Optimization method and system for optic nerve electrode array layout

Abstract

The application discloses an optic nerve electrode array layout optimization method and system. The method firstly constructs a static optic nerve fiber random distribution physiological model, then establishes an optic nerve and electrode model through a multi-physical field simulation software to obtain a potential distribution, initializes a covariance matrix self-adaptive evolution strategy algorithm by using a Python main control script and introduces biological safety and electrode manufacturing process constraints, realizes automatic coupling modeling through a Python interface, maps the potential distribution to an axon model to perform nerve dynamics simulation and fitness evaluation, and finally iteratively updates and outputs a globally optimal electrode spatial distribution parameter set. The application deeply combines three-dimensional finite element electric field simulation and nerve dynamics model, globally optimizes under strict constraints through an automatic algorithm, solves the contradiction between "high coverage rate" and "low damage" in optic nerve prosthesis design, and significantly improves the electric stimulation coverage rate and design efficiency of the optic nerve prosthesis.

Description

Technical Field

[0001] This invention belongs to the field of neural prosthesis technology, specifically relating to a method and system for optimizing the layout of optic nerve electrode arrays. Background Technology

[0002] Optic nerve prostheses are a neurorehabilitation technique used to help blind people regain visual function. In optic nerve prosthesis systems, the stimulation coverage and resolution of the artificial prosthesis directly determine the final quality of the patient's prosthetic recovery, and these two key indicators are closely related to the spatial location and number of implanted electrodes.

[0003] However, optic nerve tissue is extremely fragile, and electrode design is not simply a matter of "numerical stacking." An excessive number of electrodes and inappropriate entry points can easily cause irreversible mechanical damage to the axonal fibers inside the optic nerve, and due to overstimulation, it often fails to provide better visual coverage and resolution. Therefore, electrode design is essentially a complex problem that requires balancing "activation efficiency" with "reducing tissue damage."

[0004] Currently, traditional invasive electrode arrays (such as the Utah array) mostly use a fixed rectangular grid layout. This layout has a serious geometric mismatch with the natural anatomical shape of the optic nerve, which is round or elliptical. This structural mismatch not only easily causes additional damage to the optic nerve and its surrounding cortical sheath tissue during implantation, but also makes it difficult for the electrodes to effectively cover the peripheral area of ​​the optic nerve, thus limiting visual field coverage and creating a large stimulation blind spot in the peripheral visual field.

[0005] Therefore, to further improve the quality of artificial nervous systems while ensuring neurophysiological safety, it is necessary to design an optic nerve insertion electrode with a better structure and more reasonable arrangement, so as to break through the existing optic nerve prostheses designed based on experience.

[0006] To address the shortcomings of existing technologies in electrode layout design, such as the lack of scientific optimization methods and the difficulty in balancing high coverage and low damage, this invention proposes a method and system for optimizing the layout of optic nerve electrode arrays. This method and system aim to introduce multi-objective constraints into automated algorithms to scientifically guide electrode spatial layout schemes that meet both biosafety and electrode manufacturing process requirements. Summary of the Invention

[0007] The technical problem that this invention aims to solve is that existing optic nerve electrode arrays mostly adopt a fixed rectangular grid layout, which does not match the geometry of the natural anatomical shape of the optic nerve, easily causing tissue damage and low stimulation coverage. At the same time, existing electrode layout designs mostly rely on experience and lack scientific optimization methods, making it impossible to effectively balance the high activation rate of nerve fibers with the low risk of tissue damage.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A method for optimizing the layout of an optic nerve electrode array includes the following steps:

[0010] S1. Construct a static physiological model of random distribution of optic nerve fibers, generate multiple myelinated axon models with different diameters, and the axons are randomly and discretely distributed in the three-dimensional space of the optic nerve. The spatial distribution of this model remains unchanged during the subsequent optimization process.

[0011] S2, Based on multiphysics simulation software, a model of the optic nerve and electrodes is established to obtain the three-dimensional potential distribution in the optic nerve under current stimulation.

[0012] S3, Initialize the covariance matrix adaptive evolution strategy algorithm, set the optimization objective of maximizing the axon activation rate, and introduce constraints related to biosafety and electrode manufacturing process;

[0013] S4, through the interface between the main control script and the multiphysics simulation software, imports the electrode space coordinates generated by the algorithm into the model, simulates the physical occupancy after electrode implantation and applies current stimulation, solves and derives the potential distribution;

[0014] S5, map the potential distribution to the axon model, solve the axon transmembrane potential response, calculate the axon activation rate and safety index of the current scheme, and generate a fitness ranking table;

[0015] S6 updates the algorithm parameters based on fitness feedback, iterates until the convergence condition is met, and outputs the globally optimal set of electrode spatial distribution parameters.

[0016] Preferably, in step S1, 10,000 to 100,000 myelinated axon models are generated, with axon diameter between 1 micrometer and 8.7 micrometers and following a bimodal Gaussian distribution. Each axon adopts a multi-layer myelin model, which is divided into nodes of Ranvier and internode segments. The internode segments are further divided into paranode zones, near-paranode zones, and isodiameter myelin zones.

[0017] Preferably, in step S2, the electrode geometry is dynamically generated based on the electrode coordinates, the geometric difference operation is performed to remove the space occupied by the electrode, and a nerve-electrode coupling model is established; different conductivity parameters are set for the nerve bundle, nerve sheath, and surrounding body fluid, a stimulation current is applied to the electrode surface, and the electrode surface and adjacent nerve tissue are subjected to mesh refinement processing.

[0018] Preferably, the constraints introduced in step S3 include single-channel stimulation current intensity constraints, electrode number constraints, spatial linear distance constraints between any two electrode microneedle points, and double-layer radial distribution constraints of the electrode array.

[0019] Preferably, in step S3, the initial search step size of the algorithm is set to 100 micrometers, the population size is set to 8 to 10 individuals, the maximum number of iterations is set to 30 to 50 generations, and the convergence threshold is set to 10 to the power of negative 6. The double-layer electrode array is parameterized using a polar coordinate system, and the independent variables include the inner layer distributed polar diameter, the inner layer initial polar angle offset, the outer layer distributed polar diameter, and the outer layer initial polar angle offset.

[0020] A system for optimizing the layout of an optic nerve electrode array includes a physiological model construction module, a multiphysics modeling module, an algorithm initialization module, an automated coupling module, a fitness evaluation module, and a global optimization output module.

[0021] The physiological model building module is used to construct a static physiological model of randomly distributed optic nerve fibers, generating multiple myelinated axon models with different diameters;

[0022] The multiphysics modeling module is used to establish models of the optic nerve and electrodes, and to obtain the three-dimensional potential distribution within the optic nerve under electrical stimulation.

[0023] The algorithm initialization module is used to initialize the covariance matrix adaptive evolution strategy algorithm, set the optimization objective, and introduce constraints related to biosafety and electrode manufacturing process.

[0024] The automated coupling module is used to import the electrode space coordinates generated by the algorithm into the multiphysics model to simulate electrode implantation and solve for the potential distribution.

[0025] The fitness evaluation module is used to map the potential distribution to the axon model, calculate the axon activation rate and safety index, and generate a fitness ranking table.

[0026] The global optimization output module is used to update the algorithm parameters based on fitness feedback, iterates until convergence, and outputs the optimal electrode distribution parameter set.

[0027] Preferably, the physiological model construction module generates 10,000 to 100,000 myelinated axon models with axon diameter between 1 micrometer and 8.7 micrometers and following a bimodal Gaussian distribution. Each axon adopts a multi-layer myelin model, which is divided into nodes of Ranvier and internode segments. The internode segments are further divided into paranode zones, near-paranode zones, and isodiameter myelin zones.

[0028] Preferably, the multiphysics modeling module dynamically generates electrode geometry based on electrode coordinates, performs geometric difference operations to remove the space occupied by the electrodes to establish a nerve-electrode coupling model, sets corresponding conductivity parameters for different tissue sites, applies stimulation current to the electrode surface, and performs mesh refinement processing on the electrode surface and adjacent nerve tissue.

[0029] Preferably, the constraints introduced by the algorithm initialization module include single-channel stimulation current intensity constraints, electrode number constraints, spatial linear distance constraints between any two electrode microneedle points, and double-layer radial distribution constraints of the electrode array.

[0030] Preferably, the algorithm initialization module sets the initial search step size to 100 micrometers, the population size to 8 to 10 individuals, the maximum number of iterations to 30 to 50 generations, and the convergence threshold to 10 to the power of -6. The dual-layer electrode array is parameterized using polar coordinates, with independent variables including the inner layer distributed polar diameter, the inner layer initial polar angle offset, the outer layer distributed polar diameter, and the outer layer initial polar angle offset. The global optimization output module finally outputs the optimal parameter set containing the electrode polar diameter, polar angle, and penetration depth.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] The proposed optic nerve electrode array layout optimization method differs from traditional optimization methods that rely solely on geometric distribution or simplified physical fields. This method deeply integrates the three-dimensional finite element electric field simulation of multiphysics simulation (COMSOL) software with the neurodynamic model of neurodynamics software (NEURON). Through a Python interface, the precisely calculated extracellular potential distribution is directly applied to myelinated axon models with different diameters and random distributions. This coupled simulation more realistically simulates the stimulation effect of electrodes in complex tissue environments (such as nerve bundles, nerve sheaths, and body fluids), giving the optimization results extremely high biological credibility.

[0033] In addition, current electrode layout designs often lack scientific optimization methods and rely heavily on experience. This invention proposes a complete automated execution framework that uses a Python master script to drive population evolution, automatically performing iterative evaluation, crossover, and mutation. This algorithm-driven approach not only significantly improves design efficiency but also searches for globally optimal solutions within a broad anatomically feasible domain that are difficult to discover through human experience, providing scientific guidance for improving the performance of optic nerve prosthesis systems.

[0034] This invention transforms biosafety indicators such as single-channel safe current threshold, electrode spacing constraints, and microneedle insertion depth into dynamic constraint operators of an algorithm through weight mapping. While maximizing the overall activation rate of nerve fibers, it strictly controls the risk of tissue mechanical damage, fundamentally resolving the contradiction between "high coverage" and "low damage" in optic nerve prosthesis design, and providing a quantitative basis for the development of high-performance optic nerve prostheses. Attached Figure Description

[0035] Figure 1 This is a flowchart of the optic nerve electrode array layout optimization method proposed in this invention.

[0036] Figure 2 This is a schematic diagram of the electrode distribution before and after optimization using the optimization method proposed in this invention.

[0037] Figure 3 This is a schematic diagram showing the axon activation range before and after optimization using the optimization method proposed in this invention.

[0038] Figure 4 The image shows the activation state of the YZ-plane axons before optimization using the optimization method proposed in this invention.

[0039] Figure 5 The diagram shows the activation state of the YZ-plane axons after optimization using the optimization method proposed in this invention.

[0040] Figure 6 This is a comparison chart of the number of activated axons before and after optimization using the optimization method proposed in this invention.

[0041] Figure 7 This is a convergence curve of electrode coverage as the number of evaluations increases in the optimization method proposed in this invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] Example 1

[0044] A method for optimizing the layout of an optic nerve electrode array is described in further detail below with reference to the accompanying drawings and specific embodiments.

[0045] like Figure 1 The diagram shows a flowchart of the optic nerve electrode array layout optimization method of the present invention, which covers full-scale simulation from microscopic axonal dynamics to macroscopic three-dimensional physical fields. The method first constructs a model with a random distribution of multiple optic nerve fibers, providing a standardized biological evaluation benchmark for optimization. Then, by defining core constraints such as the number of electrodes, the radial range of the double layer, and the mechanical spacing, a rigorous search boundary is delineated. Finally, the adaptive characteristics of the Covariance Matrix Adaptive Evolutionary Strategy (CMA-ES) algorithm are utilized to efficiently explore within the high-dimensional independent variable space.

[0046] At the execution layer, complex geometric difference operations and electric potential field solutions were implemented through an automated interface between Python scripts and finite element software. Next, a neurodynamics calculation engine was used to transform the physical field information into biological activation indicators. Finally, through multiple generations of iterative evolution, the system automatically determined the optimal electrode arrangement of four inner electrodes and twelve outer electrodes. This process not only eliminates the randomness and bias of manual design layouts but also ensures the optimal balance between improving visual resolution and ensuring neural safety from a biophysical perspective, providing a scientific quantitative method for the development of next-generation invasive optic nerve prostheses.

[0047] The specific steps of the optic nerve electrode array layout optimization method of the present invention include:

[0048] S1, Construct a static physiological model of random distribution of optic nerve fibers:

[0049] Anatomical parameters of the optic nerve were obtained using the NEURON software. Within the three-dimensional spatial region of the optic nerve, a model of at least 10,000 to 100,000 myelinated axons of varying diameters was generated. These axons were randomly and discretely distributed along the x, y, and z axes of the optic nerve cross-section. This model maintained its spatial distribution during subsequent optimization, serving as a standardized testing environment for evaluating axonal activation.

[0050] Using the Python interface provided by the NEURON simulation platform, random generation was performed. Each myelinated axon is an independent root. The transverse coordinate of each myelinated axon within the optic nerve cross-section. Following a uniform distribution pattern, ensure that fibers cover the entire cross-section of the optic nerve, including the peripheral regions. Myelinated axon diameter... Settings to The distribution of myelin axon diameters follows a bimodal Gaussian distribution model to realistically simulate the physiological characteristics of the optic nerve, where thinner axons are the majority and thicker axons are only a minority.

[0051] Each randomly generated myelinated axon employs the classic McIntyre-Richardson-Greeer (MRG) multilayer myelin model to accurately simulate the skip-conduction characteristics of myelinated nerves. Each myelinated axon is constructed as a multi-segment cable model composed of multiple interconnected sections with different electrical properties. Specifically, this model includes a fixed-length unmyelinated region, the node of Ranvier, and internodal segments surrounded by multiple layers of myelin. To more realistically simulate the impedance characteristics and current conduction paths of biological tissues, the internodal segments are further finely divided into the paranodal region adjacent to the node of Ranvier, the proximal paranodal region, and the isohypermyelinated region in between. The geometric dimensions of each anatomical unit are not fixed but linearly scaled according to the diameter of the myelinated axon; for example, the internodal spacing is set to 100 times the diameter of the myelinated axon. In the NEURON simulation environment, a nonlinear ion channel mechanism, including fast sodium channels, persistent sodium channels, and slow potassium channels, is embedded at the node of Ranvier to simulate the generation and conduction of action potentials. Meanwhile, the internodal segments are endowed with linear leakage current channel properties characterized by high resistance and low capacitance to simulate the passive insulation function of the myelin sheath. Finally, through the extracellular mechanism module of the NEURON software, the spatial coordinates of each segment are mapped in real time to the three-dimensional potential distribution extracted from the multiphysics simulation, thereby realizing the dynamic driving of nerve fibers by the extracellular electric field.

[0052] After completing the above modeling of the random distribution of multiple axons, the positions of subsequent myelinated axons will not be changed, so as to count the total number of myelinated axon activations under each generation of electrode schemes, thereby eliminating the evaluation bias caused by changes in the neural model.

[0053] S2, Multiphysics Modeling Based on Multiphysics Simulation (COMSOL):

[0054] Using the interface between Python and the multiphysics simulation (COMSOL) software, the optimization variable set of the current individual is automatically imported into the COMSOL software to generate a three-dimensional finite element model of the physical structure of the optic nerve and electrode microneedles. Different conductivity parameters are set for different parts of the tissue (such as nerve bundles, nerve sheaths, and surrounding body fluids), and a stimulating current is applied at the exposure site of the electrode microneedles. The three-dimensional spatial potential distribution data within the optic nerve model under this current stimulation is solved and extracted.

[0055] After obtaining the electrode layout parameters for the current generation, the main control script automatically imports the three-dimensional coordinate data into the finite element simulation software for modeling. The system first dynamically generates a corresponding number of spherical electrode geometries within the optic nerve model based on the coordinate vectors. To accurately simulate the physical occupancy of the implanted electrodes, the system performs geometric difference operations to remove the space occupied by the electrodes from the optic nerve tissue model, thereby establishing a nerve-electrode coupling model with precise anatomical topology.

[0056] After the neural-electrode coupling model is constructed, the system automatically performs physical property assignment. First, the platinum-iridium alloy material properties with high electrical conductivity are assigned to all spherical electrode domains. Then, within the current field physical environment, the system automatically identifies and selects the entire outer surface of each spherical electrode as the stimulation interface, and uniformly applies a stimulation current of a preset intensity as the excitation source. To ensure computational accuracy, the system performs mesh refinement on the electrode surface and adjacent neural tissue regions to ensure computational convergence at points of drastic change in the electric field gradient. Finally, the system solves for and extracts the three-dimensional spatial potential distribution data within the optic nerve model and exports it in a standard data format, providing field source input for subsequent neural dynamic coupling.

[0057] S3, Algorithm initialization and constraint methods considering real biological safety and electrode manufacturing process are defined:

[0058] An automated layout optimization strategy is achieved by integrating the Covariance Matrix Adaptive Evolutionary Strategy (CMA-ES) algorithm into a Python master script. The algorithm is initially set to a random distribution and configured with an initial search step size. The size is 100 micrometers. The population size is dynamically set to 8 to 10 individuals based on a 32-dimensional independent variable space (corresponding to the y, z coordinates of the 16 optimization electrodes), and the maximum number of iterations is set to 30 to 50 generations. A convergence threshold is set for the algorithm. for Through adaptive adjustment of the covariance matrix, the system automatically explores and outputs the electrode spatial distribution scheme that achieves the global optimal activation coverage of the optic nerve.

[0059] After constructing the physiological model of random distribution of multiple axons, this embodiment further clarifies the mathematical model and physical boundaries of the optimization algorithm. First, the objective function of the optimization algorithm is set as maximizing the overall activation rate of optic nerve fibers, and its mathematical expression is:

[0060]

[0061] Symbol explanation:

[0062] - The fitness value of the current electrode layout scheme, i.e., the overall activation rate of optic nerve fibers;

[0063] - The total number of axons in a random distribution model of multiple axons;

[0064] - Step function: If the input is true, output 1; otherwise, output 0.

[0065] - : No. The action potential emission state of the root axon under electrical stimulation is considered true if an action potential is generated and conducted, otherwise it is false.

[0066] To ensure that the optimization process complies with biosafety and engineering feasibility, this embodiment establishes a series of strict physical and physiological constraints:

[0067] Electrical stimulation intensity constraint: single-channel stimulation current Must meet

[0068]

[0069] Symbol explanation:

[0070] - Single-channel stimulation current;

[0071] - The single-channel safety current threshold is set to 10 microamps to avoid electrochemical damage and thermal damage to nerve tissue caused by excessive charge injection.

[0072] Electrode number constraint: In this optimization framework, the total number of electrodes... As a preset constant before the algorithm starts, it needs to be set in advance based on the number of hardware channels of the prosthesis system. For example, It can be set to 12 or 16, or other values.

[0073] Electrode spacing constraint: To ensure the mechanical strength of the electrode base and reduce the risk of tissue tearing during microneedle insertion, the linear spatial distance between any two electrode microneedle points is constrained. Must meet

[0074]

[0075] Symbol explanation:

[0076] - : No. The and the first The linear distance between each electrode microneedle point;

[0077] - : A unit of length, micrometer.

[0078] After initial optimization, the electrode points exhibit a double-layer circular distribution, but the distribution of each layer is relatively random. To ensure the feasibility of surgical implantation, the optimization constraints should be further refined to determine a regularized electrode layout with high engineering feasibility.

[0079] Double-layer radial constraint: The electrode array is spatially divided into inner and outer layers, with electrodes in each layer uniformly distributed at equal angles. The radius of the inner layer distribution is set to... The outer layer distribution radius is To ensure the spatial hierarchy of the electrodes within the optic nerve cross-section, certain constraints must be met.

[0080]

[0081] Symbol explanation:

[0082] - Electrode array inner layer distribution of polar diameter;

[0083] - : Electrode array outer layer distribution of polar diameter;

[0084] - The edge boundary of the anatomical radius of the optic nerve.

[0085] In defining the set of independent variables for optimization, this embodiment uses polar coordinates to parameterize the double-layer electrode array. (Independent variable set) Represented as:

[0086]

[0087] Symbol explanation:

[0088] - : Finding the optimal set of independent variables;

[0089] - The inner layer of the electrode array that satisfies the above radial constraints has a polar diameter distribution.

[0090] - The initial polar angle offset of the inner layer array relative to the reference coordinate axis is used to determine the phase distribution of the inner layer electrodes in the circumferential direction.

[0091] - The outer layer of the electrode array that satisfies the above radial constraints has a polar radius distribution.

[0092] - The initial polar angle offset of the outer array relative to the reference coordinate axis is used to determine the phase distribution of the outer electrodes in the circumferential direction.

[0093] By performing a global search within the multidimensional feasible region defined by the above constraints, the algorithm can automatically iterate and output the feasible result. Electrode space layout scheme that achieves maximum value.

[0094] S4, Automated Coupled Modeling Based on Python Interface:

[0095] Using the interface between Python and the multiphysics simulation (COMSOL) software, the coordinates of multiple electrodes for the next iteration are calculated based on the covariance matrix adaptive evolution strategy (CMA-ES) algorithm. The positions of electrode points in the optic nerve are automatically set, and the potential distribution is derived after solving.

[0096] S5, Neurodynamic Simulation and Fitness Assessment:

[0097] The potential distribution matrix is ​​mapped to the axon model constructed in S1 and used as an extracellular driving source to solve the nonlinear dynamic response of the axon transmembrane potential. The axon activation rate (main objective) and safety index (constraint) of the current scheme are jointly calculated to generate a fitness ranking table containing a default flag vector, which is used to distinguish between superior and inferior electrode arrangement schemes.

[0098] After acquiring the three-dimensional potential distribution data derived from finite element simulation, the system applies it as an extracellular electrical stimulation source to the static optic nerve fiber model constructed in step S1 via an automated interface. The system first performs spatial coordinate mapping, precisely aligning the values ​​of each discrete sampling point within the potential field with the specific spatial locations of the node of Ranvier and internodal segments in each axon model. Subsequently, using a neurodynamics calculation engine, the potential at each sampling point is injected as an extramembrane potential fluctuation into the axon's cable equation. By iteratively solving the variation law of transmembrane ion current, the physiological response of nerve fibers under complex electric field environments is simulated.

[0099] After the dynamic simulation is completed, the system automatically performs a fitness assessment. The system monitors the peak membrane potential at the end of each axon model to determine whether it has successfully fired and conducted action potentials, and counts the total number of successfully activated axons in real time. Subsequently, the system calculates the proportion of activated axons to the total number of axons, obtaining the nerve fiber activation rate under the current electrode layout scheme. This rate is then used as the fitness value of the current individual and fed back to the Covariance Matrix Adaptive Evolutionary Strategy (CMA-ES) optimization algorithm. If the fitness increases, it indicates that the population scheme is effective, and the algorithm will then conduct further in-depth exploration.

[0100] S6, Adaptive update of covariance matrix and global solution output:

[0101] The Python main control script collects the fitness values ​​of all individuals in the current population, drives the optimization algorithm to perform update operations such as crossover, mutation, and selection, and generates the set of independent variables for the next generation of electrode distribution optimization. Steps S4 to S6 are executed repeatedly for joint simulation until the set maximum number of iterations is reached or the optimal fitness of the population meets the convergence threshold. The optimization loop then exits, and the electrode spatial distribution parameters with the highest global fitness are output, which is the final optimized layout scheme for the optic nerve electrode array.

[0102] The Python main control script continuously collects the fitness values ​​of all individuals in the current population, driving the Covariance Matrix Adaptive Evolutionary Strategy (CMA-ES) algorithm to perform covariance matrix updates and adaptive step size adjustments. The initial electrode adjustment step size is... Based on their fitness values, the electrode adjustment step size is gradually reduced among several populations with high fitness until the fitness gradually stabilizes and reaches its maximum. The system iteratively executes steps S4 to S6, continuously iterating and evolving to generate the next generation of electrode distribution independent variable set until a preset termination condition is met, i.e., reaching the maximum number of iterations or the change in the optimal fitness of the population is lower than the convergence threshold. During the optimization process, the algorithm automatically learns the nonlinear relationship between electric field distribution and neural activation, continuously converging to a feasible domain space with high coverage.

[0103] After the optimization loop exits, the system outputs the electrode spatial distribution parameters with the highest global fitness, which is the final determined optimized layout scheme of the optic nerve electrode array.

[0104] In this embodiment, the system ultimately outputs a set of optimal coordinates for the four inner electrodes and the twelve outer electrodes. The electrode layout before and after optimization is as follows: Figure 2 and Figure 3 As shown. Electrode coverage under the two electrode layouts is as follows. Figure 4 , Figure 5 As shown in the diagram, this scheme significantly improves the overall activation coverage of optic nerve fibers while strictly adhering to electrode spacing constraints and physiological safety boundaries. A schematic diagram illustrating the improvement is shown below. Figure 6 As shown. Finally, the system automatically generates a technical report and configuration file containing the optimal electrode diameter, polar angle, and insertion depth to guide the actual fabrication and implantation of the high-performance optic nerve prosthesis electrode array. Throughout the process, the fitness value changes as shown in the diagram. Figure 7 As shown.

[0105] Therefore, in response to the technical bottlenecks of existing technologies, such as the difficulty in balancing high nerve fiber coverage and low tissue damage risk, the optic nerve electrode array layout optimization scheme provided by this invention uses an automated algorithm to perform global optimization under strict safety and anatomical constraints, which significantly improves the electrical stimulation coverage of the optic nerve prosthesis.

[0106] Example 2

[0107] Visual nerve electrode array layout optimization system

[0108] This invention provides a specific embodiment of a optic nerve electrode array layout optimization system. The system runs on a computer device equipped with a Python environment, NEURON software, and COMSOL software. It includes interconnected physiological model construction modules, multiphysics modeling modules, algorithm initialization modules, automated coupling modules, fitness evaluation modules, and global optimization output modules.

[0109] The physiological model construction module calls the NEURON software to construct a static physiological model of random distribution of optic nerve fibers, generating multiple myelinated axon models with different diameters. The axons are randomly and discretely distributed in the three-dimensional space of the optic nerve. The spatial position of all axons is fixed as a standardized testing environment. Each axon adopts the McIntyre-Richardson-Greeer (MRG) multilayer myelin model, which divides it into nodes of Ranvier and internodes. The internodes are further divided into paranodes, proximal paranodes, and isohyetal myelin regions.

[0110] The algorithm initialization module integrates the Covariance Matrix Adaptive Evolutionary Strategy (CMA-ES) algorithm, sets the optimization objective of maximizing axon activation rate, and introduces constraints such as single-channel stimulation current intensity, number of electrodes, spatial linear distance between any two electrode microneedles, and radial distribution constraints of the double-layer electrode array. The algorithm is configured with an initial search step size of 100 micrometers, a population size of 8 to 10 individuals, a maximum number of iterations of 30 to 50 generations, and a convergence threshold of (10^{-6}). The double-layer electrode array is parameterized using polar coordinates.

[0111] The automated coupling module imports the electrode coordinates generated by the algorithm into the multiphysics simulation (COMSOL) software via a Python interface. The multiphysics modeling module simultaneously builds a nerve-electrode coupling model, sets corresponding conductivity parameters for different tissue sites, refines the mesh of the electrode surface and adjacent nerve tissue, applies stimulation current, and solves for the three-dimensional potential distribution. The fitness evaluation module maps the potential distribution to the axon model, calculates axon activation rate and safety indices, and generates a fitness ranking table. The global optimization output module drives iterative algorithm updates, ultimately outputting an optimal parameter set including electrode diameter, polar angle, and insertion depth, and automatically generating technical reports and configuration files.

[0112] The embodiments described above are merely examples of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application.

Claims

1. A method for optimizing the layout of an optic nerve electrode array, characterized in that, Includes the following steps: S1. Construct a static physiological model of random distribution of optic nerve fibers, generate multiple myelinated axon models with different diameters, and the axons are randomly and discretely distributed in the three-dimensional space of the optic nerve. The spatial distribution of this model remains unchanged during the subsequent optimization process. S2, Based on multiphysics simulation software, a model of the optic nerve and electrodes is established to obtain the three-dimensional potential distribution in the optic nerve under current stimulation. S3, Initialize the covariance matrix adaptive evolution strategy algorithm, set the optimization objective of maximizing the axon activation rate, and introduce constraints related to biosafety and electrode manufacturing process; S4, through the interface between the main control script and the multiphysics simulation software, imports the electrode space coordinates generated by the algorithm into the multiphysics model, simulates the physical occupancy after electrode implantation and applies current stimulation, solves and derives the three-dimensional potential distribution. S5, map the three-dimensional potential distribution onto the axon model, solve the axon transmembrane potential response, calculate the axon activation rate and safety index of the current scheme, and generate a fitness ranking table containing default flag vectors; S6 updates the algorithm parameters based on fitness feedback, iterates until the convergence condition is met, and outputs the globally optimal set of electrode spatial distribution parameters.

2. The optic nerve electrode array layout optimization method according to claim 1, characterized in that, In step S1, 10,000 to 100,000 myelinated axon models are generated. The axon diameter is between 1 micrometer and 8.7 micrometers and follows a bimodal Gaussian distribution. Each axon adopts a multi-layer myelin model and is divided into nodes of Ranvier and internode segments. The internode segments are further divided into paranode zones, near-paranode zones, and isodiameter myelin zones.

3. The optic nerve electrode array layout optimization method according to claim 1, characterized in that, In step S2, the electrode geometry is dynamically generated based on the electrode coordinates, and the space occupied by the electrode is eliminated by performing geometric difference operation to establish a nerve-electrode coupling model. Different conductivity parameters are set for the nerve bundle, nerve sheath, and surrounding body fluid. Stimulating current is applied to the electrode surface, and the electrode surface and adjacent nerve tissue are refined by mesh processing.

4. The optic nerve electrode array layout optimization method according to claim 1, characterized in that, The constraints introduced in step S3 include single-channel stimulation current intensity constraints, electrode number constraints, spatial linear distance constraints between any two electrode microneedle points, and double-layer radial distribution constraints of the electrode array.

5. The optic nerve electrode array layout optimization method according to claim 4, characterized in that, In step S3, the initial search step size of the algorithm is set to 100 micrometers, the population size is set to 8 to 10 individuals, the maximum number of iterations is set to 30 to 50 generations, and the convergence threshold is set to 10 to the power of negative 6. The double-layer electrode array is parameterized using polar coordinates, and the independent variables include the inner layer distributed polar diameter, the inner layer initial polar angle offset, the outer layer distributed polar diameter, and the outer layer initial polar angle offset.

6. A system for optimizing the layout of an optic nerve electrode array, characterized in that, It includes a physiological model building module, a multiphysics modeling module, an algorithm initialization module, an automated coupling module, a fitness evaluation module, and a global optimization output module; The physiological model building module is used to construct a static physiological model of randomly distributed optic nerve fibers, generating multiple myelinated axon models with different diameters; The multiphysics modeling module is used to establish models of the optic nerve and electrodes, and to obtain the three-dimensional potential distribution within the optic nerve under electrical stimulation. The algorithm initialization module is used to initialize the covariance matrix adaptive evolution strategy algorithm, set the optimization objective, and introduce constraints related to biosafety and electrode manufacturing process. The automated coupling module is used to import the electrode space coordinates generated by the algorithm into the multiphysics model to simulate electrode implantation and solve the three-dimensional potential distribution. The fitness evaluation module is used to map the three-dimensional potential distribution to the axon model, calculate the axon activation rate and safety index, and generate a fitness ranking table containing default flag vectors. The global optimization output module is used to update the algorithm parameters based on fitness feedback, iterates until convergence, and outputs the optimal electrode distribution parameter set.

7. The optic nerve electrode array layout optimization system according to claim 6, characterized in that, The physiological model construction module generates 10,000 to 100,000 myelinated axon models with axon diameter between 1 micrometer and 8.7 micrometers and following a bimodal Gaussian distribution. Each axon adopts a multi-layer myelin model, which is divided into nodes of Ranvier and internode segments. The internode segments are further divided into paranode zones, near-paranode zones, and isodiameter myelin zones.

8. The optic nerve electrode array layout optimization system according to claim 6, characterized in that, The multiphysics modeling module dynamically generates electrode geometry based on electrode coordinates, performs geometric difference operations to eliminate the space occupied by the electrodes to establish a nerve-electrode coupling model, sets corresponding conductivity parameters for different tissue sites, applies stimulation current to the electrode surface, and refines the mesh of the electrode surface and adjacent nerve tissue.

9. The optic nerve electrode array layout optimization system according to claim 6, characterized in that, The constraints introduced by the algorithm initialization module include single-channel stimulation current intensity constraints, electrode number constraints, spatial linear distance constraints between any two electrode microneedle points, and double-layer radial distribution constraints of the electrode array.

10. The optic nerve electrode array layout optimization system according to claim 9, characterized in that, The algorithm initialization module sets the initial search step size to 100 micrometers, the population size to 8 to 10 individuals, the maximum number of iterations to 30 to 50 generations, and the convergence threshold to 10 to the power of -6. The double-layer electrode array is parameterized using polar coordinates, with independent variables including the inner layer distributed polar diameter, the inner layer initial polar angle offset, the outer layer distributed polar diameter, and the outer layer initial polar angle offset. The global optimization output module finally outputs the optimal parameter set containing the electrode polar diameter, polar angle, and penetration depth.

Images