors, and the input feature dimension of the RNN branch is consistent with the parameter dimension of all time-varying feature matrices;

[0047] Step S64, the spatial feature vectors of all multi-channel three-dimensional tensors are extracted by the CNN branch, and the dynamic feature vectors of all time-varying feature matrices are extracted by the RNN branch;

[0048] Step S65, all spatial feature vectors and all dynamic feature vectors are merged in the concatenation layer of the CNN / RNN hybrid deep neural network, and the obtained joint feature represents the mapping relationship.

[0049] To achieve the above purpose, the application also provides the following technical scheme:

[0050] A computing system for the transmission and scattering process of low-frequency electromagnetic waves in the brain, the computing system is applied to the computing method as described above, and the computing system comprises:

[0051] An electromagnetic characteristic parameter acquisition module is configured to acquire dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and to obtain unified electromagnetic characteristic parameters of the whole brain through weighted average calculation;

[0052] A uniform sphere model construction module is configured to construct a uniform sphere model with a diameter matching the anatomical size of a human cranial cavity, and to define the boundary conditions of the uniform sphere model from inside to outside as an open electromagnetic boundary from tissue to air;

[0053] A differential control equation construction module is configured to construct a Maxwell differential control equation with the uniform sphere model as a calculation domain and the unified electromagnetic characteristic parameters of the whole brain;

[0054] An electromagnetic wave dynamic simulation module is configured to input external electromagnetic waves with different incident parameter combinations into the differential control equation through dynamic simulation, and to obtain a plurality of brain electromagnetic field distribution data sets;

[0055] An electromagnetic field distribution discretization module is configured to discretize all brain electromagnetic field distribution data sets into three-dimensional grid data sets, and to divide all three-dimensional grid data sets into a sample set and a validation set;

[0056] A hybrid deep neural network training module is configured to input all sample sets into a CNN / RNN hybrid deep neural network for training, to establish a mapping relationship from incident parameters to brain three-dimensional electromagnetic field distribution, and to obtain a trained neural network model;

[0057] An electromagnetic response data prediction module is configured to input all the verification set into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship.

[0058] To achieve the above object, the application further provides the following technical scheme.

[0059] An electronic device includes a processor, and a memory coupled to the processor, the memory storing program instructions executable by the processor; the processor implements the detection calculation method as described above when executing the program instructions stored in the memory.

[0060] To achieve the above object, the application further provides the following technical scheme.

[0061] A computer readable storage medium, the computer readable storage medium has program instructions stored therein, the program instructions are executed by a processor to implement the detection calculation method as described above.

[0062] Effective effect:

[0063] The application obtains the dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and obtains the unified electromagnetic characteristic parameters of the whole brain through weighted average calculation; a uniform sphere model with a diameter matching the anatomical size of the human cranial cavity is constructed, and the boundary conditions of the uniform sphere model from the inside to the outside are defined as an open electromagnetic boundary from tissue to air; the uniform sphere model is taken as a calculation domain, and the Maxwell differential control equation is constructed by using the unified electromagnetic characteristic parameters of the whole brain; the external electromagnetic wave is input into the differential control equation in different incident parameter combinations through dynamic simulation, and a plurality of electromagnetic field distribution data sets in the brain are obtained; all the electromagnetic field distribution data sets in the brain are discretized into three-dimensional grid data sets, and all the three-dimensional grid data sets are divided into a sample set and a verification set; all the sample sets are input into a CNN / RNN hybrid deep neural network for training, so as to establish a mapping relationship from the incident parameters to the three-dimensional electromagnetic field distribution in the brain, and obtain a trained neural network model; all the verification sets are input into the trained neural network model, so as to obtain electromagnetic response prediction data through the mapping relationship. The brain tissue is simplified into a homogeneous isotropic sphere model, the calculation complexity is obviously reduced, and the foundation for efficient solution is laid; then, an electromagnetic wave propagation and scattering physical model based on the Maxwell equation set is constructed, the motion law of the electromagnetic wave in the brain tissue is accurately described; then, the physical model is deeply integrated into the deep learning algorithm, and the three-dimensional electromagnetic field distribution in the brain is efficiently solved; finally, the propagation path, energy distribution and scattering field characteristics of the electromagnetic wave are analyzed, the positioning and qualitative diagnosis of brain edema, tumors and other lesions are realized, key support is provided for brain health evaluation, the efficiency bottleneck of the traditional numerical method in brain electromagnetic calculation is overcome, the timeliness and accuracy of brain disease diagnosis are improved, and reliable technical support is provided for bedside rapid diagnosis of acute conditions such as stroke. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 A step flowchart for an embodiment of the calculation method of the low-frequency electromagnetic wave conduction and scattering process in the brain;

[0065] Figure 2 A graph of the attenuation characteristics of the electromagnetic wave in the brain tissue for an embodiment of the calculation method of the low-frequency electromagnetic wave conduction and scattering process in the brain;

[0066] Figure 3 A comparison graph of the scattering energy for an embodiment of the calculation method of the low-frequency electromagnetic wave conduction and scattering process in the brain;

[0067] Figure 4 A comparison graph of the scattering energy distribution for an embodiment of the calculation method of the low-frequency electromagnetic wave conduction and scattering process in the brain;

[0068] Figure 5A forward scattering energy contrast map for an embodiment of the method of calculating the propagation and scattering of low frequency electromagnetic waves in the brain;

[0069] Figure 6 A backward scattering energy contrast map for an embodiment of the method of calculating the propagation and scattering of low frequency electromagnetic waves in the brain.

[0070] Figure 7 A functional module schematic diagram for an embodiment of the system of calculating the propagation and scattering of low frequency electromagnetic waves in the brain;

[0071] Figure 8 A structural schematic diagram for an embodiment of the electronic device;

[0072] Figure 9 A structural schematic diagram for an embodiment of the storage medium. DETAILED DESCRIPTION

[0073] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0074] The terms "first", "second", "third" in the present application are only used for description purpose, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second", "third" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise explicitly and specifically limited. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative position relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications also change accordingly. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device.

[0075] Reference to“an embodiment” herein means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase“in an embodiment” in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is expressly understood that any of the embodiments described herein can be incorporated into other embodiments.

[0076] As shown in Figure 1 The present embodiment provides an embodiment of a calculation method of the transmission and scattering process of low-frequency electromagnetic waves in the brain, in which the calculation method is applied to low-frequency electromagnetic waves incident on and propagating in target brain tissue.

[0077] Specifically, the calculation method comprises the following steps:

[0078] Step S1, obtaining the dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and calculating the unified electromagnetic characteristic parameters of the whole brain by weighted average.

[0079] Preferably, since the brain waves of human brain tissue are mainly concentrated in the δ wave (0.5-4 Hz) band, θ wave (4-8 Hz) band, α wave (8-13 Hz), β wave (13-30 Hz), and abnormal high frequency (46-50 Hz) representing pathology, the frequency band of the low-frequency electromagnetic waves of the present embodiment can be set to 0.5 Hz to 50 Hz.

[0080] Preferably, the dielectric constant can be obtained by the parallel plate capacitance method or the resonance method; the magnetic permeability parameter can be obtained by measuring the magnetization response by applying an alternating magnetic field through a magnetometer (VSM) or an impedance analyzer, and calculating the relative magnetic permeability. Since the present embodiment is a non-invasive detection and the brain tissue cannot be sampled, the dielectric constant and magnetic permeability parameters of the present embodiment can be obtained through public channels, such as PhysioNet, a biological tissue parameter database (such as Tissue Frequency Chart), and selecting the brain tissue type and setting the frequency range to directly obtain the dielectric constant and magnetic permeability parameters at the corresponding frequency.

[0081] Alternatively, the dielectric constant and magnetic permeability parameters can be obtained by the NRW (Nicolson-Ross-Weir) method, by measuring the transmission loss and phase of the brain tissue sample in a specific frequency band (such as 8 to 20 GHz) to derive the dielectric constant and magnetic permeability parameters, to ensure the calculation efficiency.

[0082] Preferably, the present embodiment preferentially adopts the direct query method to obtain the dielectric constant and magnetic permeability parameters, which can be realized by script grabbing data from public channels.

[0083] Step S2, a uniform sphere model with a diameter matching the human cranial cavity anatomical size is constructed, and the inner-to-outer boundary condition of the uniform sphere model is defined as an open electromagnetic boundary from tissue to air.

[0084] Preferably, since the human cranial cavity anatomical size is known measurement data, the diameter of the uniform sphere model ranges from 18 cm to 22 cm, and the intermediate value of 20 cm is taken as the diameter of the uniform sphere model. If higher precision is required, the individualized cranial cavity size of the patient can be obtained through CT three-dimensional reconstruction.

[0085] Preferably, the uniform sphere model can be created by electromagnetic simulation software COMSOL Multiphysics, ANSYS HFSS, and programming tool MATLAB.

[0086] Specifically, the internal medium parameters of the uniform sphere model are set to the whole brain unified electromagnetic characteristic parameters in step S1, and the external environment of the uniform sphere model is set to air, i.e., the dielectric constant and magnetic permeability parameters of the external environment are both 1.

[0087] Preferably, the open electromagnetic boundary condition is to ensure that the electromagnetic wave does not cause excessive reflection in dynamic simulation. Among them, the sphere surface satisfies the normal component continuity, and the tangential components of the electric field and the magnetic field need to satisfy the tangential component continuity.

[0088] Step S3, taking the uniform sphere model as the calculation domain, the Maxwell differential control equation is constructed by the whole brain unified electromagnetic characteristic parameters.

[0089] Preferably, since the model is a sphere, the Maxwell equation set needs to be converted to the spherical coordinate system (r, θ, ϕ), and the curl operator needs to be expanded according to the spherical coordinates.

[0090] Step S4, through dynamic simulation, the external electromagnetic wave is input into the differential control equation with different incidence parameter combinations to obtain several brain electromagnetic field distribution data sets.

[0091] Preferably, the discrete frequency points can be sampled according to a logarithmic interval, for example, 10 points per decade, and the unit amplitude (1 V / m) and zero phase are set as the reference, which can be expanded to multiple amplitude and phase combinations with ±30° phase difference. Under the condition of uniformly covering the whole space in the spherical coordinate system, take a sampling point every 15° in θ∈[0, π] and φ∈[0, 2π].

[0092] Preferably, dynamic simulation can be performed by electromagnetic simulation software COMSOL Multiphysics, which has an AC / DC module built-in for low-frequency electromagnetic design, supports solving Maxwell equations under quasi-static approximation, and provides a biomedical material library that can directly call brain tissue dielectric parameters and support parameterized scanning function for batch simulation of different incident angle / frequency combinations.

[0093] Preferably, since COMSOL Multiphysics can directly export a three-dimensional electric field / magnetic field distribution matrix (real part + imaginary part), the three-dimensional electric field / magnetic field distribution matrix (real part + imaginary part) can be directly defined as the brain electromagnetic field distribution dataset of the present embodiment, that is, one three-dimensional electric field / magnetic field distribution matrix (real part + imaginary part) is obtained for one incident parameter combination.

[0094] Step S5, discretize all brain electromagnetic field distribution datasets into three-dimensional grid datasets, and divide all three-dimensional grid datasets into sample sets and validation sets.

[0095] Preferably, since the present embodiment adopts a uniform sphere model, the three-dimensional network can adopt regular hexahedral elements to improve computational efficiency, and the voxel size needs to be less than one-tenth of the electromagnetic wave wavelength when dividing the three-dimensional grid.

[0096] Preferably, for continuous electromagnetic field data E(x,y,z) and H(x,y,z), sampling can be performed according to grid node coordinates (X,Y,Z) and stored in complex format of real part + imaginary part or amplitude / phase format.

[0097] Preferably, the division ratio of the sample set and the validation set can be set to 6:4, and if a test set is needed, the sample set: validation set: test set = 6:2:2. At the same time, layering is performed according to incident parameters (frequency, angle) to ensure consistent distribution in each layer in the training / validation / test set.

[0098] Preferably, the format of the three-dimensional grid dataset is a three-dimensional tensor structure, that is, the dimension is Nr×Nθ×Nϕ×C, where C is the number of channels.

[0099] Preferably, the key pseudo code is as follows:

[0100] # Three-dimensional grid discretization

[0101] def discretize_to_grid(field_data, grid_resolution=1e-3):

[0102] Nr, Ntheta, Nphi = int(0.2 / grid_resolution), int(0.2 / grid_resolution), int(0.2 / grid_resolution) # diameter 20cm sphere

[0103] grid_data = np.zeros((Nr, Ntheta, Nphi, 6))

[0104] # 6 channels: Er, Etheta, Ephi, Hr, Htheta, Hphi

[0105] for i, j, k in product(range(nr), range(ntheta), range(nphi)):

[0106] grid_data[i,j,k] = interpolate_field(field_data, (i*dr, j*dtheta, k*dphi))

[0107] return grid_data

[0108] # Dataset split

[0109] def split_dataset(grid_datasets, train_ratio=0.6, val_ratio=0.2):

[0110] n_samples = len(grid_datasets)

[0111] indices = np.random.permutation(n_samples)

[0112] train_idx = indices[:int(n_samples*train_ratio)]

[0113] val_idx = indices[int(n_samples*train_ratio):int

[0114] :::ml-data{name=citationList}

[0115] ```json

[0116] Step S6, input all sample sets into the CNN / RNN hybrid deep neural network for training to establish a mapping relationship from the incident parameters to the three-dimensional electromagnetic field distribution in the brain, and obtain a trained neural network model.

[0117] Preferably, the three-dimensional convolution branch CNN of the CNN / RNN hybrid deep neural network is used to receive the discretized three-dimensional grid data set, and the Nr×Nθ×Nϕ×C three-dimensional tensor structure in the above, C corresponding to the number of channels of the CNN branch.

[0118] Wherein, the CNN branch can adopt a first convolution kernel of 5×5×5, a step of 2, a channel number of 64, and ReLU activation; and three layers of 3×3×3 convolution are subsequently stacked, each layer being followed by batch normalization and maximum pooling.

[0119] Wherein, the time sequence processing branch (RNN) inputs the time-varying feature matrix below into a bidirectional LSTM with a hidden unit number of 128 to capture the frequency domain correlation.

[0120] Next, the flattened features (about 1024 dimensions) output by the CNN are spliced with the time sequence features (256 dimensions) output by the RNN, and a fully connected layer (512 nodes) is used for nonlinear fusion to obtain the mapping relationship.

[0121] Preferably, a composite loss function can be designed: the first term measures the deviation of the predicted electric field value from the standard numerical solution, and the second term forces the network prediction result to satisfy the basic conservation law of electromagnetic field, and the formula is as follows:

[0122]

[0123] Wherein represents the loss function, represents the network predicted electric field, represents the reference solution, represents the weight coefficient, is a 2-norm operator.

[0124] Preferably, the key pseudo code is as follows:

[0125] class HybridModel(nn.Module):

[0126] def __init__(self):

[0127] super().__init__()

[0128] # CNN branch

[0129] self.cnn = nn.Sequential(

[0130] nn.Conv3d(6, 64, kernel_size=5, stride=2),

[0131] nn.BatchNorm3d(64),

[0132] nn.ReLU(),

[0133] nn.MaxPool3d(2),

[0134] # ...more convolutional layers )

[0136] # RNN branch

[0137] self.rnn = nn.LSTM(input_size=6, hidden_size=128,bidirectional=True)

[0138] # Fusion Layer

[0139] self.fc = nn.Sequential(

[0140] nn.Linear(1024+256, 512),

[0141] nn.SiLU(),

[0142] nn.Linear(512, Nr*Nθ*Nϕ*3) # Outputs a three-dimensional field )

[0144] def forward(self, x_grid, x_params):

[0145] cnn_feat = self.cnn(x_grid).flatten(1)

[0146] rnn_out, _ = self.rnn(x_params)

[0147] rnn_feat = rnn_out.mean(dim=1)

[0148] fused = torch.cat([cnn_feat, rnn_feat], dim=1)

[0149] return self.fc(fused).view(-1, Nr,Nθ,Nϕ,3)

[0150] Step S7, input all the verification set into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship.

[0151] Further, step S7, input all the verification set into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship, and then further comprising the following steps:

[0152] Step S10, extract the electric field attenuation rate and distribution broadening degree in the electromagnetic response prediction data, and compare them with the preset healthy brain tissue benchmark database respectively.

[0153] Step S20, if the comparison result is abnormal increase of dielectric parameters and concentric circle diffusion of electric field, it is determined as brain edema.

[0154] Step S30, if the comparison result is the presence of local high absorption area accompanied by loss of scattering directionality, it is determined as brain tumor.

[0155] Three types of diagnostic features are extracted from the comparison results: the attenuation percentage of local electric field intensity relative to the healthy benchmark; the energy ratio change of electromagnetic wave forward scattering and backward scattering; and the distribution broadening degree of scattering energy in a specific angle interval.

[0156] Among them, the brain edema diagnosis standard: when the target area meets the abnormal increase of medium parameters, the concentric circle diffusion of electric field attenuation and the absence of shadow effect at the same time, it is determined as edema lesion.

[0157] Among them, the brain tumor diagnosis standard: when the target area appears local high absorption area accompanied by electromagnetic wave propagation path distortion, and the scattering energy distribution appears directional loss, it is determined as tumor lesion.

[0158] Preferably, referring to Figures 2 to 6 , respectively shows the comparison diagram of various electromagnetic information results of normal brain, brain edema and brain tumor, wherein, Figure 2 is the attenuation characteristic diagram of electromagnetic wave in brain tissue; Figure 3 is the scattering energy comparison diagram; Figure 4 is the scattering energy distribution comparison diagram; Figure 5 is the forward scattering energy comparison diagram; Figure 6 is the backward scattering energy comparison diagram.

[0159] Further, step S1, obtain the dielectric constant and magnetic permeability parameters of the target brain tissue at the preset frequency band, and obtain the unified electromagnetic characteristic parameters of the whole brain through weighted average calculation, which specifically includes the following steps:

[0160] Step S11: Divide the target brain tissue into regions based on the preset voxel size, and collect the dielectric constant and permeability parameters of each region in the preset frequency band.

[0161] Preferably, the voxel size needs to be less than one-tenth of the electromagnetic wave wavelength. If the 20cm diameter model of this embodiment is used, the voxel size is generally between 1mm and 4mm. The frequency band of the low-frequency electromagnetic wave can be set to 0.5Hz to 50Hz according to the above discussion.

[0162] Preferably, the dielectric constant and permeability parameters of this embodiment can be obtained through public channels, such as PhysioNet and biological tissue parameter databases (such as TissueFrequencyChart). By selecting the brain tissue type and setting the frequency range, the dielectric constant and permeability parameters at the corresponding frequency can be obtained directly.

[0163] Step S12, the uniform dielectric constant and uniform magnetic permeability parameters of the whole brain are obtained by weighted averaging using equation (1):

[0164] (1).

[0165] in, To achieve a unified dielectric constant for the entire brain, The first target brain tissue Each partition voxel For the first The dielectric constant of each voxel in the partition. To achieve a unified magnetic permeability parameter for the entire brain, For the first The permeability parameters of each voxel in the partition.

[0166] Step S13: Integrate the uniform dielectric constant and uniform magnetic permeability parameters of the whole brain for the same voxels into uniform electromagnetic property parameters of the whole brain.

[0167] Further, step S3, using a uniform sphere model as the computational domain, constructs Maxwell's differential governing equations through unified electromagnetic property parameters of the whole brain, specifically including the following steps:

[0168] Step S31, construct Maxwell's differential governing equations using equation (2):

[0169] (2).

[0170] in, For vector differential operators, For curl, The electric field intensity vector in the computational domain. The imaginary unit, For the magnetic permeability parameter of the computational domain, a magnetic field intensity vector of the calculation domain, a time harmonic oscillation frequency of the calculation domain, a permittivity of the calculation domain, a current density of the excitation source.

[0171] In step S32, a constraint condition of the Maxwell differential control equation is defined according to formula (3) to meet the energy dissipation characteristics of the external electromagnetic wave at the far-field boundary:

[0172] (3).

[0173] wherein, a propagation distance of the excitation source, a scattering field of the excitation source, a wave number of the external electromagnetic wave.

[0174] Further, in step S4, the external electromagnetic wave is input into the differential control equation in different incident parameter combinations through dynamic simulation to obtain a plurality of electromagnetic field distribution data sets in the brain, including:

[0175] In step S41, a plurality of multi-dimensional parameter matrices are constructed based on a plurality of incident parameter combinations of the external electromagnetic wave.

[0176] Preferably, the incident parameter combinations can be combined according to a frequency range, an amplitude and a phase, and an incident direction.

[0177] wherein, the frequency range: discrete frequency points are divided according to a preset frequency band (such as 1 Hz-1 MHz), and are usually sampled at a logarithmic interval (such as 10 points / decade); the amplitude and the phase: a unit amplitude (1 V / m) and a zero phase are set as a reference, and can be extended to a plurality of amplitude and phase combinations (such as a phase difference of ±30°); the incident direction: the full space is uniformly covered in a spherical coordinate system, and discrete angles of θ∈[0, π] and φ∈[0, 2π] are taken (such as one sampling point every 15°).

[0178] In step S42, the differential control equation is converted into a discrete algebraic equation set through a finite difference time domain method.

[0179] Preferably, the electromagnetic field components can be arranged in a staggered grid, the electric field components are located at the center of the grid edge, the magnetic field components are located at the center of the grid surface, a space surrounding relationship is formed, and a frog jumping type is adopted in time to realize the alternative update of the electric field and the magnetic field components, and the time step interval is Δt / 2.

[0180] Preferably, the Maxwell equation is converted into a component form in a spherical coordinate system (r, θ, φ), and the radial and angular derivatives need to be processed:

[0181] .

[0182] where the pole region (θ = 0, π) needs to be locally grid-refined or coordinate-transformed to avoid singularity, the notation of the formula has appeared in the above, and the meaning of the notation will not be repeated here.

[0183] Preferably, since the present embodiment is open boundary, a perfect matched layer (PML) is added as an absorbing layer outside the calculation domain, and reflection-free is achieved by complex coordinate stretching,

[0184] Preferably, the core Python code is as follows:

[0185] # Pseudo code example: FDTD core iteration in spherical coordinates

[0186] for n in time_steps:

[0187] # Update magnetic field components

[0188] H_phi += (dt / mu) * ( (E_r[theta+1] - E_r[theta]) / (r*dtheta) - (r*E_theta[r+1] - r*E_theta[r]) / (r*dr) )

[0189] # Update electric field components (including PML absorption terms)

[0190] E_r *= (1 - sigma_r*dt / eps) / (1 + sigma_r*dt / eps)

[0191] E_r += (dt / eps) / (1 + sigma_r*dt / eps) * ( (H_phi / (r*sin_theta) -H_phi_prev) / (r*dphi) - (H_theta[r] - H_theta[r-1]) / dr )

[0192] # Boundary processing

[0193] apply_pml_boundary(E, H)

[0194] Step S43, respectively input each multi-dimensional parameter matrix into a discrete algebraic equation set, and solve to obtain an intracranial electromagnetic field distribution dataset.

[0195] Further, step S5, discretize all intracranial electromagnetic field distribution datasets into three-dimensional grid datasets, and divide all three-dimensional grid datasets into sample sets and validation sets, specifically including the following steps:

[0196] Step S51: Define a right-handed coordinate system with its origin located at the center of the uniform sphere model, based on the uniform sphere model. The Z-axis of the right-handed coordinate system is aligned with the anatomical sagittal axis of the target brain tissue.

[0197] Preferably, since the three-dimensional mesh is a regular hexahedron, spherical coordinates are not applicable. In this embodiment, a right-handed coordinate system is established to conform to the orientation of the three-dimensional mesh.

[0198] Preferably, the sagittal axis is one of the three basic axes in human anatomy, extending in the anteroposterior direction and parallel to the ground, forming a spatial positioning reference together with the coronal axis (left-right direction) and the vertical axis (up-down direction). In brain models, the sagittal axis usually corresponds to the midline in the anteroposterior direction, passing through key anatomical landmarks such as the anterior commissure (AC) and posterior commissure (PC).

[0199] Preferably, the anterior commissure (AC) is located anterior to the junction of the left and right hemispheres below the corpus callosum, and the posterior commissure (PC) is located dorsally to the aqueduct of the midbrain. The direction of the sagittal axis can be precisely defined by the AC-PC line, which coincides with the anatomical sagittal axis in a standard brain model.

[0200] Preferably, after determining the anatomical sagittal axis, the origin of the right-hand coordinate system is the geometric center of the uniform sphere model, the Z-axis extends from front to back along the AC-PC line (sagittal axis), the X-axis is defined as the left-right direction (i.e., the coronal axis) according to the right-hand rule, and the Y-axis is determined by X×Z=Y and points in the up-down direction (i.e., the vertical axis).

[0201] Step S52: Divide the right-handed coordinate system based on the preset voxel size to form several three-dimensional meshes.

[0202] Step S53: Map the current brain electromagnetic field distribution dataset to a three-dimensional grid in a right-handed coordinate system using trilinear interpolation.

[0203] Preferably, the trilinear interpolation method is a method for linearly interpolating discrete sampled data on a three-dimensional tensor product grid. The grid allows for non-overlapping arrangement in each dimension, but it is not a triangular finite element analysis grid. This method obtains the target point value by selecting eight adjacent data points around the interpolation point and performing weighted linear interpolation along the x, y, and z axes. Since the formula is existing technology, it is not provided in this embodiment.

[0204] Preferably, the core Python code is as follows:

[0205] import numpy as np

[0206] def trilinear_interp(volume, points):

[0207] """Core function for 3D mesh data interpolation"

[0208] Args:

[0209] volume: Original 3D array (shape: DxHxW)

[0210] points: An array of target point coordinates (shape: Nx3)

[0211] Returns:

[0212] An array of interpolated values ​​(shape: N,)

[0213] """

[0214] # Coordinate normalization

[0215] points = np.asarray(points)

[0216] dims = np.array(volume.shape) - 1

[0217] grid_coords = points * dims

[0218] # Get the indexes of the 8 corner points

[0219] coords = np.floor(grid_coords).astype(int)

[0220] delta = grid_coords - coords

[0221] x0, y0, z0 = coords.T

[0222] x1, y1, z1 = np.clip(coords + 1, 0, dims).T

[0223] # Trilinear Interpolation Calculation

[0224] c000 = volume[x0, y0, z0]

[0225] c100 = volume[x1, y0, z0]

[0226] c010 = volume[x0, y1, z0]

[0227] # ...the other 5 corner points

[0228] c00 = c000*(1-delta[:,0]) + c100*delta[:,0]

[0229] c01 = c010*(1-delta[:,0]) + c110*delta[:,0]

[0230] #... other plane interpolations

[0231] return c0*(1-delta[:,2]) + c1*delta[:,2]

[0232] Step S54, zero padding is performed on the infinite-valued three-dimensional grid, and a voxel dataset with physical labels is obtained based on all three-dimensional grids, i.e., a three-dimensional grid dataset.

[0233] Step S55, the current voxel dataset is divided into a sample set and a validation set.

[0234] Preferably, the division ratio of the sample set and the validation set can be set to 6:4, and if a test set is needed to be added, the sample set: validation set: test set = 6:2:2. At the same time, layering is performed according to the incident parameters (frequency, angle), and it is ensured that each layer is uniformly distributed in the training / validation / test set.

[0235] Further, step S6, all sample sets are input into a CNN / RNN hybrid deep neural network for training to establish a mapping relationship from the incident parameters to the three-dimensional electromagnetic field distribution in the brain, and a trained neural network model is obtained, which specifically includes the following steps:

[0236] Step S61, all brain electromagnetic field distribution datasets are organized into multi-channel three-dimensional tensors according to the spatial dimensions of the calculation domain.

[0237] Preferably, the three-dimensional convolution branch CNN of the CNN / RNN hybrid deep neural network is used to receive the discretized three-dimensional grid dataset, and the Nr×Nθ×Nϕ×C three-dimensional tensor structure in the above, C corresponds to the channel number of the CNN branch.

[0238] Step S62, all multi-dimensional parameter matrices are organized into time-varying feature matrices that are time-sequentially aligned with all multi-channel three-dimensional tensors.

[0239] Step S63, the input channel number of the CNN branch of the CNN / RNN hybrid deep neural network is defined to be consistent with the dimension of all multi-channel three-dimensional tensors, and the input feature dimension of the RNN branch is consistent with the parameter dimension of all time-varying feature matrices.

[0240] Step S64, spatial feature vectors of all multi-channel three-dimensional tensors are extracted by the CNN branch, and dynamic feature vectors of all time-varying feature matrices are extracted by the RNN branch.

[0241] Preferably, the CNN branch can adopt a first convolution kernel of 5x5x5, a step of 2, a number of channels of 64, and ReLU activation; and then 3 layers of 3x3x3 convolution are stacked, each layer being followed by batch normalization and maximum pooling.

[0242] In the time sequence processing branch (RNN), a time-varying feature matrix is input into a bidirectional LSTM with a number of hidden units of 128 to capture frequency domain correlation.

[0243] Step S65, all spatial feature vectors and all dynamic feature vectors are merged in a concatenation layer of the CNN / RNN hybrid deep neural network, and the obtained joint feature representation is the mapping relationship.

[0244] Preferably, the flattened features (about 1024 dimensions) output by the CNN and the time sequence features (256 dimensions) output by the RNN are spliced, nonlinear fusion is performed through a fully connected layer (512 nodes), and the mapping relationship is obtained.

[0245] The embodiment obtains the dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and obtains the unified electromagnetic characteristic parameters of the whole brain through weighted average calculation; a uniform sphere model with a diameter matching the anatomical size of the human cranial cavity is constructed, and the boundary conditions of the uniform sphere model from the inside to the outside are defined as an open electromagnetic boundary from the tissue to the air; the Maxwell differential control equation is constructed by taking the uniform sphere model as the calculation domain and by using the unified electromagnetic characteristic parameters of the whole brain; the external electromagnetic wave is input into the differential control equation in different incident parameter combinations through dynamic simulation, and a plurality of electromagnetic field distribution data sets in the brain are obtained; all the electromagnetic field distribution data sets in the brain are discretized into three-dimensional grid data sets, and all the three-dimensional grid data sets are divided into a sample set and a verification set; all the sample sets are input into the CNN / RNN hybrid deep neural network for training, so as to establish a mapping relationship from the incident parameters to the three-dimensional electromagnetic field distribution in the brain, and obtain a trained neural network model; all the verification sets are input into the trained neural network model, so as to obtain electromagnetic response prediction data through the mapping relationship. The brain tissue is simplified into a homogeneous isotropic sphere model in the embodiment, which obviously reduces the calculation complexity and lays a foundation for efficient solution; then, the electromagnetic wave propagation and scattering physical model based on the Maxwell equation set is constructed, which accurately describes the motion law of the electromagnetic wave in the brain tissue; then, the physical model is deeply integrated into the deep learning algorithm, and the three-dimensional electromagnetic field distribution in the brain is efficiently solved; finally, the propagation path, energy distribution and scattering field characteristics of the electromagnetic wave are analyzed, the positioning and qualitative diagnosis of brain edema, tumors and other lesions are realized, key support is provided for brain health evaluation, the efficiency bottleneck of the traditional numerical method in brain electromagnetic calculation is overcome, the timeliness and accuracy of brain disease diagnosis are improved, and reliable technical support is provided for bedside rapid diagnosis of acute conditions such as stroke.

[0246] As shown in Figure 7 , the embodiment provides an embodiment of a computing system for the transmission and scattering process of low-frequency electromagnetic waves in the brain, which is applied to the computing method in the above embodiment in the embodiment.

[0247] Specifically, the computing system comprises an electromagnetic characteristic parameter acquisition module 1, a uniform sphere model construction module 2, a differential control equation construction module 3, an electromagnetic wave dynamic simulation module 4, an electromagnetic field distribution discretization module 5, a hybrid deep neural network training module 6 and an electromagnetic response data prediction module 7 which are electrically or signal connected in sequence.

[0248] The electromagnetic characteristic parameter acquisition module 1 is configured to acquire dielectric constant and magnetic permeability parameters of the target brain tissue at a preset frequency band, and obtain unified electromagnetic characteristic parameters of the whole brain through weighted average calculation; the uniform sphere model construction module 2 is configured to construct a uniform sphere model with a diameter matching the anatomical size of the human cranial cavity, and define the boundary conditions of the uniform sphere model from inside to outside as an open electromagnetic boundary from tissue to air; the differential control equation construction module 3 is configured to construct Maxwell differential control equations by taking the uniform sphere model as a calculation domain and using the unified electromagnetic characteristic parameters of the whole brain; the electromagnetic wave dynamic simulation module 4 is configured to input external electromagnetic waves with different incident parameter combinations into the differential control equations through dynamic simulation to obtain a plurality of electromagnetic field distribution data sets in the brain; the electromagnetic field distribution discretization module 5 is configured to discretize all the electromagnetic field distribution data sets in the brain into three-dimensional grid data sets, and divide all the three-dimensional grid data sets into a sample set and a validation set; the hybrid deep neural network training module 6 is configured to input all the sample set into a CNN / RNN hybrid deep neural network for training, so as to establish a mapping relationship from incident parameters to three-dimensional electromagnetic field distribution in the brain, and obtain a trained neural network model; and the electromagnetic response data prediction module 7 is configured to input all the validation set into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship.

[0249] Further, the computing system comprises an electromagnetic response data comparison module, a brain edema determination module and a brain tumor determination module which are electrically or signal connected in sequence.

[0250] The electromagnetic response data comparison module is configured to extract the electric field attenuation rate and the distribution broadening degree in the electromagnetic response prediction data, and compare them with a preset healthy brain tissue benchmark database respectively; the brain edema determination module is configured to determine brain edema if the comparison result is that the dielectric parameter abnormally increases and the electric field presents concentric circle diffusion; and the brain tumor determination module is configured to determine brain tumor if the comparison result is that a local high absorption area appears accompanied by a lack of scattering directionality.

[0251] Further, the electromagnetic characteristic parameter acquisition module 1 specifically comprises a first electromagnetic characteristic parameter acquisition unit, a second electromagnetic characteristic parameter acquisition unit and a third electromagnetic characteristic parameter acquisition unit which are electrically or signal connected in sequence; and the third electromagnetic characteristic parameter acquisition unit is electrically or signal connected with the uniform sphere model construction module 2.

[0252] The first electromagnetic characteristic parameter acquisition unit is configured to divide the target brain tissue based on a preset voxel size, and acquire dielectric constant and magnetic permeability parameters of each division at a preset frequency band.

[0253] The second electromagnetic characteristic parameter acquisition unit is configured to obtain unified dielectric constant and magnetic permeability parameters of the whole brain through weighted average calculation according to formula (1):

[0254] (1).

[0255] wherein, is a whole brain uniform permittivity, is a first partition voxel of the target brain tissue, is a permittivity of the first partition voxel, is a whole brain uniform permeability parameter, is a first partition voxel permeability parameter.

[0256] The third electromagnetic property parameter acquisition unit is configured to integrate the whole brain uniform permittivity and the whole brain uniform permeability parameter of the same partition voxel into a whole brain uniform electromagnetic property parameter.

[0257] Further, the differential control equation construction module 3 specifically comprises a first differential control equation construction unit and a second differential control equation construction unit connected in sequence in an electrical or signal manner; the first differential control equation construction unit is electrically or signal connected with the uniform sphere model construction module 2, and the second differential control equation construction unit is electrically or signal connected with the electromagnetic wave dynamic simulation module 4.

[0258] The first differential control equation construction unit is configured to construct the Maxwell differential control equation through equation (2):

[0259] (2).

[0260] wherein, is a vector differential operator, is a curl, is an electric field intensity vector of a calculation domain, is an imaginary unit, is a permeability parameter of the calculation domain, is a magnetic field intensity vector of the calculation domain, is a time-harmonic oscillation frequency of the calculation domain, is a permittivity of the calculation domain, is an excitation source current density.

[0261] The second differential control equation construction unit is configured to define a constraint condition of the Maxwell differential control equation according to equation (3) to satisfy an energy dissipation characteristic of the external electromagnetic wave at the far-field boundary:

[0262] (3).

[0263] wherein, is a propagation distance of the excitation source, a scattered field of an excitation source, a wave number of an external electromagnetic wave.

[0264] Further, the electromagnetic wave dynamic simulation module 4 specifically comprises a first electromagnetic wave dynamic simulation unit, a second electromagnetic wave dynamic simulation unit, and a third electromagnetic wave dynamic simulation unit connected in sequence; the first electromagnetic wave dynamic simulation unit is electrically or signal connected with the second differential control equation construction unit; and the third electromagnetic wave dynamic simulation unit is electrically or signal connected with the electromagnetic field distribution discretization module 5.

[0265] The first electromagnetic wave dynamic simulation unit is configured to construct a plurality of multi-dimensional parameter matrices based on a plurality of incident parameter combinations of an external electromagnetic wave; the second electromagnetic wave dynamic simulation unit is configured to convert the differential control equation into a discrete algebraic equation set by using a finite difference time domain method; and the third electromagnetic wave dynamic simulation unit is configured to input each multi-dimensional parameter matrix into the discrete algebraic equation set respectively, and solve to obtain an intracerebral electromagnetic field distribution data set.

[0266] Further, the electromagnetic field distribution discretization module 5 specifically comprises a first electromagnetic field distribution discretization unit, a second electromagnetic field distribution discretization unit, a third electromagnetic field distribution discretization unit, a fourth electromagnetic field distribution discretization unit, and a fifth electromagnetic field distribution discretization unit connected in sequence; the first electromagnetic field distribution discretization unit is electrically or signal connected with the third electromagnetic wave dynamic simulation unit; and the fifth electromagnetic field distribution discretization unit is electrically or signal connected with the hybrid deep neural network training module 6.

[0267] The first electromagnetic field distribution discretization unit is configured to define a right-handed coordinate system with an origin located at the center of a uniform sphere model based on the uniform sphere model, and the Z-axis of the right-handed coordinate system is aligned with the anatomical sagittal axis of the target brain tissue; the second electromagnetic field distribution discretization unit is configured to divide the right-handed coordinate system based on a preset voxel size to form a plurality of three-dimensional grids; the third electromagnetic field distribution discretization unit is configured to map the current intracerebral electromagnetic field distribution data set into the three-dimensional grids of the right-handed coordinate system by using a trilinear interpolation method; the fourth electromagnetic field distribution discretization unit is configured to perform zero value padding on the infinite value three-dimensional grids, and obtain a voxel data set with physical labels based on all the three-dimensional grids, i.e., a three-dimensional grid data set; and the fifth electromagnetic field distribution discretization unit is configured to divide the current voxel data set into a sample set and a validation set.

[0268] Furthermore, the hybrid deep neural network training module 6 specifically includes a first hybrid deep neural network training unit, a second hybrid deep neural network training unit, a third hybrid deep neural network training unit, a fourth hybrid deep neural network training unit, and a fifth hybrid deep neural network training unit that are electrically or signal-connected in sequence; the first hybrid deep neural network training unit and the fifth electromagnetic field distribution discretization unit are electrically or signal-connected.

[0269] The training unit comprises five components: a first hybrid deep neural network (HNN) training unit, a second hybrid deep neural network training unit, and a third hybrid deep neural network training unit. The first unit organizes all brain electromagnetic field distribution datasets into multi-channel three-dimensional tensors according to the spatial dimension of the computational domain. The second unit organizes all multi-dimensional parameter matrices into time-varying feature matrices aligned with the temporal sequence of all multi-channel three-dimensional tensors. The third unit defines the number of input channels of the CNN branch of the CNN / RNN hybrid deep neural network as consistent with the dimension of all multi-channel three-dimensional tensors, and the input feature dimension of the RNN branch as consistent with the parameter dimension of all time-varying feature matrices. The fourth unit extracts spatial feature vectors from all multi-channel three-dimensional tensors through the CNN branch and extracts dynamic feature vectors from all time-varying feature matrices through the RNN branch. The fifth unit merges all spatial feature vectors and all dynamic feature vectors at the concatenation layer of the CNN / RNN hybrid deep neural network, and the resulting joint feature representation is the mapping relationship.

[0270] It should be noted that this embodiment is a functional module embodiment based on the above method embodiment. For the preferred, extended, limited, exemplified and principle explanation parts of this embodiment, please refer to the above embodiment. This embodiment will not repeat them.

[0271] The embodiment obtains the dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and obtains the unified electromagnetic characteristic parameters of the whole brain through weighted average calculation; a uniform sphere model with a diameter matching the anatomical size of the human cranial cavity is constructed, and the boundary conditions of the uniform sphere model from the inside to the outside are defined as an open electromagnetic boundary from the tissue to the air; the Maxwell differential control equation is constructed by taking the uniform sphere model as the calculation domain and by using the unified electromagnetic characteristic parameters of the whole brain; the external electromagnetic wave is input into the differential control equation in different incident parameter combinations through dynamic simulation, and a plurality of electromagnetic field distribution data sets in the brain are obtained; all the electromagnetic field distribution data sets in the brain are discretized into three-dimensional grid data sets, and all the three-dimensional grid data sets are divided into a sample set and a verification set; all the sample sets are input into a CNN / RNN hybrid deep neural network for training, so as to establish a mapping relationship from the incident parameters to the three-dimensional electromagnetic field distribution in the brain, and obtain a trained neural network model; all the verification sets are input into the trained neural network model, so as to obtain electromagnetic response prediction data through the mapping relationship. The brain tissue is simplified into a homogeneous isotropic sphere model in the embodiment, which obviously reduces the calculation complexity and lays a foundation for efficient solution; then, an electromagnetic wave propagation and scattering physical model based on the Maxwell equation set is constructed, which accurately describes the motion law of the electromagnetic wave in the brain tissue; then, the physical model is deeply integrated into the deep learning algorithm, and the three-dimensional electromagnetic field distribution in the brain is efficiently solved; finally, the propagation path, energy distribution and scattering field characteristics of the electromagnetic wave are analyzed, the positioning and qualitative diagnosis of brain edema, tumors and other lesions are realized, key support is provided for brain health evaluation, the efficiency bottleneck of the traditional numerical method in brain electromagnetic calculation is overcome, the timeliness and accuracy of brain disease diagnosis are improved, and reliable technical support is provided for bedside rapid diagnosis of acute conditions such as stroke.

[0272] Figure 8 FIG. 1 is a structural schematic diagram of an electronic device according to an embodiment of the present application. Figure 8 As shown in FIG. 1, the electronic device 8 includes a processor 81 and a memory 82 coupled to the processor 81.

[0273] The memory 82 stores program instructions for implementing the federated learning-based government data group collaborative energy-saving method of any of the above embodiments.

[0274] The processor 81 is configured to execute the program instructions stored in the memory 82 to perform federated learning-based government data group collaborative energy saving.

[0275] The processor 81 can also be called a CPU (Central Processing Unit). The processor 81 can be an integrated circuit chip having a processing capability of signals. The processor 81 can also be a general processor, a DSP (Digital Signal Processor), an ASIC (Application-Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. The general processor can be a microprocessor or the processor can also be any conventional processor.

[0276] Further, Figure 9 A structural schematic diagram of the storage medium of an embodiment of the present application is shown in FIG. 9. Figure 9 The storage medium 9 of the embodiment of the present application stores program instructions 91 capable of implementing all the methods described above, wherein the program instructions 91 can be stored in the storage medium in the form of a software product, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a ROM (Read-Only Memory), a RAM (Random Access Memory), a magnetic disk or an optical disk, and various media capable of storing program codes, or a terminal device such as a computer, a server, a mobile phone, a tablet, etc.

[0277] In several embodiments provided in the present application, it should be understood that the disclosed system, system and method can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, another division manner can be used, for example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, systems or units, and can be in the form of electrical, mechanical, signal or other forms.

[0278] In addition, the various functional units in the embodiments of the present application can be integrated in one processing unit, or each can exist physically as a separate unit, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware, or in the form of a software functional unit. The above is only an implementation of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for calculating the conduction and scattering process of low-frequency electromagnetic waves in the brain, the method being applied to low-frequency electromagnetic waves incident on and propagating in target brain tissue, characterized in that... The calculation method includes: Step S1: Obtain the dielectric constant and permeability parameters of the target brain tissue in a preset frequency band, and calculate the unified electromagnetic property parameters of the whole brain by weighted average. Step S2: Construct a uniform spherical model with a diameter matching the anatomical dimensions of the human cranial cavity, and define the boundary conditions of the uniform spherical model from the inside to the outside as an open electromagnetic boundary from tissue to air. Step S3: Using the uniform sphere model as the computational domain and the unified electromagnetic property parameters of the whole brain, construct the Maxwell differential control equations; Step S4: By dynamically simulating external electromagnetic waves with different combinations of incident parameters, inputting them into the differential control equation, several datasets of electromagnetic field distribution in the brain are obtained. Step S5: Discretize all brain electromagnetic field distribution datasets into three-dimensional grid datasets, and divide all three-dimensional grid datasets into sample sets and validation sets; Step S6: Input all sample sets into a CNN / RNN hybrid deep neural network for training to establish a mapping relationship from incident parameters to the three-dimensional electromagnetic field distribution in the brain, and obtain the trained neural network model. Step S7: Input all validation sets into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship.

2. The calculation method according to claim 1, characterized in that, Step S1: Obtain the dielectric constant and permeability parameters of the target brain tissue in a preset frequency band, and calculate the unified electromagnetic property parameters of the whole brain by weighted average, including: Step S11: The target brain tissue is divided into partitions based on a preset voxel size, and the dielectric constant and permeability parameters of each partition are collected in a preset frequency band. Step S12, the uniform dielectric constant and uniform magnetic permeability parameters of the whole brain are obtained by weighted averaging using equation (1): （1）； in, To achieve a unified dielectric constant for the entire brain, The first of the target brain tissue Each partition voxel For the first The dielectric constant of each voxel in the partition. To achieve a unified magnetic permeability parameter for the entire brain, For the first The permeability parameters of each voxel in the partition; Step S13: Integrate the unified dielectric constant and unified magnetic permeability parameters of the whole brain for the same voxels into unified electromagnetic property parameters of the whole brain.

3. The calculation method according to claim 1, characterized in that, Step S3, using the uniform sphere model as the computational domain and the unified electromagnetic property parameters of the whole brain as the construction domain, constructs the Maxwell differential governing equations, including: Step S31, construct the Maxwell differential governing equation using equation (2): （2）； in, For vector differential operators, For curl, Let be the electric field intensity vector of the computational domain. The imaginary unit, Let be the permeability parameter of the computational domain. Let be the magnetic field strength vector of the computational domain. The time-resonance oscillation frequency of the computational domain. The dielectric constant of the computational domain is given. The excitation source current density; Step S32, according to equation (3), the constraint conditions of the Maxwell differential governing equations are defined to satisfy the energy dissipation characteristics of the external electromagnetic wave at the far-field boundary: （3）； in, To measure the propagation distance of the stimulus source, The scattered field of the excitation source, The wave number of the external electromagnetic wave.

4. The calculation method according to claim 1, characterized in that, Step S4: By dynamically simulating external electromagnetic waves with different combinations of incident parameters, inputting them into the differential control equation, several datasets of brain electromagnetic field distributions are obtained, including: Step S41: Construct several multidimensional parameter matrices based on several combinations of incident parameters of external electromagnetic waves; Step S42: The differential control equations are converted into a system of discrete algebraic equations using the finite difference time-domain method. Step S43: Input each multidimensional parameter matrix into the discrete algebraic equation system and solve to obtain a brain electromagnetic field distribution dataset.

5. The calculation method according to claim 2, characterized in that, Step S5: Discretize all brain electromagnetic field distribution datasets into three-dimensional mesh datasets, and divide all three-dimensional mesh datasets into sample sets and validation sets, including: Step S51: Define a right-handed coordinate system with its origin located at the center of the uniform sphere model based on the uniform sphere model, and align the Z-axis of the right-handed coordinate system with the anatomical sagittal axis of the target brain tissue. Step S52: Divide the right-handed coordinate system based on the preset voxel size to form several three-dimensional meshes; Step S53: Map the current brain electromagnetic field distribution dataset to the three-dimensional grid of the right-hand coordinate system using trilinear interpolation. Step S54: Zero-value filling is performed on the three-dimensional mesh with countless values. Based on all the three-dimensional meshes, a voxel dataset with physical labels is obtained, which is a three-dimensional mesh dataset. Step S55: Divide the current voxel dataset into a sample set and a validation set.

6. The calculation method according to claim 1, characterized in that, Step S6: Input all sample sets into a CNN / RNN hybrid deep neural network for training to establish a mapping relationship from incident parameters to the three-dimensional electromagnetic field distribution in the brain, resulting in a trained neural network model, including: Step S61: Organize all brain electromagnetic field distribution datasets into multi-channel three-dimensional tensors according to the spatial dimension of the computational domain; Step S62: Organize all multidimensional parameter matrices into time variation feature matrices aligned with the time series of all multi-channel three-dimensional tensors; Step S63: Define the number of input channels of the CNN branch of the CNN / RNN hybrid deep neural network to be consistent with the dimension of all multi-channel three-dimensional tensors, and define the dimension of the input features of the RNN branch to be consistent with the parameter dimension of all time-varying feature matrices; Step S64: Extract the spatial feature vectors of all multi-channel three-dimensional tensors through the CNN branch, and extract the dynamic feature vectors of all time-varying feature matrices through the RNN branch; Step S65: All spatial feature vectors and all dynamic feature vectors are merged in the splicing layer of the CNN / RNN hybrid deep neural network, and the resulting joint feature representation is the mapping relationship.

7. A computational system for the conduction and scattering process of low-frequency electromagnetic waves in the brain, said computational system being applied to the computational method as described in any one of claims 1 to 6, characterized in that, The computing system includes: The electromagnetic property parameter acquisition module is used to acquire the dielectric constant and magnetic permeability parameters of the target brain tissue in a preset frequency band, and to obtain the unified electromagnetic property parameters of the whole brain by weighted average calculation. The uniform sphere model construction module is used to construct a uniform sphere model with a diameter matching the anatomical size of the human cranial cavity, and defines the boundary conditions of the uniform sphere model from the inside to the outside as an open electromagnetic boundary from tissue to air. The differential control equation construction module is used to construct Maxwell's differential control equations using the uniform sphere model as the computational domain and the unified electromagnetic property parameters of the whole brain. The electromagnetic wave dynamic simulation module is used to input external electromagnetic waves with different combinations of incident parameters into the differential control equation through dynamic simulation to obtain several brain electromagnetic field distribution datasets. The electromagnetic field distribution discretization module is used to discretize all brain electromagnetic field distribution datasets into three-dimensional grid datasets, and to divide all three-dimensional grid datasets into sample sets and validation sets. The hybrid deep neural network training module is used to input all sample sets into the CNN / RNN hybrid deep neural network for training, in order to establish the mapping relationship from the incident parameters to the three-dimensional electromagnetic field distribution in the brain, and obtain the trained neural network model. The electromagnetic response data prediction module is used to input all validation sets into the trained neural network model to obtain electromagnetic response prediction data through the mapping relationship.

8. An electronic device, characterized in that, The method includes a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the computation method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that, when executed by a processor, enable the computation method as described in any one of claims 1 to 6.

Images