Neurological rehabilitation electric stimulation device and control method thereof

Abstract

The application provides a neurology rehabilitation electric stimulation device and a control method thereof. Based on a patient head finite element model, three-dimensional electric field components in the brain when a unit current is applied by a candidate electrode contact area relative to a return electrode contact area are solved respectively, and an electric field transfer matrix is established. A target area direction projection matrix is formed according to a brain area target point and a target electric field direction, and is reduced in order under a reconstruction error limit. With the target area direction electric field mean value and the non-target area electric field upper quantile value as two optimization indexes, a candidate current vector is obtained under the constraints of current conservation, output channel and current density. The candidate current vector is quantized according to the device current step, the brain electric field is calculated again by using the un-reduced electric field transfer matrix, the executable current configuration that meets the constraints is retained, and the constant current output and abnormal stop are controlled accordingly. By adopting the above scheme, the consistency between the model calculation result and the device executable current configuration can be improved within the applicable range of the finite element model and the rated parameters of the device.

Description

Technical Field

[0001] This application relates to the field of neurological rehabilitation technology, and more specifically, to a neurological rehabilitation electrical stimulation device and its control method. Background Technology

[0002] Transcranial electrical stimulation (TCS) injects controlled current into the head tissues via scalp electrodes. Under low-frequency quasi-static conditions, the tissue potential satisfies the current continuity equation, and the brain electric field depends on the head tissue morphology, tissue conductivity, electrode contact location, and current in each channel. Individualized parameter configurations are typically based on finite element models established from magnetic resonance imaging (MRI) images to calculate the brain electric field generated by different current configurations, and accordingly select the current parameters for the stimulation device output.

[0003] In existing individualized parameter configuration methods, inconsistencies may still arise between model calculations and device outputs. For example, if the electric field vector magnitudes of grid nodes are directly used as linearly superimposed quantities, or if low-dimensional results containing only target area information are used to estimate the electric field in the non-target area, the resulting electric field evaluation will deviate from the vector superposition law of quasi-static electric fields. Furthermore, if the optimization process does not take into account the device's current step size, the number of output channels, and electrode current density limitations, continuous numerical solutions may alter the electric fields in the target and non-target areas after conversion to device output values. Therefore, improving the consistency between model calculation results and the device's executable current configuration within the applicable range of finite element models and device rated parameters has become a challenge for the industry. Summary of the Invention

[0004] This application provides a neurological rehabilitation electrical stimulation device and its control method, which can improve the consistency between the model calculation results and the device's executable current configuration within the applicable range of the finite element model and the device's rated parameters.

[0005] In a first aspect, this application provides a control method for a neurological rehabilitation electrical stimulation device, comprising: A finite element simulation model containing the electrical conductivity of head tissue and the equivalent contact area of ​​electrodes on the outer surface of the scalp is obtained. Non-zero unit currents are applied sequentially between the equivalent contact areas of candidate electrodes and the equivalent contact areas of return electrodes. The signed electric field components obtained by arranging the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix. Obtain the target points, target node weights and target electric field directions in the brain region. Form a target region direction projection matrix based on the electric field transfer matrix. Reduce the order of the target region direction projection matrix under reconstruction error constraints to obtain the target region reduced order matrix. Using the electrode current vector as a variable, the mean electric field in the target area direction is calculated based on the target area reduction matrix, and the quantile value of the electric field in the non-target area is calculated based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target area direction and decrease the quantile value of the electric field in the non-target area. Under the constraints of current conservation, total injection current, single electrode current, electrode current density and number of output channels, candidate current vectors are obtained. The candidate current vector is quantized according to the current step size of the neurological rehabilitation electrical stimulation device. The quantized intracranial electric field is recalculated using the electric field transfer matrix. The executable current configuration that satisfies the constraints is retained. Channel control data is generated according to the executable current configuration. After the electrode contact impedance verification is passed, the constant current output is controlled and the output is stopped when the actual current or output voltage is abnormal.

[0006] In some embodiments, obtaining a finite element simulation model that includes the electrical conductivity of head tissue and the equivalent contact area of ​​electrodes on the outer surface of the scalp specifically includes: Receive T1-weighted and T2-weighted magnetic resonance images of the head of the target patient and register the two types of images to the same spatial coordinate system; The registered images are segmented to obtain a three-dimensional head model that includes at least the scalp, skull, cerebrospinal fluid, gray matter, and white matter. Configure low-frequency conductivity parameters for each tissue region in the three-dimensional head model, and record the source of the conductivity parameters; Based on the preset electrode positioning coordinates, an equivalent contact area for a return electrode and multiple equivalent contact areas for candidate electrodes are defined on the corresponding boundary of the outer scalp surface of the three-dimensional head model. The three-dimensional head model with completed conductivity configuration and electrode equivalent contact area delineation is meshed, and mesh convergence verification is performed based on the difference in intracerebral electric field components per unit current before and after mesh refinement, thus obtaining the finite element simulation model.

[0007] In some embodiments, a non-zero unit current is applied sequentially between the equivalent contact region of the candidate electrode and the equivalent contact region of the return electrode. The signed electric field components obtained according to the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix, which specifically includes: In the finite element simulation model, a low-frequency quasi-static current field is set, the outer surface of the head except for the equivalent contact area of ​​each electrode is set as an electrically insulating boundary, and a specification condition for determining the zero potential point is set. The equivalent contact region of a candidate electrode is sequentially set as a non-zero unit current injection boundary, the equivalent contact region of the return electrode is set as an equal amount of reverse current boundary, and the equivalent contact regions of the remaining candidate electrodes are set as zero normal current boundaries. The signed electric field components of each grid node in the brain in three coordinate directions are calculated separately when the non-zero unit current is applied to each candidate electrode. The column vector of electric field components corresponding to each candidate electrode is formed according to the fixed order of grid nodes and coordinate directions. The electric field component column vectors are divided by the non-zero unit current and arranged in the order of candidate electrodes to obtain the electric field transfer matrix; the product of the electric field transfer matrix and the candidate electrode current vectors is used to calculate the signed electric field components in the brain, and the return electrode current is determined by the current conservation condition.

[0008] In some embodiments, a target region direction projection matrix is ​​formed based on the electric field transfer matrix, and the target region direction projection matrix is ​​reduced in order under reconstruction error constraints to obtain a target region reduced-order matrix. Specifically, this includes: The brain region target points are registered to the finite element simulation model to obtain the target region mesh node set; according to the target electric field direction of each target region mesh node, the three coordinate direction rows corresponding to the electric field transfer matrix are projected to form the target region direction projection matrix. Singular value decomposition is performed on the target area projection matrix; The truncation order is determined based on the relative reconstruction error between the target area direction projection matrices before and after truncation, and the truncation order is increased when the relative reconstruction error exceeds a preset error limit. The target region reduction matrix is ​​constructed using the determined truncation order and the corresponding singular vectors and singular values, while the unreduced electric field transfer matrix is ​​retained.

[0009] In some embodiments, using the electrode current vector as a variable, the mean electric field in the target region direction is calculated based on the target region reduced-order matrix, and the upper quantile of the electric field in the non-target region is calculated based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target region direction and decrease the upper quantile of the electric field in the non-target region. Specifically, these objectives include: Nodes in the brain grid that do not belong to the target area grid node set are designated as non-target area grid nodes. The directional electric fields of each target area grid node calculated by the target area reduction matrix according to the target area node weights are weighted and summed to obtain the mean value of the target area directional electric field. The electric field vector amplitude of each non-target area grid node is calculated according to the electric field transfer matrix, and the preset upper quantile value of the electric field vector amplitude is calculated to obtain the upper quantile value of the non-target area electric field. Increasing the mean electric field value in the target area and decreasing the quantile value in the electric field in the non-target area are two independent optimization objectives, and the maximum value of the electric field vector amplitude of the grid nodes in the non-target area is not more than a preset limit is used as a constraint.

[0010] In some embodiments, obtaining the candidate current vector under constraints of current conservation, total injected current, single electrode current, electrode current density, and number of output channels specifically includes: A constrained multi-objective model is established with the following constraints: the algebraic sum of the currents of each electrode is zero; the sum of the currents of the same polarity does not exceed the upper limit of the total injected current; the current of a single electrode does not exceed the corresponding upper limit; the ratio of the current of a single electrode to the corresponding contact area does not exceed the upper limit of the electrode current density; and the number of non-zero current electrodes does not exceed the number of output channels. Set the population size, maximum number of iterations, crossover parameter, and mutation parameter for the non-dominated sorting genetic algorithm, and set a fixed reference point for calculating the Pareto front hypervolume; The electrode currents of individuals in the population are encoded with real numbers, and non-dominated sorting is performed using the constraint dominance relationship. Individuals with larger crowding distances are preferentially retained in the same non-dominated layer to obtain the Pareto solution set for the current iteration. The population update is repeated until the relative hypervolume change rate of the Pareto front for a preset number of consecutive rounds does not exceed a preset convergence limit or the maximum number of iterations is reached. The electrode current vector in the final Pareto solution set is then used as the candidate current vector.

[0011] In some embodiments, the process of recalculating the quantized intracranial electric field using the electric field transfer matrix, retaining the executable current configuration that satisfies the constraints, and generating channel control data based on the executable current configuration includes: Each candidate current vector is quantized according to the current step size, and the mean electric field in the target area, the upper quantile of the electric field in the non-target area, and the maximum electric field in the non-target area are recalculated using the electric field transfer matrix. Candidate current vectors that do not meet the constraints are deleted. Select executable current configurations from the remaining candidate current vectors to generate channel control data including channel identifier, current amplitude, current polarity, rise time, sustain time, and fall time; detect the electrode contact impedance of each channel, output stimulation current after impedance verification, and perform feedback control based on actual current and output voltage.

[0012] Secondly, this application provides a neurological rehabilitation electrical stimulation device, comprising: The acquisition module is used to acquire a finite element simulation model containing the electrical conductivity of head tissue and the equivalent contact area of ​​electrodes on the outer surface of the scalp. Non-zero unit currents are applied sequentially between the equivalent contact areas of candidate electrodes and the equivalent contact areas of return electrodes. The signed electric field components obtained according to the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix. The processing module is used to obtain the target points in the brain region, the weights of the target nodes, and the direction of the target electric field. It forms a target direction projection matrix based on the electric field transfer matrix and reduces the order of the target direction projection matrix under the constraint of reconstruction error to obtain the target reduced order matrix. The processing module is also used to calculate the mean electric field in the target area direction based on the target area reduction matrix, using the electrode current vector as a variable, and to calculate the quantile value of the electric field in the non-target area based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target area direction and decrease the quantile value of the electric field in the non-target area. Under the constraints of current conservation, total injection current, single electrode current, electrode current density and number of output channels, candidate current vectors are obtained. The execution module is used to quantize the candidate current vector according to the current step size of the neurological rehabilitation electrical stimulation device, recalculate the quantized intracranial electric field using the electric field transfer matrix, retain the executable current configuration that satisfies the constraints, generate channel control data according to the executable current configuration, control constant current output after the electrode contact impedance verification is passed, and stop output when the actual current or output voltage is abnormal.

[0013] Thirdly, this application provides a computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the control method of the neurological rehabilitation electrical stimulation device described above.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described control method for a neurological rehabilitation electrical stimulation device.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The neurological rehabilitation electrical stimulation device and its control method provided in this application first obtain the signed three-dimensional electric field components by solving for non-zero unit current, and establish an electric field transfer matrix accordingly. Then, a target area direction projection matrix is ​​formed based on the target point in the brain, the target area node weights, and the target electric field direction. Under the reconstruction error constraint, a target area reduced-order matrix is ​​obtained. In the optimization process, the target area direction electric field is calculated by the target area reduced-order matrix, and the non-target area electric field is calculated by the unreduced electric field transfer matrix. After the candidate current vector is quantized according to the device current step size, the whole brain electric field is recalculated by the electric field transfer matrix. When the recalculation result satisfies the device constraints and the non-target area electric field limits, an executable current configuration is formed and used for constant current output and abnormal stop control.

[0016] Therefore, the signed electric field components conform to the linear superposition condition of low-frequency quasi-static electric fields, and thus the electric field transfer matrix can be used to recalculate the intracranial electric field corresponding to different electrode current vectors. The order reduction of the target area projection matrix is ​​limited by the reconstruction error, and the non-target area electric field is still calculated by the unreduced matrix, which can retain the spatial information required for non-target area electric field evaluation while reducing the amount of repetitive calculation in the target area. Incorporating the current step size, current density, and number of output channels into the constraints, and recalculating the electric field for the quantized candidate current vectors, can reduce the deviation between the continuous optimization results and the actual device settings. Monitoring the actual current, output voltage, and electrode contact impedance is used to identify constant current output anomalies. Therefore, this application can improve the consistency between the model calculation results and the device's executable current configuration within the applicable range of the finite element model and device rated parameters. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart of a control method for a neurological rehabilitation electrical stimulation device according to some embodiments of this application; Figure 2 This is an exemplary flowchart illustrating the determination of a finite element simulation model according to some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the establishment of a target region reduction matrix according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a neurological rehabilitation electrical stimulation device according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing a control method for a neurological rehabilitation electrical stimulation device according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 The figure is an exemplary flowchart of a control method for a neurological rehabilitation electrical stimulation device according to some embodiments of this application. The control method for the neurological rehabilitation electrical stimulation device mainly includes the following steps: In step 101, a finite element simulation model containing the electrical conductivity of the head tissue and the equivalent contact area of ​​the electrodes on the outer surface of the scalp is obtained. Non-zero unit currents are applied sequentially between the equivalent contact areas of the candidate electrodes and the equivalent contact areas of the return electrodes. The signed electric field components obtained according to the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix.

[0020] In some embodiments, reference Figure 2The figure is an exemplary flowchart illustrating the determination of a finite element simulation model according to some embodiments of this application. The finite element simulation model, which includes the electrical conductivity of the head tissue and the equivalent contact area of ​​the electrodes on the outer surface of the scalp, can be obtained in this application by the following steps: In step 1011, T1-weighted magnetic resonance images and T2-weighted magnetic resonance images of the target patient's head are received, and the two types of images are registered to the same spatial coordinate system; In step 1012, the registered image is segmented to obtain a three-dimensional head model that includes at least the scalp, skull, cerebrospinal fluid, gray matter, and white matter. In step 1013, conductivity parameters under low-frequency conditions are configured for each tissue region in the three-dimensional head model, and the source of the conductivity parameters is recorded. In step 1014, the equivalent contact area of ​​the reflux electrode and the equivalent contact area of ​​the candidate electrode are defined on the model boundary corresponding to the outer surface of the scalp. In step 1015, the three-dimensional head model with completed conductivity configuration and electrode equivalent contact area delineation is meshed, and mesh convergence verification is performed based on the difference in intracerebral electric field components of unit current calculation before and after mesh refinement, thus obtaining the finite element simulation model.

[0021] In practice, receiving T1-weighted and T2-weighted MRI images of the target patient's head and registering them to the same spatial coordinate system can be achieved as follows: receiving T1-weighted and T2-weighted MRI images acquired by the imaging equipment according to a predetermined examination procedure; verifying the coverage, voxel size, and motion artifacts of the two types of images; and then unifying the image coordinates and performing resampling and spatial registration. Taking 3T MRI equipment as an example, three-dimensional T1-weighted and T2-weighted sequences covering the entire brain can be used; the scanning parameters are determined by the predetermined imaging examination procedure and are not automatically set by the electrical stimulation control device. The registered images are saved in a medical digital imaging and communication format or a volume data format converted from that format.

[0022] Furthermore, tissue segmentation of the registered images to obtain a three-dimensional head model including at least the scalp, skull, cerebrospinal fluid, gray matter, and white matter can be achieved as follows: Voxel regions of the scalp, skull, cerebrospinal fluid, gray matter, and white matter are marked in the registered T1-weighted and T2-weighted images; lesion regions that are identifiable in the images and alter conductive boundaries are individually labeled. The junctions between the skull and cerebrospinal fluid, and between the cerebrospinal fluid and gray matter are manually verified to correct for holes, tissue overlaps, and discontinuous boundaries. Then, closed three-dimensional tissue regions are generated based on the tissue labels to obtain the three-dimensional head model.

[0023] In addition, in specific implementation, configuring low-frequency conductivity parameters for each tissue region in the three-dimensional head model and recording the source of these conductivity parameters can be achieved in the following way: configuring conductivity suitable for a low-frequency quasi-static current field for the scalp, skull, cerebrospinal fluid, gray matter, and white matter respectively, and recording the values, units, literature sources, and corresponding tissues. As an example, the conductivity of the scalp, skull, cerebrospinal fluid, white matter, and gray matter can be set to 0.465 S / m, 0.010 S / m, 1.650 S / m, 0.126 S / m, and 0.276 S / m respectively; these values ​​are modeling parameters and are not expressed as measured conductivity of patient tissues.

[0024] In addition, in specific implementation, the equivalent contact area of ​​the return electrode and the equivalent contact area of ​​the candidate electrode on the model boundary corresponding to the outer surface of the scalp can be defined in the following way: The coordinates of the preset electrode positioning system are registered to the three-dimensional head model. Multiple non-overlapping equivalent contact areas are defined on the model boundary corresponding to the outer surface of the scalp according to the contact contour of the electrode to be used, and the number, center position, and contact area of ​​each area are recorded. One area is designated as the equivalent contact area of ​​the return electrode, and the remaining areas are designated as the equivalent contact areas of the candidate electrodes. The equivalent contact area is the area where the finite element boundary conditions apply and does not represent the installation of a physical electrode on the virtual model; the position of the physical electrode used during device output is determined by the area number in the final executable current configuration.

[0025] The 3D head model, after completing the conductivity configuration and defining the equivalent contact area of ​​the electrodes, is meshed. Mesh convergence is verified based on the difference in intracranial electric field components per unit current before and after mesh refinement. The finite element simulation model can be implemented as follows: Tetrahedral meshes are created for the 3D head model. Local mesh refinement is performed in the thin skull layer, thin cerebrospinal fluid layer, tissue junctions, and the equivalent contact area of ​​the electrodes, with fixed numbers assigned to the intracranial mesh nodes. An equal and opposite non-zero test current is applied to a candidate equivalent contact area of ​​the electrodes and the equivalent contact area of ​​the return electrodes. After completing one electric field solution, the element size of the local area is further reduced, and the relative difference in the intracranial electric field component vectors before and after refinement is calculated. If the relative difference does not exceed a preset mesh convergence limit, the current mesh is used; otherwise, refinement continues. This yields a finite element simulation model containing tissue regions, tissue conductivity, equivalent contact areas of the electrodes, node coordinates, and element connection relationships.

[0026] In some embodiments, a non-zero unit current is applied sequentially between the equivalent contact region of the candidate electrode and the equivalent contact region of the return electrode. The signed electric field components obtained by arranging them according to the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix. This can be achieved by the following steps: In the finite element simulation model, a low-frequency quasi-static current field is set, the outer surface of the head except for the equivalent contact area of ​​each electrode is set as an electrically insulating boundary, and a specification condition for determining the zero potential point is set. The equivalent contact region of a candidate electrode is sequentially set as a non-zero unit current injection boundary, the equivalent contact region of the return electrode is set as an equal amount of reverse current boundary, and the equivalent contact regions of the remaining candidate electrodes are set as zero normal current boundaries. The signed electric field components of each grid node in the brain in three coordinate directions are calculated separately when the non-zero unit current is applied to each candidate electrode. The column vector of electric field components corresponding to each candidate electrode is formed according to the fixed order of grid nodes and coordinate directions. The electric field component column vectors are divided by the non-zero unit current and arranged in the order of candidate electrodes to obtain the electric field transfer matrix; the product of the electric field transfer matrix and the candidate electrode current vectors is used to calculate the signed electric field components in the brain, and the return electrode current is determined by the current conservation condition.

[0027] In specific implementation, a low-frequency quasi-static current field is set in the finite element simulation model, and the outer surface of the head, except for the equivalent contact area of ​​each electrode, is set as an electrically insulating boundary. The specification conditions for determining the zero potential point can be implemented in the following way: solving the current continuity equation in each tissue region. Where σ is the tissue conductivity and φ is the electric potential; according to Calculate the electric field E. Set the outer surface of the head, excluding the equivalent contact area of ​​the electrodes, as a zero-normal current boundary. In each unit current example, the injection current and return current are equal in magnitude and opposite in direction, making the algebraic sum of all boundary currents zero. Additionally, set a reference node potential that does not affect the electric field gradient to zero to eliminate arbitrary constants in the potential solution under pure current boundary conditions, and to avoid simultaneously forcing the entire return electrode contact area to a fixed potential.

[0028] In practice, the following method can be used to sequentially set a candidate electrode equivalent contact region as a non-zero unit current injection boundary, set the return electrode equivalent contact region as an equal reverse current boundary, and set the remaining candidate electrode equivalent contact regions as zero normal current boundaries: sequentially select a candidate electrode equivalent contact region as the current injection region, apply a non-zero standard current I0 to this region, apply -I0 to the return electrode equivalent contact region, and set the normal current of the remaining candidate electrode equivalent contact regions to zero. Here, I0 is used to establish the unit current response, and can be 1mA or other non-zero values ​​allowed by the solver; I0 must not be zero, and each case only changes the current injection region, while keeping the conductivity, mesh, and other boundary conditions the same.

[0029] In specific implementation, the signed electric field components in the three coordinate directions of each grid node in the brain are calculated separately when the non-zero unit current is applied to each candidate electrode. The electric field component column vectors corresponding to each candidate electrode are formed according to a fixed order of grid nodes and coordinate directions. This can be achieved as follows: For each unit current example, the electric field of N grid nodes in the brain is calculated. For the i-th grid node, the three signed components Ex,i,Ey,i, and Ez,i are retained and arranged in a fixed order from Ex,1,Ey,1,Ez,1 to Ex,N,Ey,N,Ez,N to form a column vector of length 3N. The node electric field amplitude is calculated after the three components are linearly superimposed; the electric field amplitude is not used as a linear response quantity of the current.

[0030] It should be noted that the electric field component column vector in this application consists of three-dimensional signed electric field components obtained from unit current calculations, and its sign and coordinate direction are used for subsequent current superposition and target area direction projection.

[0031] In specific implementation, the column vector of each electric field component is divided by the non-zero unit current, and arranged according to the candidate electrode order to obtain the electric field transfer matrix. The product of the electric field transfer matrix and the candidate electrode current vector is used to calculate the signed electric field components in the brain. The return electrode current is determined by the current conservation condition and can be implemented as follows: the column vector of electric field components corresponding to each candidate electrode is divided by I0, and arranged according to the candidate electrode number to form the electric field transfer matrix L. If there are M equivalent contact areas of electrodes including the return electrode, then the dimension of L is 3N×(M-1). The currents of the remaining M-1 electrodes are arranged according to the candidate electrode number to form a vector I. The return electrode current is taken as the negative of the algebraic sum of each element in I. The electric field component vector in the brain is calculated according to E=LI. This basis representation can represent multi-electrode current configurations where the algebraic sum of all currents is zero.

[0032] It should be noted that the electric field transfer matrix L represents the linear relationship between the candidate electrode current and the signed electric field components in the brain. This linear relationship is derived from the low-frequency quasi-static governing equations when the tissue conductivity is fixed; the nodal electric field amplitude is equal to the square root of the sum of the squares of the corresponding three components, and needs to be calculated after the components are superimposed.

[0033] In step 102, the target points, target node weights, and target electric field directions of the brain region are obtained. A target direction projection matrix is ​​formed based on the electric field transfer matrix. The target direction projection matrix is ​​then reduced in order under the constraint of reconstruction error to obtain a target reduced-order matrix.

[0034] In practice, obtaining the brain region target points, target region node weights, and target electric field direction can be achieved as follows: The brain region target points and target electric field direction, determined by medical professionals in the rehabilitation plan, are received, along with the image coordinate system used by the brain region target points. The brain region target points are registered to the finite element simulation model to obtain the target region mesh node set. The target electric field direction can be taken as the cortical normal corresponding to each target region node, or as the direction vector given in the rehabilitation plan; each direction vector is normalized. Non-negative target region node weights are set according to the unit volume or cortical area corresponding to the node, and all weights are normalized so that their sum is one.

[0035] In some embodiments, reference Figure 3 The figure is an exemplary flowchart illustrating the establishment of a target area reduction matrix according to some embodiments of this application. In this application, a target area direction projection matrix is ​​formed based on the electric field transfer matrix, and the target area direction projection matrix is ​​reduced in order under reconstruction error constraints to obtain the target area reduction matrix. This can be achieved by the following steps: In step 1021, a target area direction projection matrix is ​​formed based on the target area grid node set and the target electric field direction; In step 1022, singular value decomposition is performed on the target area projection matrix; In step 1023, the truncation order is determined based on the relative reconstruction error before and after truncation; In step 1024, a target region reduction matrix is ​​established based on the truncation order, while the unreduced electric field transfer matrix is ​​retained.

[0036] In practice, the target area direction projection matrix can be formed based on the target area grid node set and the target electric field direction in the following way: Let the target area grid node set be S. For the i-th node in S, extract the rows corresponding to the x, y, and z coordinate directions from the electric field transfer matrix L; multiply the three rows by the x, y, and z components of the target electric field direction of that node, and then add them together to obtain the current response row of that node along the target electric field direction. Arrange the current response rows according to the target area node order to form the target area direction projection matrix A. This process is based on the spatial registration result of the target area and the target electric field direction, and does not select nodes based on the correlation of electric field amplitudes across different tissue boundaries.

[0037] It should be noted that each row of the target area direction projection matrix A corresponds to a target area grid node, and its product with the electrode current vector I is the signed electric field component of that node along the target electric field direction.

[0038] In practice, determining the truncation order based on the relative reconstruction error before and after truncation can be achieved in the following way: performing singular value decomposition on the target area projection matrix A to obtain... The truncation matrix is ​​obtained by selecting the first r singular values ​​and their corresponding singular vectors in descending order of singular values. ; Calculate the relative reconstruction error Select the minimum r value that ensures εr does not exceed the preset error limit; this r value is the cutoff order. As an example, the preset error limit can be set to 5%, but its specific value should be determined and recorded based on the required calculation accuracy.

[0039] In addition, in specific implementation, the target region order reduction matrix is ​​established based on the truncation order, and the unreduced electric field transfer matrix is ​​retained. This can be achieved in the following way: within the range of values ​​that satisfy the device current constraint, multiple sets of verification current vectors are selected, and AI and ArI are calculated respectively; if the relative error of the electric field in the target region direction corresponding to any verification current vector exceeds the verification limit, r is increased and the calculation is repeated until both the reconstruction error and the verification error meet the requirements. The final Ar is used as the target region order reduction matrix; when calculating ArI, the following steps are performed sequentially. , and This reduces repetitive matrix operations.

[0040] It should be noted that the target area reduction matrix Ar is only used to calculate the electric field in the target area direction; the unreduced electric field transfer matrix L is still used to calculate the electric field in the non-target area and to perform whole-brain electric field recalculation on the quantized candidate current vector.

[0041] In step 103, the mean electric field in the target area is calculated based on the target area reduction matrix, using the electrode current vector as a variable. The quantile value of the electric field in the non-target area is calculated based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target area and decrease the quantile value of the electric field in the non-target area. The candidate current vector is obtained under the constraints of current conservation, total injected current, single electrode current, electrode current density and number of output channels.

[0042] In practice, the upper limit of the total injection current, the upper limit of the current of each channel, the current step size, the number of rated output channels, and the upper limit of the compliance voltage are read from the rated parameters of the stimulation device, and the upper limit of the electrode current density is set according to the contact area of ​​each electrode. Taking the electrode current vector I as a variable, it is required that the algebraic sum of all electrode currents is zero, the sum of currents of the same polarity does not exceed the upper limit of the total injection current, the current of each electrode does not exceed the upper limit of the corresponding channel, the ratio of each electrode current to the contact area does not exceed the upper limit of the current density, and the number of non-zero current electrodes does not exceed the number of rated output channels. This yields the set of current values ​​for the optimization stage; the current step size is used in the candidate current vector quantization stage.

[0043] It should be noted that the set of permissible current values ​​for the device is determined by the device's rated parameters and current conservation conditions, and unverified fixed effective electric field values ​​are not used to replace the calculation results of the patient model.

[0044] In some embodiments, using the electrode current vector as a variable, the mean electric field in the target region direction is calculated based on the target region reduced-order matrix, and the upper quantile of the electric field in the non-target region is calculated based on the electric field transfer matrix. The following steps can be used to achieve the two optimization objectives of increasing the mean electric field in the target region direction and decreasing the upper quantile of the electric field in the non-target region: Nodes in the brain grid that do not belong to the target area grid node set are designated as non-target area grid nodes. The directional electric fields of each target area grid node calculated by the target area reduction matrix according to the target area node weights are weighted and summed to obtain the mean value of the target area directional electric field. The electric field vector amplitude of each non-target area grid node is calculated according to the electric field transfer matrix, and the preset upper quantile value of the electric field vector amplitude is calculated to obtain the upper quantile value of the non-target area electric field. Increasing the mean electric field value in the target area and decreasing the quantile value in the electric field in the non-target area are two independent optimization objectives, and the maximum value of the electric field vector amplitude of the grid nodes in the non-target area is not more than a preset limit is used as a constraint.

[0045] In specific implementation, nodes in the brain grid that do not belong to the target area grid node set are considered as non-target area grid nodes. The directional electric fields of each target area grid node calculated using the target area reduction matrix, based on the target area node weights, are weighted and summed to obtain the mean electric field value of the target area direction. The electric field vector amplitude of each non-target area grid node is calculated using the electric field transfer matrix, and a preset upper quantile value of the electric field vector amplitude is calculated. The upper quantile value of the non-target area electric field can be obtained as follows: Substituting the electrode current vector I into the target area reduction matrix Ar, the signed electric field components of each target area node along the target electric field direction are obtained, and then weighted and summed according to the target area node weights to obtain the mean electric field value of the target area direction. Substituting the same electrode current vector I into the unreduced electric field transfer matrix L, the electric field components in the three coordinate directions of each brain node are obtained. For non-target area nodes, the square root of the sum of the squares of the three components is calculated to obtain the non-target area electric field amplitude sequence, and the preset upper quantile value and maximum value of this sequence are calculated. As an example, the upper quantile value can be taken as the 95th percentile value to reduce the impact of a single grid outlier on the overall evaluation.

[0046] It should be noted that the mean electric field in the target area is used to characterize the overall level of the electric field in the target area along the target direction; the upper quantile of the electric field in the non-target area is used to characterize the overall level of the higher electric field amplitude in the non-target area; and the maximum electric field value in the non-target area is used to limit the local extrema.

[0047] In practice, increasing the mean electric field in the target area and decreasing the quantile value of the electric field in the non-target area can be implemented as two independent optimization objectives in the following way: Increasing the mean electric field in the target area is the first optimization objective, and decreasing the quantile value of the electric field in the non-target area is the second optimization objective. The two objectives are calculated separately and compared using their dominance relationship, without merging them into a single scalar objective using fixed weights before optimization. The maximum value of the electric field in the non-target area not exceeding the limit determined by the rehabilitation plan is used as a hard constraint. When scaling the objective values, upper and lower bounds determined and recorded within the set of allowable current values ​​for the same device are used. The scaling transformation is only used to calculate the congestion distance and does not change the dominance relationship between the two objectives.

[0048] In addition, in specific implementation, the constraint that the maximum value of the electric field vector amplitude of the non-target area grid nodes does not exceed a preset limit can be implemented in the following way: the candidate electrode current is treated as a real variable, and the return electrode current is determined by the algebraic sum of the remaining electrode currents. For each current vector, the total injection current, single electrode current, electrode current density, number of non-zero current electrodes, and maximum value of the non-target area electric field are checked in sequence; if any constraint is not satisfied, the constraint violation is recorded, and the current vector is retained without relaxing the device's rated limits. After optimization, the candidate current vectors are quantized according to the device current step size, and all constraints are rechecked based on the quantization results.

[0049] In some embodiments, obtaining the candidate current vector under constraints of current conservation, total injected current, single electrode current, electrode current density, and number of output channels can be achieved by the following steps: A constrained multi-objective model is established with the following constraints: the algebraic sum of the currents of each electrode is zero; the sum of the currents of the same polarity does not exceed the upper limit of the total injected current; the current of a single electrode does not exceed the corresponding upper limit; the ratio of the current of a single electrode to the corresponding contact area does not exceed the upper limit of the electrode current density; and the number of non-zero current electrodes does not exceed the number of output channels. Set the population size, maximum number of iterations, crossover parameter, and mutation parameter for the non-dominated sorting genetic algorithm, and set a fixed reference point for calculating the Pareto front hypervolume; The electrode currents of individuals in the population are encoded with real numbers, and non-dominated sorting is performed using the constraint dominance relationship. Individuals with larger crowding distances are preferentially retained in the same non-dominated layer to obtain the Pareto solution set for the current iteration. The population update is repeated until the relative hypervolume change rate of the Pareto front for a preset number of consecutive rounds does not exceed a preset convergence limit or the maximum number of iterations is reached. The electrode current vector in the final Pareto solution set is then used as the candidate current vector.

[0050] In practical implementation, the constraints are: the algebraic sum of all electrode currents must be zero; the total sum of currents of the same polarity must not exceed the upper limit of the total injected current; the single electrode current must not exceed its corresponding upper limit; the ratio of the single electrode current to its corresponding contact area must not exceed the upper limit of the electrode current density; and the number of non-zero current electrodes must not exceed the number of output channels. The constrained multi-objective model can be established as follows: the target vector is composed of the mean electric field in the target area and the quantile values ​​in the non-target area electric field; and the constraint set is composed of current conservation, total injected current, single electrode current, electrode current density, number of output channels, and the maximum limit of the non-target area electric field. The model output is a set of electrode current vectors that satisfy all constraints and are not mutually dominant on the two objectives.

[0051] Furthermore, in practical implementation, setting the population size, maximum number of iterations, crossover parameters, and mutation parameters of the non-dominated sorting genetic algorithm, and setting a fixed reference point for calculating the Pareto front hypervolume, can be achieved as follows: Set the population size and maximum number of iterations of the non-dominated sorting genetic algorithm based on the number of candidate electrodes, and set the crossover and mutation parameters. Perform a pre-run using at least two different random initial populations, comparing the target range covered by the obtained Pareto fronts; when the difference in target range exceeds a preset stability limit, increase the population size or the maximum number of iterations. Record the algorithm parameters, random seed, and fixed reference point used for hypervolume calculation in each optimization attempt.

[0052] It should be noted that the optimization algorithm parameters are used to control the calculation process, and the convergence of the algorithm cannot be assumed solely based on the number of variables or constraints.

[0053] In practice, the electrode currents of individuals in the population are encoded with real numbers, and non-dominated sorting is performed using constraint dominance relationships. Within the same non-dominated layer, individuals with larger crowding distances are prioritized. The Pareto solution set for the current iteration can be obtained as follows: An initial population with real-number encoding is generated within the current value set during the optimization phase. The two objective values ​​and constraint violation amounts for each individual are calculated. During sorting, feasible solutions are prioritized over infeasible solutions; between two infeasible solutions, the one with the smaller constraint violation amount is prioritized; between two feasible solutions, the non-dominated layer is determined according to the dominance relationship between the two objectives. The crowding distance is calculated for individuals within the same non-dominated layer, and individuals with larger crowding distances are prioritized, ensuring that the retained solutions are distributed at different positions on the Pareto front.

[0054] In practice, population updates are repeated until the relative hypervolume change rate of the Pareto front for a preset number of consecutive iterations does not exceed a preset convergence limit or the maximum number of iterations is reached. The electrode current vector in the final Pareto solution set is used as the candidate current vector. This can be achieved as follows: After each round of population updates, the hypervolume of the current Pareto front is calculated using a fixed reference point, and the rate of change of the current hypervolume relative to the previous round is calculated. Optimization terminates when the relative hypervolume change rate for a preset number of consecutive iterations does not exceed the preset convergence limit; otherwise, it terminates when the maximum number of iterations is reached. The Pareto solution satisfying all constraints at termination is used as the candidate current vector, without using conditions such as small crowding distance or high distribution density to filter out front boundary solutions.

[0055] In step 104, the candidate current vector is quantized according to the current step size of the neurological rehabilitation electrical stimulation device, the quantized intracranial electric field is recalculated using the electric field transfer matrix, the executable current configuration that satisfies the constraints is retained, channel control data is generated according to the executable current configuration, constant current output is controlled after the electrode contact impedance verification is passed, and output is stopped when the actual current or output voltage is abnormal.

[0056] In some embodiments, the candidate current vector is quantized according to the current step size of the neurological rehabilitation electrical stimulation device, the quantized intracranial electric field is recalculated using the electric field transfer matrix, and the executable current configuration that satisfies the constraints is retained. The generation of channel control data based on the executable current configuration can be achieved through the following steps: Each candidate current vector is quantized according to the current step size, and the mean electric field in the target area, the upper quantile of the electric field in the non-target area, and the maximum electric field in the non-target area are recalculated using the electric field transfer matrix. Candidate current vectors that do not meet the constraints are deleted. Select executable current configurations from the remaining candidate current vectors to generate channel control data including channel identifier, current amplitude, current polarity, rise time, sustain time, and fall time; detect the electrode contact impedance of each channel, output stimulation current after impedance verification, and perform feedback control based on actual current and output voltage.

[0057] In specific implementation, each candidate current vector is quantized according to the current step size, and the quantized mean electric field in the target area direction, the upper quantile of the electric field in the non-target area, and the maximum electric field in the non-target area are recalculated using the electric field transfer matrix. Candidate current vectors that do not meet the constraints are deleted in the following way: Quantize the candidate current vectors according to the device current step size, recalculate the return electrode current, and delete vectors that do not meet the requirements for current conservation, total injected current, single electrode current, electrode current density, or number of output channels. For the remaining vectors, the whole-brain electric field is recalculated using the unreduced electric field transfer matrix L, and the mean electric field in the target area direction, the upper quantile of the electric field in the non-target area, and the maximum electric field in the non-target area are recalculated. Vectors that do not meet the lower limit of the electric field in the target area direction or the limit of the electric field in the non-target area after recalculation are deleted. Then, executable current configurations are selected from the remaining vectors according to the preset target priority order. If there are no remaining vectors, the message "no executable configuration" is returned and automatic output is disabled.

[0058] In addition, in specific implementation, an executable current configuration is selected from the remaining candidate current vectors to generate channel control data including channel identifier, current amplitude, current polarity, rise time, sustain time, and fall time. The electrode contact impedance of each channel is detected, and the stimulation current is output after impedance verification. Feedback control based on the actual current and output voltage can be implemented as follows: Output channels are assigned according to the electrode region number, current amplitude, and current polarity in the executable current configuration, and rise time, sustain time, and fall time are generated. Before output, the device-allowed detection current is applied to each connected electrode, and the electrode contact impedance is calculated based on the detection voltage and detection current. Further checks are made to ensure that the target current and the expected output voltage corresponding to the contact impedance are below the device's compliance voltage limit and that a preset voltage margin is maintained. Startup is prohibited if the impedance or expected output voltage does not meet the requirements. During output, the actual current and output voltage of each channel are collected. When the deviation of the actual current from the target current exceeds the limit, the contact impedance exceeds the limit, the output voltage reaches the compliance voltage limit, or an open circuit is detected at the electrode, the current is reduced according to the fall time, and output is stopped.

[0059] In another aspect, in some embodiments, this application provides a neurological rehabilitation electrical stimulation device, with reference to... Figure 4 The figure is a schematic diagram of the structure of a neurological rehabilitation electrical stimulation device 400 according to some embodiments of this application. The neurological rehabilitation electrical stimulation device 400 includes: a data acquisition module 401, a processing module 402, and an execution module 403, which are described below: The acquisition module 401 in this application is mainly used to acquire a finite element simulation model that includes the electrical conductivity of the head tissue and the equivalent contact area of ​​the electrodes on the outer surface of the scalp. Non-zero unit current is applied between the equivalent contact area of ​​the candidate electrode and the equivalent contact area of ​​the return electrode in sequence. The signed electric field components are obtained according to the grid nodes and coordinate directions in the brain to obtain the electric field transfer matrix. Processing module 402 in this application is mainly used to obtain brain region target points, target region node weights and target electric field directions, form a target region direction projection matrix according to the electric field transfer matrix, and reduce the order of the target region direction projection matrix under the reconstruction error constraint to obtain a target region reduced order matrix. It should be noted that the processing module 402 in this application is also used to calculate the mean electric field in the target area direction based on the target area reduction matrix, using the electrode current vector as a variable, and to calculate the quantile value of the electric field in the non-target area based on the electric field transfer matrix, with the two optimization objectives of increasing the mean electric field in the target area direction and decreasing the quantile value of the electric field in the non-target area, and to obtain the candidate current vector under the constraints of current conservation, total injected current, single electrode current, electrode current density and number of output channels; The execution module 403 in this application is mainly used to quantize the candidate current vector according to the current step size of the neurological rehabilitation electrical stimulation device, recalculate the quantized intracranial electric field using the electric field transfer matrix, retain the executable current configuration that satisfies the constraints, generate channel control data according to the executable current configuration, control constant current output after the electrode contact impedance verification is passed, and stop output when the actual current or output voltage is abnormal.

[0060] In its implementation, the neurological rehabilitation electrostimulation device includes a multi-channel constant current output circuit, a current acquisition circuit, a voltage acquisition circuit, an electrode contact impedance detection circuit, and an electrode interface. The finite element model data can be generated by an external modeling device and then input into the neurological rehabilitation electrostimulation device. The electrode contact impedance detection circuit, connected to the electrode interface, is used to detect the electrode contact state before and during output; the multi-channel constant current output circuit, also connected to the electrode interface, is used to output current according to the control data of each channel.

[0061] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the control method of the above-described neurological rehabilitation electrical stimulation device.

[0062] In some embodiments, reference Figure 5 This figure is a schematic diagram of the structure of a computer device for implementing a control method for a neurological rehabilitation electrical stimulation device according to some embodiments of this application. The control method for the neurological rehabilitation electrical stimulation device in the above embodiments can be achieved through... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0063] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0064] The communication bus 502 can be used to transmit information between the aforementioned components.

[0065] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0066] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. The control method of the neurological rehabilitation electrical stimulation device in the above embodiments can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0067] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0068] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0069] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0070] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described control method for a neurological rehabilitation electrical stimulation device.

[0071] In summary, the control method disclosed in this application establishes an electric field transfer matrix using signed three-dimensional electric field components, and establishes a target region reduced-order matrix based on the target point in the brain, the target region node weights, and the target electric field direction. In the candidate current vector optimization, the electric field in the target region direction is calculated using the target region reduced-order matrix, while the electric field in the non-target region is calculated using the unreduced electric field transfer matrix. After the candidate current vector is quantized according to the device current step size, the whole brain electric field is recalculated using the unreduced electric field transfer matrix, and an executable current configuration is formed when the constraints are satisfied. The device outputs a constant current according to the executable current configuration and performs an abnormal stop based on the actual current, output voltage, and electrode contact impedance.

[0072] Therefore, the signed electric field components conform to the linear superposition condition of low-frequency quasi-static electric fields, and thus the electric field transfer matrix can be used to recalculate the intracranial electric field corresponding to different electrode current vectors. The order reduction of the target area projection matrix is ​​limited by the reconstruction error, and the non-target area electric field is still calculated by the unreduced matrix, which can retain the spatial information required for non-target area electric field evaluation while reducing the amount of repetitive calculation in the target area. Incorporating the current step size, current density, and number of output channels into the constraints, and recalculating the electric field for the quantized candidate current vectors, can reduce the deviation between the continuous optimization results and the actual set values ​​of the device. Monitoring the actual current, output voltage, and electrode contact impedance is used to identify abnormal constant current output. Therefore, this application can improve the consistency between the model calculation results and the device's executable current configuration within the applicable range of the finite element model and the device's rated parameters. It should be noted that the electric field index is a physical quantity predicted by the model and is not directly equivalent to the rehabilitation efficacy.

[0073] Although exemplary embodiments of this application have been described, those skilled in the art can make changes and modifications without departing from the technical concept of this application.

Claims

1. A control method for a neurological rehabilitation electrical stimulation device, characterized in that, include: A finite element simulation model containing the electrical conductivity of head tissue and the equivalent contact area of ​​electrodes on the outer surface of the scalp is obtained. Non-zero unit currents are applied sequentially between the equivalent contact areas of candidate electrodes and the equivalent contact areas of return electrodes. The signed electric field components obtained by arranging the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix. Obtain the target points, target node weights and target electric field directions in the brain region. Form a target region direction projection matrix based on the electric field transfer matrix. Reduce the order of the target region direction projection matrix under reconstruction error constraints to obtain the target region reduced order matrix. Using the electrode current vector as a variable, the mean electric field in the target area direction is calculated based on the target area reduction matrix, and the quantile value of the electric field in the non-target area is calculated based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target area direction and decrease the quantile value of the electric field in the non-target area. Under the constraints of current conservation, total injection current, single electrode current, electrode current density and number of output channels, candidate current vectors are obtained. The candidate current vector is quantized according to the current step size of the neurological rehabilitation electrical stimulation device. The quantized intracranial electric field is recalculated using the electric field transfer matrix. The executable current configuration that satisfies the constraints is retained. Channel control data is generated according to the executable current configuration. After the electrode contact impedance verification is passed, the constant current output is controlled and the output is stopped when the actual current or output voltage is abnormal.

2. The method as described in claim 1, characterized in that, Obtaining a finite element simulation model that includes the electrical conductivity of head tissue and the equivalent contact area of ​​electrodes on the outer surface of the scalp specifically includes: Receive T1-weighted and T2-weighted magnetic resonance images of the head of the target patient and register the two types of images to the same spatial coordinate system; The registered images are segmented to obtain a three-dimensional head model that includes at least the scalp, skull, cerebrospinal fluid, gray matter, and white matter. Configure low-frequency conductivity parameters for each tissue region in the three-dimensional head model, and record the source of the conductivity parameters; Based on the preset electrode positioning coordinates, an equivalent contact area for a return electrode and multiple equivalent contact areas for candidate electrodes are defined on the corresponding boundary of the outer scalp surface of the three-dimensional head model. The three-dimensional head model with completed conductivity configuration and electrode equivalent contact area delineation is meshed, and mesh convergence verification is performed based on the difference in intracerebral electric field components per unit current before and after mesh refinement, thus obtaining the finite element simulation model.

3. The method as described in claim 1, characterized in that, A non-zero unit current is applied sequentially between the equivalent contact region of the candidate electrode and the equivalent contact region of the return electrode. The signed electric field components, obtained according to the grid nodes and coordinate directions in the brain, yield the electric field transfer matrix, which specifically includes: In the finite element simulation model, a low-frequency quasi-static current field is set, the outer surface of the head except for the equivalent contact area of ​​each electrode is set as an electrically insulating boundary, and a specification condition for determining the zero potential point is set. The equivalent contact region of a candidate electrode is sequentially set as a non-zero unit current injection boundary, the equivalent contact region of the return electrode is set as an equal amount of reverse current boundary, and the equivalent contact regions of the remaining candidate electrodes are set as zero normal current boundaries. The signed electric field components of each grid node in the brain in three coordinate directions are calculated separately when the non-zero unit current is applied to each candidate electrode. The column vector of electric field components corresponding to each candidate electrode is formed according to the fixed order of grid nodes and coordinate directions. The electric field component column vectors are divided by the non-zero unit current and arranged in the order of candidate electrodes to obtain the electric field transfer matrix; the product of the electric field transfer matrix and the candidate electrode current vectors is used to calculate the signed electric field components in the brain, and the return electrode current is determined by the current conservation condition.

4. The method as described in claim 1, characterized in that, The target area direction projection matrix is ​​formed based on the electric field transfer matrix, and the target area direction projection matrix is ​​reduced in order under the reconstruction error constraint to obtain the target area reduced order matrix, which specifically includes: The brain region target points are registered to the finite element simulation model to obtain the target region mesh node set; according to the target electric field direction of each target region mesh node, the three coordinate direction rows corresponding to the electric field transfer matrix are projected to form the target region direction projection matrix. Singular value decomposition is performed on the target area projection matrix; The truncation order is determined based on the relative reconstruction error between the target area direction projection matrices before and after truncation, and the truncation order is increased when the relative reconstruction error exceeds a preset error limit. The target region reduction matrix is ​​constructed using the determined truncation order and the corresponding singular vectors and singular values, while the unreduced electric field transfer matrix is ​​retained.

5. The method as described in claim 1, characterized in that, Using the electrode current vector as a variable, the mean electric field in the target region is calculated based on the target region reduction matrix, and the upper quantile of the electric field in the non-target region is calculated based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target region and decrease the upper quantile of the electric field in the non-target region. Specifically, these objectives include: Nodes in the brain grid that do not belong to the target area grid node set are designated as non-target area grid nodes. The directional electric fields of each target area grid node calculated by the target area reduction matrix according to the target area node weights are weighted and summed to obtain the mean value of the target area directional electric field. The electric field vector amplitude of each non-target area grid node is calculated according to the electric field transfer matrix, and the preset upper quantile value of the electric field vector amplitude is calculated to obtain the upper quantile value of the non-target area electric field. Increasing the mean electric field value in the target area and decreasing the quantile value in the electric field in the non-target area are two independent optimization objectives, and the maximum value of the electric field vector amplitude of the grid nodes in the non-target area is not more than a preset limit is used as a constraint.

6. The method as described in claim 1, characterized in that, The candidate current vectors obtained under constraints of current conservation, total injected current, single electrode current, electrode current density, and number of output channels specifically include: A constrained multi-objective model is established with the following constraints: the algebraic sum of the currents of each electrode is zero; the sum of the currents of the same polarity does not exceed the upper limit of the total injected current; the current of a single electrode does not exceed the corresponding upper limit; the ratio of the current of a single electrode to the corresponding contact area does not exceed the upper limit of the electrode current density; and the number of non-zero current electrodes does not exceed the number of output channels. Set the population size, maximum number of iterations, crossover parameter, and mutation parameter for the non-dominated sorting genetic algorithm, and set a fixed reference point for calculating the Pareto front hypervolume; The electrode currents of individuals in the population are encoded with real numbers, and non-dominated sorting is performed using the constraint dominance relationship. Individuals with larger crowding distances are preferentially retained in the same non-dominated layer to obtain the Pareto solution set for the current iteration. The population update is repeated until the relative hypervolume change rate of the Pareto front for a preset number of consecutive rounds does not exceed a preset convergence limit or the maximum number of iterations is reached. The electrode current vector in the final Pareto solution set is then used as the candidate current vector.

7. The method as described in claim 1, characterized in that, The process of recalculating the quantized intracranial electric field using the electric field transfer matrix, retaining the executable current configuration that satisfies the constraints, and generating channel control data based on the executable current configuration includes: Each candidate current vector is quantized according to the current step size, and the mean electric field in the target area, the upper quantile of the electric field in the non-target area, and the maximum electric field in the non-target area are recalculated using the electric field transfer matrix. Candidate current vectors that do not meet the constraints are deleted. Select executable current configurations from the remaining candidate current vectors to generate channel control data including channel identifier, current amplitude, current polarity, rise time, sustain time, and fall time; detect the electrode contact impedance of each channel, output stimulation current after impedance verification, and perform feedback control based on actual current and output voltage.

8. A neurological rehabilitation electrical stimulation device, characterized in that, include: The acquisition module is used to acquire a finite element simulation model containing the electrical conductivity of head tissue and the equivalent contact area of ​​electrodes on the outer surface of the scalp. Non-zero unit currents are applied sequentially between the equivalent contact areas of candidate electrodes and the equivalent contact areas of return electrodes. The signed electric field components obtained according to the grid nodes and coordinate directions in the brain are used to obtain the electric field transfer matrix. The processing module is used to obtain the target points in the brain region, the weights of the target nodes, and the direction of the target electric field. It forms a target direction projection matrix based on the electric field transfer matrix and reduces the order of the target direction projection matrix under the constraint of reconstruction error to obtain the target reduced order matrix. The processing module is also used to calculate the mean electric field in the target area direction based on the target area reduction matrix, using the electrode current vector as a variable, and to calculate the quantile value of the electric field in the non-target area based on the electric field transfer matrix. The two optimization objectives are to increase the mean electric field in the target area direction and decrease the quantile value of the electric field in the non-target area. Under the constraints of current conservation, total injection current, single electrode current, electrode current density and number of output channels, candidate current vectors are obtained. The execution module is used to quantize the candidate current vector according to the current step size of the neurological rehabilitation electrical stimulation device, recalculate the quantized intracranial electric field using the electric field transfer matrix, retain the executable current configuration that satisfies the constraints, generate channel control data according to the executable current configuration, control constant current output after the electrode contact impedance verification is passed, and stop output when the actual current or output voltage is abnormal.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the control method of the neurological rehabilitation electrical stimulation device as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the control method of the neurological rehabilitation electrical stimulation device as described in any one of claims 1 to 7.

Images