Method and system for optimizing position of electrode plate by particle swarm optimization based on coordinate transformation

Through the particle swarm optimization algorithm of coordinate transformation and inertial weight adjustment, the problem of global search capability and stability in electrode sheet position optimization is solved, the accuracy and consistency of electrode sheet position optimization is improved, and the damage and allergies to human tissue are reduced. reaction.

CN119006581APending Publication Date: 2024-11-22NANJING UNIV OF AERONAUTICS & ASTRONAUTICS

Patent Information

Application Number
CN202410908011.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing particle swarm optimization algorithm has low global search capabilities and instability when optimizing electrode sheet positions, resulting in low accuracy and inconsistent results.

Method used

Through coordinate transformation technology, the search area of ​​the algorithm is transformed from the boundary to the inside, combined with the adaptive adjustment of inertial weights, the global search capability and stability of the particle swarm algorithm are improved, and the hybrid particle swarm optimization algorithm model and Delaunay triangulation technology are optimized Electrode sheet position.

Benefits of technology

It significantly improves the global search capability and result accuracy of the algorithm, ensures the reliability and consistency of the algorithm, and reduces the damage and allergic reactions of the electrode sheet to human tissues.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006581A_ABST
    Figure CN119006581A_ABST
Patent Text Reader

Abstract

The invention discloses a method and system for optimizing the position of an electrode plate through a particle swarm algorithm based on coordinate transformation, and the method comprises the steps: (1) segmenting an obtained CT or MRI data set, segmenting a normal tissue region and a foreign matter region, and carrying out the grid discretization of a segmentation result; (2) optimizing the positions of the known number of electrode plates attached to the object by adopting a coordinate transformation particle swarm optimization algorithm to enable the electric field intensity of the foreign matter area to reach preset intensity; according to the method, the search region of the algorithm is converted from the boundary to the interior, the characteristic of high search capability of the particle swarm algorithm in the region is fully played, the global search capability of the algorithm is improved, the solution precision is improved, and the stability of the particle swarm algorithm is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of medical equipment, and in particular to a method for optimizing electrode positions based on a particle swarm algorithm of coordinate transformation. Background Art

[0002] Since patch therapy requires long-term attachment of electrodes to human tissue, this can cause damage and allergies to human tissue, and the electrodes generate a lot of heat when working. If there are too many electrodes, the heat generated can easily cause damage to human tissue. Therefore, when choosing to attach electrodes, it is necessary to minimize the number of electrodes used and optimize their attachment positions, so as to meet the needs of electric field therapy while alleviating adverse reactions such as damage to human tissue and allergies caused to patients.

[0003] In recent years, the particle swarm optimization algorithm in the intelligent algorithm has been introduced to optimize the position of the electrode sheet, but the existing particle swarm optimization algorithm for optimizing the position of the electrode sheet still has the following defects:

[0004] 1. Low global search capability: The traditional particle swarm algorithm is affected by the fact that the solution space is at the boundary of the region rather than inside the region, and its global search capability is limited, which often leads to the algorithm converging quickly and the result accuracy is not high.

[0005] 2. Instability: As an intelligent algorithm, the traditional particle swarm algorithm has a certain degree of randomness. On the one hand, this randomness can help overcome the local optimal solution, but on the other hand, it also makes the algorithm result unstable. This instability needs to be overcome as much as possible to ensure the reliability and consistency of the algorithm. Summary of the invention

[0006] Purpose of the invention: In view of the above shortcomings, the present invention aims to transform the search area of ​​the algorithm from the boundary to the interior, give full play to the strong search ability of the particle swarm algorithm within the region, enhance the global search ability of the algorithm, and improve the accuracy of the solution; at the same time, it can improve the stability of the algorithm.

[0007] Technical solution: To solve the above problems, the present invention discloses a method for optimizing the position of electrode sheets based on a particle swarm algorithm of coordinate transformation, comprising the following steps:

[0008] (1) Segment the acquired CT or MRI data set to separate normal tissue areas and foreign matter areas, and discretize the segmentation results into grids;

[0009] (2) The electrode sheets attached to the object are taken as a particle swarm, and the coordinate transformation particle swarm optimization algorithm is used to optimize the positions of a known number of electrode sheets attached to the object so that the electric field strength in the heterogeneous area reaches a preset strength; the electrode sheets attached to the object are used to release current to the object at the attachment position, and the object senses and generates a corresponding electric field.

[0010] Furthermore, step (2) includes the following sub-steps:

[0011] (2.1) Assume that in a D-dimensional search space, for a population X with N particles, X = [X1, X2, ..., X N ], a hybrid particle swarm optimization algorithm model is established, the specific model is:

[0012]

[0013] Where V id (t+1) represents the velocity of the i-th particle at the next moment; V id (t) represents the velocity of the i-th particle at the current moment; w is the inertia factor; P id (t) represents the optimal position of the i-th particle at the current moment; P gd (t) represents the global optimal position at the current moment; c1 represents the individual learning factor of each particle; c2 represents the social learning factor of each particle; r1 and r2 are both random numbers between [0,1]; X id (t) represents the position of the i-th particle at the current moment;

[0014] (2.2) Initialize the particle swarm parameters and randomly generate the speed and position of N initial particles. The particle swarm parameters specifically include: the lower and upper bounds of the initial particle speed are set between [-π / 2, π / 2], the initial speed v i Randomly generate between [-π / 2,π / 2]; the number of pairs of electrode sheets is set to M, and an initial particle is M pairs of electrode sheets; wherein, the coordinates of the two center positions of a pair of electrode sheets are set to (x, y, z) and (x′, y′, z′), respectively, the current applied to the electrode sheet at the position of (x, y, z) is I, and the current applied to the electrode sheet at the position of (x′, y′, z′) is -I;

[0015] (2.3); Perform coordinate transformation on the particle position, and the two center position coordinates (x, y, z) and (x′, y′, z′) are converted from Cartesian coordinates to spherical coordinates through spherical coordinate transformation;

[0016] (2.4) Calculate the particle fitness function F for each of the N initial particles. The calculation formula is:

[0017] F=E tumor

[0018] In the formula, E tumor is the average electric field strength value in the heterogeneous region;

[0019] (2.5) Adaptively adjust the particle's inertia weight w, update the particle's velocity and position, and recalculate the particle's fitness; the inertia weight w changes in segments with the number of iterations k, and the adjustment formula is:

[0020]

[0021] In the formula, w min is the minimum value of the inertia weight, w max is the maximum value of the inertia weight; k is the current iteration number, k1 is the iteration number of the first stage of iteration, and k2 is the iteration number of the second stage of iteration;

[0022] (2.6) According to the calculation results of step (2.5), determine whether the particles are aggregated, that is, whether the particle search process is completed. If the particles are aggregated, the bundle algorithm is used to operate the output particles; otherwise, return to step (2.5) and repeat the iteration.

[0023] Furthermore, in step (2.4), the coordinates after coordinate transformation are assumed to be and Considering that the electrode positions are all at the boundary, only the last two variables are needed to determine the point, so the coordinates are and The spherical coordinate transformation formula is: Correspondingly, in 2D, the coordinate transformation is polar coordinate transformation. The particle positions before and after the transformation are (x, y), (x′, y′) and (r, θ), (r′, θ′). At this time, only the latter variable is needed in the coordinates to determine the point, so the coordinates are (θ) and (θ′); the polar coordinate transformation formula is: θ=arctan2(y,x).

[0024] Furthermore, in step (2.4), calculate E tumor Specifically, the head electric field intensity distribution equation is based on the quasi-static field described by the Maxwell differential equations. The Maxwell equations are:

[0025]

[0026] When a current with an angular frequency of ω is injected into the scalp electrode, the electric field intensity E in the field is expressed as:

[0027] E(x,t)=Re(e jωt E c (x)

[0028] The constitutive relation of Maxwell's equations and the definition of the electric field strength are:

[0029]

[0030] Combining the above three formulas, we get:

[0031]

[0032] Taking the divergence of both ends of the above equation, we get

[0033]

[0034] When the capacitance effect is neglected, At this time, the imaginary part is ignored, and we can get:

[0035]

[0036] In order to uniquely determine the distribution of electric potential, appropriate boundary conditions are set on the boundary. On the boundary where the electrode is placed, denoted as Γ1, assuming that the electric potential is known, denoted as g, set the Dirichlet boundary condition:

[0037] φ=g,on Γ1.

[0038] On the remaining boundary, denoted as Γ2, there is no current, which means J·υ=0. The following Neumann boundary conditions are obtained on Γ2:

[0039] On Γ2.

[0040] In summary, we can get the forward BVP model:

[0041]

[0042] In the formula, B represents magnetic induction intensity, J represents current density, E represents electric field intensity, ρ represents charge density; ε represents dielectric constant of dielectric, μ represents magnetic permeability, σ represents electrical conductivity; E(x, t) is the space-time variation field, E c (x) is a spatially varying complex-valued field, ω represents the angular frequency of the injected current in the field; φ is the potential distribution function in the field;

[0043] The electric potential distribution φ in the entire grid scene is calculated according to the forward BVP model, and then the electric field intensity in the entire grid scene is calculated by the following formula:

[0044] In Ω.

[0045] Then, the average electric field strength value E in the foreign matter area is obtained based on the head gridding result. tumor .

[0046] Furthermore, in step (1), the segmentation result is discretized into a grid using Delaunay triangulation, specifically:

[0047] In two dimensions, the overall region boundary is x 2 +y 2 =1, the inner small area boundary is (x-0.2) 2 +(y-0.2) 2 =0.1 2 , in three dimensions, the overall region boundary is x 2 +y 2 +z 2 =0.06 2 , the inner small area boundary is x 2 +(y-0.04) 2 +z 2 =0.01 2 .

[0048] Delaunay triangulation is to triangulate the point set P in the plane so that no point in P is strictly inside the circumcircle of any triangle. Specifically, the Delaunay edge is first defined: Assume that the edge set {E∈R 2 (v i ,v j )}, the two endpoints of an edge e are v i 、v j , circle C is v i and v j If the connected line segments are the diameter of a circle and the circle does not contain other points in the point set, then the edge e is a Delaunay edge. Then the definition of the Delaunay triangulation from the Delaunay edge is: when there is a triangulation T for the point set V on the plane, and the edge set E contained in the triangulation T satisfies that it only contains Delaunay edges, then the triangulation T is the Delaunay triangulation of the point set V.

[0049] Furthermore, in step (2), for the two-dimensional case, M=3 is taken, and the fitness results are calculated several times by the original particle swarm algorithm and the coordinate transformation particle swarm algorithm respectively; for the three-dimensional case, M=2 is taken, and the fitness results are calculated several times by the original particle swarm algorithm and the coordinate transformation particle swarm algorithm respectively under the condition of giving prior information; the prior information includes fixing the position of the positive electrode sheet, fixing the position of the negative electrode sheet, or fixing the position of one of the positive and negative electrode sheets.

[0050] Furthermore, this deep learning-based high-resolution reconstruction method of electrical impedance imaging is used to optimize the attachment position of the electrode patch on the object, and is not used as a diagnostic criterion and is not intended for diagnostic or therapeutic purposes.

[0051] In addition, the present invention also provides a system for optimizing the position of an electrode sheet using a particle swarm algorithm based on coordinate transformation, comprising a data processing module and an electrode sheet position optimization module:

[0052] The data processing module is used to segment the acquired CT or MRI data set, segment the normal tissue area and the foreign body area, and perform grid discretization on the segmentation result;

[0053] The electrode sheet position optimization module is used to take the electrode sheet attached to the object as a whole as a particle swarm, and use the coordinate transformation particle swarm optimization algorithm to optimize the position of a known number of electrode sheets attached to the object so that the electric field strength in the heterogeneous area reaches a preset strength; the electrode sheet attached to the object is used to release current to the object at the attachment position, and the object senses and generates a corresponding electric field.

[0054] The electrode position optimization module specifically includes:

[0055] The hybrid particle swarm optimization algorithm model building module is used to find the optimal particle swarm optimization algorithm for a population X=[X1,X2,……,X N ], a hybrid particle swarm optimization algorithm model is established, the specific model is:

[0056]

[0057] Where V id (t+1) represents the velocity of the i-th particle at the next moment; V id (t) represents the velocity of the i-th particle at the current moment; w is the inertia factor; P id (t) represents the optimal position of the i-th particle at the current moment; P gd (t) represents the global optimal position at the current moment; c1 represents the individual learning factor of each particle; c2 represents the social learning factor of each particle; r1 and r2 are both random numbers between [0,1]; X id (t) represents the position of the i-th particle at the current moment;

[0058] Initialize the hybrid particle swarm optimization algorithm model to initialize the particle swarm parameters and randomly generate the speed and position of N initial particles. The particle swarm parameters specifically include: the lower and upper bounds of the initial particle speed are set between [-π / 2,π / 2], the initial speed v i Randomly generate between [-π / 2,π / 2]; the number of pairs of electrode sheets is set to M, and an initial particle is M pairs of electrode sheets; wherein, the coordinates of the two center positions of a pair of electrode sheets are set to (x, y, z) and (x′, y′, z′), respectively, the current applied to the electrode sheet at the position of (x, y, z) is I, and the current applied to the electrode sheet at the position of (x′, y′, z′) is -I;

[0059] The particle position coordinate transformation module is used to transform the particle position coordinates. The two center position coordinates (x, y, z) and (x′, y′, z′) are transformed from Cartesian coordinates to spherical coordinates through spherical coordinate transformation.

[0060] The particle fitness calculation module is used to calculate the particle fitness function F for each of the N initial particles. The calculation formula is:

[0061] F=E tumo1

[0062] In the formula, E tumor is the average electric field strength value in the heterogeneous region;

[0063] The particle inertia weight and position update module is used to adaptively adjust the particle's inertia weight w, update the particle's velocity and position, and recalculate the particle's fitness. The inertia weight w changes in segments with the number of iterations k, and the adjustment formula is:

[0064]

[0065] In the formula, w min is the minimum value of the inertia weight, w max is the maximum value of the inertia weight; k is the current iteration number, k1 is the iteration number of the first stage of iteration, and k2 is the iteration number of the second stage of iteration;

[0066] The particle fitness judgment module is used to judge whether the particles are aggregated, that is, whether the particle search process is finished. If the particles are aggregated, the bundle algorithm is used to operate the output particles; otherwise, the particle inertia weight and position update module are returned to repeat the iteration.

[0067] In the particle position coordinate transformation module, the coordinates after coordinate transformation are set to and Considering that the electrode positions are all at the boundary, only the last two variables are needed to determine the point, so the coordinates are and The spherical coordinate transformation formula is: Correspondingly, in 2D, the coordinate transformation is polar coordinate transformation. The particle positions before and after the transformation are (x, y), (x′, y′) and (r, θ), (r′, θ′). At this time, only the latter variable is needed in the coordinates to determine the point, so the coordinates are (θ) and (θ′); the polar coordinate transformation formula is: θ=arctan2(y,x).

[0068] Beneficial effects: 1. High global search capability: In the actual application environment where the electrode sheet generates an electric field, in order to improve the global search capability of the particle swarm algorithm, the present invention introduces the coordinate transformation technology. This technology transforms the solution space of the algorithm from the region boundary to the interior of the region, effectively utilizing the strong search capability of the particle swarm algorithm within the region. This improvement not only improves the global search capability of the algorithm, but also significantly improves the accuracy of the algorithm results.

[0069] 2. High algorithm stability: This invention uses coordinate transformation technology and combines it with the segmented weight method to significantly improve the search capability of the improved particle swarm algorithm without increasing excessive computational costs. This improvement enables the algorithm to stably converge to the optimal solution, ensuring the reliability and consistency of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 Schematic diagram of the dual-branch collaborative fusion strategy network construction method of the present invention

[0071] Figure 2 It is a schematic diagram of regional grid division in a two-dimensional situation of the present invention.

[0072] Figure 3 It is a schematic diagram of the regional grid division in the three-dimensional situation of the present invention.

[0073] Figure 4 It is a schematic diagram of the optimization results of the two-dimensional situation of the present invention.

[0074] Figure 4 (a) is the regional potential distribution diagram of the optimization result of the two-dimensional case of the present invention.

[0075] Figure 4 (b) is the regional electric field intensity distribution diagram of the two-dimensional optimization result of the present invention.

[0076] Figure 5 It is a schematic diagram of the three-dimensional optimization result of the present invention.

[0077] Figure 5 (a) is the regional potential distribution diagram of the three-dimensional optimization result of the present invention.

[0078] Figure 5 (b) is a regional electric field intensity distribution diagram of the three-dimensional optimization result of the present invention. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following embodiments and appended Figures 1 to 5 , the present invention is further described in detail. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the present invention. The specific embodiment process of the present invention is as follows Figure 1 As shown, the steps are as follows:

[0080] Step 1: Give the boundaries of the overall area and the internal small area through a function expression, and use Delaunay triangulation to discretize the segmentation results. Specifically:

[0081] In two dimensions, the overall region boundary is x 2 +y 2 =1, the inner small area boundary is (x-0.2) 2 +(y-0.2) 2 =0.1 2 In three dimensions, the overall region boundary is x 2 +y 2 +z 2 =0.06 2 , the inner small area boundary is x 2 +(y-0.04) 2 +z 2 =0.01 2 .

[0082] Delaunay triangulation is to triangulate the point set P in the plane so that no point in P is strictly inside the circumcircle of any triangle. Specifically, first define the Delaunay edge: Assume that the edge set {E∈R 2 (v i ,v j )}, the two endpoints of an edge e are v i 、v j , circle C is v i and v j If the connected line segments are the diameter of a circle, and the circle does not contain other points in the point set, then the edge e is a Delaunay edge. From the Delaunay edge to the definition of Delaunay triangulation: when there is a triangulation T for the point set V on the plane, and the edge set E contained in the triangulation T satisfies that it only contains Delaunay edges, then the triangulation T is the Delaunay triangulation of the point set V. Compared with general triangulation, Delaunay triangulation can ensure that the result of the discrete data is unique. The final segmentation results are respectively meshed and discretized. The results are as follows Figure 2-3 shown.

[0083] Step 2: The electrode sheet attached to the head is taken as a particle swarm, and the coordinate transformation particle swarm optimization algorithm is used to optimize the positions of a known number of electrode sheets attached to the object so that the electric field strength in the heterogeneous area is as large as possible. Specifically, it includes:

[0084] (1) Establish a coordinate transformation particle swarm optimization algorithm model. The coordinate transformation particle swarm optimization algorithm obtains the global optimal solution based on the movement of particles in the search space; suppose that in a D-dimensional search space, the population consists of N particles X = [X1, X2, ..., X N ], the i-th particle exchanges information between populations according to the historical optimal solution and the global optimal solution in each iterative evolution process to update the position and speed of the particle. Its mathematical model is:

[0085]

[0086] Where V id (t+1) represents the velocity of the i-th particle at the next moment; V id (t) represents the speed of the i-th particle at the current moment; w is the inertia factor. When w is large, the particle swarm has a strong global optimization ability and the particle convergence speed is slow. On the contrary, the particle swarm has a strong local optimization ability and the particle convergence speed is fast. P id (t) represents the optimal position of the i-th particle at the current moment; P gd (t) represents the global optimal position at the current moment; c1 represents the individual learning factor of each particle; c2 represents the social learning factor of each particle; r1 and r2 are both random numbers between [0,1]; X id (t) represents the position of the i-th particle at the current moment.

[0087] (2) Initialize the particle swarm parameters and randomly generate the speed and position of N initial particles. The particle swarm parameters include: the number of particles N is set to 10; the lower and upper bounds of the speed are set between [-π / 2, π / 2], and the initial speed v i Generate a random initial velocity in the upper and lower bounds; c1 and c2 are selected to be 2; r1 and r2 are selected to be random numbers between [0,1]; the initial value and final value of the inertia weight w are selected to be 0.9 and 0.4 respectively; the number of pairs of electrode sheets is set to M, one initial particle is M pairs of electrode sheets, and for one initial particle, the center position range of the electrode sheet is set to the range determined on the scene surface grid in combination with prior information; wherein, the coordinates of the center position of the electrode sheet are set to (x, y, z) and (x′, y′, z′), respectively; the current applied to the electrode sheet at the position of (x, y, z) is 900 mA, and the current applied to the electrode sheet at the position of (x′, y′, z′) is -900 mA.

[0088] (3) Perform coordinate transformation on the particle position. Assume that the original coordinates of the center position of the electrode sheet are (x, y, z) and (x′, y′, z′), and convert the Cartesian coordinates into spherical coordinates through spherical coordinate transformation. Assume that the coordinates after coordinate transformation are and where r, θ, are radial distance, polar angle and azimuth respectively. Considering that the electrode positions are all at the boundary, only the last two variables are needed to determine the point, so the coordinates are and The spherical coordinate transformation formula is: Correspondingly, in 2D, the coordinate transformation is polar coordinate transformation. The particle positions before and after the transformation are (x, y), (x′, y′) and (r, θ), (r′, θ′), where r and θ are radial distance and azimuth respectively. At this time, only the latter variable is needed to determine the point, and the coordinates are (θ) and (θ′). The polar coordinate transformation formula is: θ=arctan2(y,x).

[0089] (4) Calculate the particle fitness function F for each of the 10 initial particles, and obtain the optimal position of the individual particle and the global optimal position. The specific calculation formula of the fitness function is:

[0090] F=E tumor (2)

[0091] In the formula, E tumor is the average electric field strength value in the heterogeneous region;

[0092] The electric potential φ can be obtained by solving the spatial electric potential distribution equation using the finite element simulation method. Specifically, the head electric field intensity distribution equation is based on the quasi-static field described by the Maxwell differential equations, as shown in formula (3):

[0093]

[0094] When a current with an angular frequency of ω is injected into the scalp electrode, the electric field strength E in the field can be expressed as:

[0095] E(x,t)=Re(e jωt E c (x)) (4)

[0096] The constitutive relation of Maxwell's equations and the definition of the electric field strength are:

[0097]

[0098] Combining (3)-(5) we can get:

[0099]

[0100] Taking the divergence on both ends of the above equation, we can get

[0101]

[0102] When the capacitance effect can be neglected, At this time, the imaginary part can be ignored, and we can get:

[0103]

[0104] In order to uniquely determine the distribution of electric potential, it is necessary to set appropriate boundary conditions on the boundary. On the boundary where the electrode is placed, denoted as Γ1, assuming that the electric potential is known, denoted as g, the Dirichlet boundary condition can be set:

[0105] φ=g, on Γ1. (9)

[0106] On the remaining boundary, denoted as Γ2, there is no current, which means J·υ=0. The following Neumann boundary conditions can be obtained on Γ2:

[0107]

[0108] In summary, by combining equations (8)-(10), we can get the forward BVP (Boundary Value Problem) model:

[0109]

[0110] In the formula, B represents magnetic induction intensity, J represents current density, E represents electric field intensity, ρ represents charge density; ε represents dielectric constant of dielectric, μ represents magnetic permeability, σ represents electrical conductivity; E(x, t) is the space-time variation field, E c (x) is a spatially varying complex-valued field, ω represents the angular frequency of the injected current in the field, and φ is the potential distribution function in the field.

[0111] According to model (11), the electric potential distribution φ in the entire grid scene is calculated, and the electric field intensity E in the entire grid scene can be calculated by combining the following formula:

[0112] In Ω.

[0113] Then, according to the head meshing result, the mean electric field strength E on the grid nodes of the foreign body area (brain glioma area) is obtained. tumor .

[0114] (5) Adaptively adjust the particle's inertia weight w, update the particle's velocity and position, and recalculate the particle's fitness. The inertia weight w changes in segments with the number of iterations k, and the adjustment formula is:

[0115]

[0116] In the formula, wmin is the minimum value of the inertia weight, w max is the maximum value of the inertia weight; k is the current iteration number, k1 is the iteration number of the first stage of iteration, and k2 is the iteration number of the second stage of iteration.

[0117] Specifically, since the electrode sheet is attached to the scalp, P in the calculation formula (1) id (t)-X id (t) and P gd (t)-X id (t) are all geodesics, i.e., the surface distance between two points on the head surface, through w×v i The particle velocity is updated, the position of the particle on the scalp surface is calculated for the next iteration, and the average field strength in the glioma area is recalculated.

[0118] (6) According to the calculation result of step (5), determine whether the particles are aggregated, that is, whether the particle search process is completed. If the particles are aggregated, the algorithm is terminated and the output particles are operated; otherwise, return to step (5) and repeat the iteration.

[0119] Finally, for the two-dimensional case, M = 3, k1 = 200, k2 = 50, and the fitness results obtained by calculating 8 times by the original particle swarm algorithm and the coordinate transformation particle swarm algorithm are shown in Table 1:

[0120] Table 1. Fitness results of original and coordinate transformed particle swarm algorithm in two dimensions Unit: V / cm

[0121]

[0122] The average fitness result of the original particle swarm algorithm is 196.3883, and the average fitness result of the coordinate transformation particle swarm algorithm is 198.5963, an increase of 2.208%.

[0123] For the three-dimensional case, M=2, k1=200, k2=50, given certain prior information, such as fixing the position of the positive electrode sheet, fixing the position of the negative electrode sheet, or fixing the position of one of the positive and negative electrode sheets, the fitness results obtained by calculating 8 times by the original particle swarm algorithm and the coordinate transformation particle swarm algorithm are shown in Tables 2 to 4:

[0124] Table 2. The fitness results of the original and coordinate transformed particle swarm algorithm when the positive electrode position is fixed in three dimensions. Unit: V / cm

[0125]

[0126] The average fitness result of the original particle swarm algorithm is 151.7001, and the average fitness result of the coordinate transformation particle swarm algorithm is 155.2819, an increase of 2.307%.

[0127] Table 3. Fitness results of original and coordinate transformed particle swarm algorithm with fixed negative electrode sheet position in three dimensions. Unit: V / cm

[0128]

[0129] The average fitness result of the original particle swarm algorithm is 154.6540, and the average fitness result of the coordinate transformation particle swarm algorithm is 155.2819, an increase of 0.404%.

[0130] Table 4. The fitness results of the original and coordinate transformed particle swarm algorithm in the case of fixing the position of one positive and one negative electrode in three dimensions. Unit: V / cm

[0131]

[0132] The average fitness result of the original particle swarm algorithm is 151.7551, and the average fitness result of the coordinate transformation particle swarm algorithm is 155.2819, an increase of 2.271%.

[0133] Since the fitness indicates the quality of the electrode position, a high fitness means that the electrode position is relatively good. In the above verification results, 8 results are given for each verification, and the optimization effect is reflected by comparing the mean of the algorithm results.

[0134] Figure 4 and Figure 5 The simulation results of electrode placement obtained by using coordinate transformation particle swarm algorithm in two dimensions and three dimensions are shown respectively. The left figure shows the distribution of electric field in the region, and the right figure shows the distribution of field intensity in the region.

Claims

1. A method for optimizing the position of an electrode sheet using a particle swarm algorithm based on coordinate transformation, characterized in that: The following steps are involved: (1) Segment the acquired CT or MRI data set to separate normal tissue areas and foreign matter areas, and discretize the segmentation results into grids; (2) The electrode sheets attached to the object are taken as a particle swarm, and the coordinate transformation particle swarm optimization algorithm is used to optimize the positions of a known number of electrode sheets attached to the object so that the electric field strength in the heterogeneous area reaches a preset strength; the electrode sheets attached to the object are used to release current to the object at the attachment position, and the object senses and generates a corresponding electric field.

2. The method for optimizing the position of electrode sheets by using a particle swarm algorithm based on coordinate transformation according to claim 1, characterized in that: Step (2) includes the following sub-steps: (2.1) Assume that in a D-dimensional search space, for a population of N particles X = [X q ,X2,……,X N ], a hybrid particle swarm optimization algorithm model is established, the specific model is: Where V id (t+1) represents the velocity of the i-th particle at the next moment; V id (t) represents the velocity of the i-th particle at the current moment; w is the inertia factor; P id (t) represents the optimal position of the i-th particle at the current moment; P gd (t) represents the global optimal position at the current moment; c1 represents the individual learning factor of each particle; c2 represents the social learning factor of each particle; r1 and r2 are both random numbers between [0,1]; X id (t) represents the position of the i-th particle at the current moment; (2.2) Initialize the particle swarm parameters and randomly generate the speed and position of N initial particles. The particle swarm parameters include: the lower and upper bounds of the initial particle speed are set between [-π / 2, π / 2], the initial speed v i Randomly generate between [-π / 2,π / 2]; the number of pairs of electrode sheets is set to M, and an initial particle is M pairs of electrode sheets; wherein, the coordinates of the two center positions of a pair of electrode sheets are set to (x, y, z) and (x′, y′, z′), respectively, the current applied to the electrode sheet at the position of (x, y, z) is I, and the current applied to the electrode sheet at the position of (x′, y′, z′) is -I; (2.3) Perform coordinate transformation on the particle position. The two center position coordinates (x, y, z) and (x′, y′, z′) are converted from Cartesian coordinates to spherical coordinates through spherical coordinate transformation. (2.4) Calculate the particle fitness function F for each of the N initial particles. The calculation formula is: F=E tumor In the formula, E tumor is the average electric field strength value in the heterogeneous region; (2.5) Adaptively adjust the particle's inertia weight w, update the particle's velocity and position, and recalculate the particle's fitness; the inertia weight w changes in segments with the number of iterations k, and the adjustment formula is: In the formula, w min is the minimum value of the inertia weight, w max is the maximum value of the inertia weight; k is the current iteration number, k1 is the iteration number of the first stage of iteration, and k2 is the iteration number of the second stage of iteration; (2.6) According to the calculation results of step (2.5), determine whether the particles are aggregated, that is, whether the particle search process is over. If the particles are already aggregated, end the algorithm and operate the output particles; otherwise, return to step (2.5) and repeat the iteration.

3. The method for optimizing electrode position based on coordinate transformation particle swarm algorithm according to claim 2, characterized in that: In step (2.3), the coordinates after coordinate transformation are and Considering that the electrode positions are all at the boundary, only the last two variables are needed to determine the point, so the coordinates are and The spherical coordinate transformation formula is: Correspondingly, in two dimensions, the coordinate transformation is polar coordinate transformation. The particle positions before and after the transformation are (x, y), (x′, y′) and (r, θ), (r′, θ′). At this time, only the latter variable is needed in the coordinates to determine the point, so the coordinates are (θ) and (θ′); the polar coordinate transformation formula is: θ=arctan2(y,x).

4. The method for optimizing electrode position based on coordinate transformation particle swarm algorithm according to claim 3, characterized in that: In step (2.4), calculate E tumor Specifically, the head electric field intensity distribution equation is based on the quasi-static field described by the Maxwell differential equations. The Maxwell equations are: When a current with an angular frequency of ω is injected into the scalp electrode, the electric field intensity E in the field is expressed as: E(x,t)=Re(e jωt HAVE BEEN c (x)) The constitutive relation of Maxwell's equations and the definition of the electric field strength are: Combining the above three formulas, we get: Taking the divergence of both ends of the above equation, we get When the capacitance effect is neglected, At this time, the imaginary part is ignored, and we can get: In order to uniquely determine the distribution of electric potential, appropriate boundary conditions are set on the boundary. On the boundary where the electrode is placed, denoted as Γ1, assuming that the electric potential is known, denoted as g, set the Dirichlet boundary condition: φ=g,on Γ1. On the remaining boundary, denoted as Γ2, there is no current, which means J·υ = 0. The following Neumann boundary conditions are obtained on Γ2: On Γ2. In summary, we can get the forward BVP model: In the formula, B represents magnetic induction intensity, J represents current density, E represents electric field intensity, ρ represents charge density; ε represents dielectric constant of dielectric, μ represents magnetic permeability, σ represents electrical conductivity; E(x, t) is the space-time variation field, E c (x) is a spatially varying complex-valued field, ω represents the angular frequency of the injected current in the field; φ is the potential distribution function in the field; The electric potential distribution φ in the entire grid scene is calculated according to the forward BVP model, and then the electric field intensity in the entire grid scene is calculated by the following formula: In Ω. Then, the average electric field strength value E in the foreign matter area is obtained based on the head gridding result. tumor .

5. The high-resolution reconstruction method of electrical impedance imaging based on deep learning according to claim 1, characterized in that: In step (1), the segmentation result is discretized into a grid using Delaunay triangulation, specifically: In two dimensions, the overall region boundary is x 2 +y 2 =1, the inner small area boundary is (x-0.2) 2 +(y-0.2) 2 =0.1 2 , in three dimensions, the overall region boundary is x 2 +y 2 +z 2 =0.06 2 , the inner small area boundary is x 2 +(y-0.04) 2 +z 2 =0.01 2 . Delaunay triangulation is to triangulate the point set P in the plane so that no point in P is strictly inside the circumcircle of any triangle. Specifically, the Delaunay edge is first defined: Assume that the edge set {E∈R 2 (v i ,v j )}, the two endpoints of an edge e are v i 、v j , circle C is v i and v j If the connected line segments are the diameter of a circle and the circle does not contain other points in the point set, then the edge e is a Delaunay edge. Then the definition of the Delaunay triangulation from the Delaunay edge is: when there is a triangulation T for the point set V on the plane, and the edge set E contained in the triangulation T satisfies that it only contains Delaunay edges, then the triangulation T is the Delaunay triangulation of the point set V.

6. The high-resolution reconstruction method of electrical impedance imaging based on deep learning according to claim 2, characterized in that: In step (2), for the two-dimensional case, M=3 is taken, and the fitness results are calculated several times by the original particle swarm algorithm and the coordinate transformation particle swarm algorithm respectively; for the three-dimensional case, M=2 is taken, and the fitness results are calculated several times by the original particle swarm algorithm and the coordinate transformation particle swarm algorithm respectively under the condition of giving prior information; the prior information includes fixing the position of the positive electrode sheet, fixing the position of the negative electrode sheet, or fixing the position of one of the positive and negative electrode sheets.

7. The high-resolution reconstruction method of electrical impedance imaging based on deep learning according to any one of claims 1 to 6, characterized in that: This deep learning-based high-resolution reconstruction method of electrical impedance imaging is used to optimize the attachment position of the electrode on the object. It is not used as a diagnostic criterion and has non-diagnostic and non-therapeutic purposes.

8. A system for optimizing electrode position based on particle swarm optimization algorithm based on coordinate transformation, characterized in that: Including data processing module and electrode position optimization module: The data processing module is used to segment the acquired CT or MRI data set, segment the normal tissue area and the foreign body area, and perform grid discretization on the segmentation result; The electrode sheet position optimization module is used to take the electrode sheet attached to the object as a whole as a particle swarm, and use the coordinate transformation particle swarm optimization algorithm to optimize the position of a known number of electrode sheets attached to the object so that the electric field strength in the heterogeneous area reaches a preset strength; the electrode sheet attached to the object is used to release current to the object at the attachment position, and the object senses and generates a corresponding electric field.

9. The system for optimizing electrode position based on particle swarm algorithm based on coordinate transformation according to claim 8, characterized in that: The electrode position optimization module specifically includes: The hybrid particle swarm optimization algorithm model building module is used to find the optimal particle swarm optimization algorithm for a population X=[X1,X2,……,X N ], a hybrid particle swarm optimization algorithm model is established, the specific model is: Where V id (t+1) represents the velocity of the i-th particle at the next moment; V id (t) represents the velocity of the i-th particle at the current moment; w is the inertia factor; P id (t) represents the optimal position of the i-th particle at the current moment; P gd (t) represents the global optimal position at the current moment; c1 represents the individual learning factor of each particle; c2 represents the social learning factor of each particle; r1 and r2 are both random numbers between [0,1]; X id (t) represents the position of the i-th particle at the current moment; Initialize the hybrid particle swarm optimization algorithm model to initialize the particle swarm parameters and randomly generate the speed and position of N initial particles. The particle swarm parameters specifically include: the lower and upper bounds of the initial particle speed are set between [-π / 2,π / 2], the initial speed v i Randomly generate between [-π / 2,π / 2]; the number of pairs of electrode sheets is set to M, and an initial particle is M pairs of electrode sheets; wherein, the coordinates of the two center positions of a pair of electrode sheets are set to (x, y, z) and (x′, y′, z′), respectively, the current applied to the electrode sheet at the position of (x, y, z) is I, and the current applied to the electrode sheet at the position of (x′, y′, z′) is -I; The particle position coordinate transformation module is used to transform the particle position coordinates. The two center position coordinates (x, y, z) and (x′, y′, z′) are transformed from Cartesian coordinates to spherical coordinates through spherical coordinate transformation. The particle fitness calculation module is used to calculate the particle fitness function F for each of the N initial particles. The calculation formula is: F=E tumor In the formula, E tumor is the average electric field strength value in the heterogeneous region; The particle inertia weight and position update module is used to adaptively adjust the particle's inertia weight w, update the particle's velocity and position, and recalculate the particle's fitness. The inertia weight w changes in segments with the number of iterations k, and the adjustment formula is: In the formula, w min is the minimum value of the inertia weight, w max is the maximum value of the inertia weight; k is the current iteration number, k1 is the iteration number of the first stage of iteration, and k2 is the iteration number of the second stage of iteration; The particle fitness judgment module is used to judge whether the particles are aggregated, that is, whether the particle search process is finished. If the particles are aggregated, the bundle algorithm is used to operate the output particles; otherwise, the particle inertia weight and position update module are returned to repeat the iteration.

10. The system for optimizing electrode position based on coordinate transformation particle swarm algorithm according to claim 8, characterized in that: In the particle position coordinate transformation module, the coordinates after coordinate transformation are set to and Considering that the electrode positions are all at the boundary, only the last two variables are needed to determine the point, so the coordinates are and The spherical coordinate transformation formula is: Correspondingly, in 2D, the coordinate transformation is polar coordinate transformation. The particle positions before and after the transformation are (x, y), (x′, y′) and (r, θ), (r′, θ′). At this time, only the latter variable is needed in the coordinates to determine the point, so the coordinates are (θ) and (θ′); the polar coordinate transformation formula is: θ=arctan2(y,x).

Citation Information

Patent Citations

  • Brain power supply positioning method, system and device based on population diversity control

    CN112294339A

  • Electrode plate attaching method and system

    CN115619808A

  • Method for optimizing support vector machine on basis of particle swarm optimization algorithm

    WO2018072351A1

Cited By

  • Underwater passive electric field positioning method and system

    CN119757892A