Intelligent control method and system based on in-vitro culture biological neural network
By scanning the electrode regions and calculating saliency parameters of in vitro biological neuronal networks, combined with a pre-trained classification model, dynamic regulation of the neural state space was achieved. This solved the problem of low localization accuracy in in vitro neuronal networks, improved control accuracy and network plasticity, and made the network suitable for complex task environments.
Patent Information
- Application Number
- CN202510945687.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-09
- Publication Date
- 2025-10-21
- Estimated Expiration
- 2045-07-09
AI Technical Summary
Existing in vitro neural networks have difficulty locating representative neurons in the final actual control, resulting in reduced control accuracy.
By scanning the channels of the biochip to select the electrode area, stimulation feedback is obtained and significance parameters are calculated, channel vectors are constructed, and actions are output using a pre-trained classification model to achieve dynamic regulation of the neural state space. In vitro BNNs are used as a physical reserve pool for nonlinear calculations and integration of synaptic connection networks.
It improves the control accuracy of neural networks, enhances the plasticity and adaptability of the network, and is able to maintain stable control strategies in complex environments. It is suitable for multi-objective coordinated control, bionic robot navigation, and human-computer interaction systems.
Smart Images

Figure CN120818639A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of biological neuron technology, and in particular to an intelligent control method and system based on in vitro cultured biological neuron networks. Background Art
[0002] In recent years, with the rapid development of brain-like intelligence and neural engineering, biological neural networks cultured in vitro, as a new type of information processing carrier, have gradually demonstrated great potential in neural signal modeling, intelligent decision-making control, and biological computing, becoming one of the important research directions of brain-like systems. In vitro biological neural networks refer to neural networks formed by self-organizing and connecting biological neurons in a non-physiological environment through in vitro culture technology. These networks not only have the nonlinear computing power and synaptic plasticity unique to biological nervous systems, but can also be finely manipulated and observed under experimental conditions. Compared with traditional artificial neural networks, in vitro biological neural networks have unique advantages in energy efficiency, parallel processing, real-time learning, and autonomous adaptation, and have therefore become an important breakthrough in the intersection of brain science research and artificial intelligence.
[0003] Electrical stimulation is one of the most mature and widely used methods for constructing real-time interactions between in vitro biological neural networks (BNNs) and external systems. This method applies electrical signals to neurons via external electrodes, triggering changes in membrane potential and thus inducing a network response. Multi-electrode array (MEA) technology, in particular, has become an indispensable core interface tool in BNN research due to its high temporal and spatial resolution and multi-channel simultaneous stimulation and recording capabilities. It not only supports high-precision observation and quantitative analysis of neural activity but also provides an efficient channel for information exchange between BNNs and virtual systems, enabling neural networks to possess dynamic processing capabilities within a closed perception-feedback loop.
[0004] In recent years, BNNs have demonstrated excellent plasticity and intelligent behavior in a variety of typical interactive tasks using MEAs. For example, the Brainoware system, by culturing human brain organoids on MEAs and building a hardware architecture based on reservoir computing, has successfully achieved speech recognition and nonlinear equation prediction tasks, demonstrating the memory and mapping capabilities of BNNs in time series data processing. However, existing in vitro neural networks have difficulty locating representative neurons in actual control, resulting in reduced control accuracy. Summary of the Invention
[0005] In view of this, embodiments of the present invention provide an intelligent control method and system based on an in vitro cultured biological neuronal network to eliminate or improve one or more defects in the prior art.
[0006] One aspect of the present invention provides an intelligent control method based on an in vitro cultured biological neuronal network, the method comprising the following steps:
[0007] Extract and culture neurons from organisms in vitro, and construct biochips using microelectrode array chips;
[0008] Scanning the biochip, and selecting an electrode region of the biochip based on a firing rate obtained by scanning a channel of the biochip, wherein the electrode region of the biochip includes an electrode region 1 and an electrode region 2;
[0009] Acquiring a real-time task screen to be determined, determining a relative distance based on the position of the subject and the target in the real-time task screen to be determined, and selecting electrode region 1 or electrode region 2 for applying stimulation to the biochip and a stimulation intensity based on the relative distance;
[0010] Acquiring feedback from the biochip to stimulation and constructing it into an initial feedback vector, calculating a significance parameter for each channel in the initial feedback vector, selecting each channel based on the significance parameter of each channel, and constructing a channel vector based on the selected channels;
[0011] The channel vector is input into a pre-trained classification model, and the action is output by the classification model.
[0012] Using the above scheme, this approach induces in vitro biological neural networks (BNNs) to respond specifically to external inputs, achieving dynamic control of the neural state space and providing stable, distinguishable, high-dimensional state vectors for intelligent control systems. In this approach, BNNs are modeled as a physically implemented reservoir computing system (RC). Reservoir computing is a structure for time series data processing. Its core mechanism is to map inputs to a high-dimensional state space through a reservoir with nonlinear dynamic characteristics, while the output is extracted by a linearly trainable readout layer. Traditional RC systems often rely on recurrent neural networks (RNNs) as reservoirs. In vitro BNNs, due to their natural biophysical properties, are ideal vehicles for constructing physical reservoirs. In vitro BNNs possess two key characteristics: nonlinear computational capability and network plasticity. Neurons integrate spatial and temporal information through a complex network of synaptic connections, demonstrating highly nonlinear signal processing capabilities. At the same time, the neural state will undergo temporary changes after receiving input and gradually return to the ground state in the absence of continuous stimulation, reflecting typical short-term memory characteristics. Ultimately, the significance parameters of each channel can be calculated through feedback from the biochip, and the channel can be selected to construct a channel vector, which can locate representative neurons and ensure control accuracy.
[0013] In some embodiments of the present invention, in the step of obtaining the feedback of the biochip to a stimulus and constructing it into an initial feedback vector, and calculating a significance parameter for each channel in the initial feedback vector:
[0014] constructing the channel number of each channel in which a Spike response occurs within a first preset time period after the biochip is stimulated as an initial feedback vector;
[0015] For each channel in the initial feedback vector, a significance parameter is calculated based on the number of spike responses of the channel in the initial feedback vector and the number of spike responses of the channel in a second preset time period before the channel is stimulated.
[0016] In some embodiments of the present invention, in the step of calculating the significance parameter based on the number of spike responses of the channel in the initial feedback vector and the number of spike responses of the channel in the second preset time period before stimulation, the significance parameter is calculated using the following formula:
[0017]
[0018] Among them, R i represents the number of Spike responses of channel i in the initial feedback vector; B irepresents the number of Spike responses of channel i in the second preset time period before being stimulated, S i represents the significance parameter of channel i.
[0019] In some embodiments of the present invention, in the step of selecting each channel based on the significance parameter of each channel and constructing a channel vector based on the selected channel, the significance parameter is compared with a preset threshold to determine the selected channel, and the channel number of the selected channel is used as the value of each dimension in the channel vector to obtain the channel vector.
[0020] In some embodiments of the present invention, the steps of extracting and culturing neurons from an organism in vitro and constructing a biochip using a microelectrode array chip include:
[0021] The fetal rat cortical tissue was peeled off and cut into small pieces, washed in DMEM culture medium, and transferred to a digestion solution containing 0.125% trypsin and deoxyribonuclease I for digestion at 37°C for 20 minutes.
[0022] After sufficient dissociation by pipetting, a single-cell suspension was obtained, and the cells were separated by centrifugation at 80 × g for 5 minutes, and the supernatant was removed;
[0023] The obtained cells were resuspended in complete culture medium containing Neurobasal-Plus medium, B27-Plus, GlutaMax, and penicillin-streptomycin double antibody;
[0024] Using a multi-electrode array chip, cells were seeded and a biological neural network cultured in vitro was constructed. A freshly prepared 1% Terg-a-zyme solution was used to clean the chip surface at room temperature for 2 hours. The chip was then rinsed three times with deionized water and thoroughly dried. The dried chip was sterilized in 70% ethanol for 12 to 16 hours, rinsed three times with sterile deionized water, and placed in a 90mm sterile culture dish. A high-humidity chamber was constructed with a 30mm wet dish. 0.6mL of complete culture medium was added to each chip for pre-culture, and the chip was placed in a 37°C, 5% CO2 incubator for two days to improve surface adaptability.
[0025] After the pre-culture, the chip surface was double-coated in sequence. First, 0.1 mg / mL poly-D-lysine (PDL, Sigma-Aldrich, #P6407) was used for incubation at 37°C for 3 hours, then rinsed three times with sterile water and dried; then 0.02 mg / mL laminin was used for incubation at 37°C for 1 hour as the second layer of coating. After incubation, no washing was done and cells were directly inoculated to obtain a biochip.
[0026] In some embodiments of the present invention, during the cell seeding process, 50 μL of neuronal cell suspension is dripped onto the central electrode area of the chip. The suspension concentration is set between 12,000–16,000 cells / μL depending on the cell type. After seeding, the chip is covered with a sterilized cover and placed in a 37°C incubator for 2 hours to promote cell attachment. Subsequently, 0.6 mL of complete culture medium is slowly added along the side of the chip. The culture environment is controlled at 37°C, 5% CO2, and relative humidity greater than 95%.
[0027] After cells have completed attachment, they enter the routine culture phase: starting from the first day, the culture medium is changed three times a week, each time replacing 50% of the original volume. Neural electrical activity is recorded between 4 and 24 hours after the culture medium is changed, and the time interval between the change and the recording is kept consistent as much as possible. As the culture time progresses, neurons gradually grow, differentiate, and form a stable synaptic connection network. An in vitro neural network with bioelectrical activity is formed on the chip, providing a basis for subsequent electrical stimulation, training, and control experiments.
[0028] In some embodiments of the present invention, in the step of selecting the electrode regions of the biochip based on the firing rates obtained by scanning the channels of the biochip:
[0029] Scan and record the overall activity of the biochip and calculate the spontaneous discharge rate of each channel;
[0030] According to the discharge activity, the first preset number of channels with the highest discharge rate are selected as candidate recording channels, and a two-dimensional spatial discharge heat map is constructed;
[0031] A sliding window-based two-dimensional firing density search method was used to divide the two-dimensional spatial discharge heat map into non-overlapping windows of fixed size, and the average discharge rate of the channel was calculated by traversing each window.
[0032] All windows are sorted in descending order according to the average discharge rate, and the window area ranked in the first interval is selected as the candidate stimulation area, and two candidate stimulation areas with a distance greater than the first distance are selected as electrode area 1 and electrode area 2 respectively.
[0033] In some embodiments of the present invention, the step of selecting the electrode area of the biochip based on the firing rate scanned from the channel of the biochip also includes selecting electrode area three and electrode area four from the candidate stimulation areas, wherein electrode area three is used to apply stimulation based on a special category of target bodies; and electrode area four is used to apply stimulation based on multiple target bodies.
[0034] In some embodiments of the present invention, the step of selecting the electrode area of the biochip based on the firing rate scanned from the channel of the biochip also includes selecting electrode area five from the candidate stimulation areas, and the electrode area five is used to apply stimulation after the body movement is successful.
[0035] In some embodiments of the present invention, electrode area three is used to represent the target body: if the target body position is of a special category, a 100mV pulse is synchronously applied to the third pole, otherwise it is not activated, which is used to guide the priority response to key targets. Electrode area four is used to encode the complexity of the local environment: when there are multiple target bodies in the scene at the same time, a 200mV pulse is applied to electrode area four; if there is only one target body, it is not activated. This encoding can guide BNNs to enhance processing priority or increase response intensity in multi-target situations, which helps the system maintain a stable control strategy in a complex environment. Electrode area five carries historical hit feedback: if the previous action was successful, a 200mV reward pulse is applied to electrode area five 1s after the action is completed to simulate short-term memory modulation.
[0036] In some embodiments of the present invention, the method further comprises training the biochip, comprising:
[0037] Symmetrical bidirectional pulses with the same preset width and multiple preset pulse amplitudes are applied to electrode area 1 and electrode area 2 respectively, forming multiple stimulation patterns. Each pattern is repeated at equal time intervals. Multiple stimulation patterns are performed sequentially to form one round of training.
[0038] A perturbation stimulation mechanism is introduced in each round of training. A preset number of random repeated pulses are added between each mode training. The pulse amplitude is within a preset range, the interval is the same, and the stimulation form is a symmetrical biphasic pulse.
[0039] The second aspect of the present invention also provides an intelligent control system based on an in vitro cultured biological neuronal network, the system comprising a computer device, the computer device comprising a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.
[0040] The third aspect of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps implemented by the aforementioned intelligent control method based on in vitro cultured biological neuronal networks.
[0041] Additional advantages, objects, and features of the present invention will be described in part in the following description and will become apparent to those skilled in the art after studying the following or may be learned by practice of the present invention. The objects and other advantages of the present invention may be particularly pointed out and attained in the description and drawings.
[0042] Those skilled in the art will understand that the purposes and advantages that can be achieved by the present invention are not limited to the above specific descriptions, and the above and other purposes that can be achieved by the present invention will be more clearly understood based on the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention.
[0044] Figure 1 This is a schematic diagram of an implementation method of the intelligent control method based on in vitro cultured biological neuronal networks;
[0045] Figure 2 This is a schematic diagram of the overall architecture of the intelligent control method based on in vitro cultured biological neuronal networks;
[0046] Figure 3 This is the closed-loop control schematic diagram of this scheme.
[0047] Figure 4 This is a schematic diagram of the Spike firing rate activity graph;
[0048] Figure 5 This is a schematic diagram of the biochip training for this scheme;
[0049] Figure 6 This is a statistical diagram of the discharge activity of the biochip in this scheme;
[0050] Figure 7 This is a histogram diagram of the electrode channel distribution of this scheme;
[0051] Figure 8 This is a schematic diagram of the electrode channel correlation heat map of this scheme;
[0052] Figure 9 This is a comparison chart of the average correlation coefficients of this scheme;
[0053] Figure 10 This is the global efficiency comparison chart of this solution;
[0054] Figure 11 This is a comparison chart of the modularity coefficient of this scheme;
[0055] Figure 12 There is a random stimulus classification recognition accuracy graph for this scheme. DETAILED DESCRIPTION
[0056] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.
[0057] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, the accompanying drawings only show structures and / or processing steps closely related to the solutions according to the present invention, while other details that are not closely related to the present invention are omitted.
[0058] The existing technology as a whole has problems such as insufficient utilization of stimulus information, insufficient mining of spatiotemporal features, and limited task complexity, and has not yet fully demonstrated its potential in dynamic perception and adaptive learning.
[0059] like Figure 1-3 As shown, the present invention proposes an intelligent control method based on an in vitro cultured biological neuronal network, the method comprising the following steps:
[0060] Step S100, extracting and culturing neurons from an organism in vitro, and constructing a biochip using a microelectrode array chip;
[0061] Step S200, scanning the biochip, and selecting an electrode region of the biochip based on the emission rate obtained by scanning the channels of the biochip, wherein the electrode region of the biochip includes an electrode region 1 and an electrode region 2;
[0062] Step S300, obtaining a real-time task screen to be determined, determining a relative distance based on the position of the subject and the target in the real-time task screen to be determined, and selecting electrode region 1 or electrode region 2 for applying stimulation to the biochip, as well as the stimulation intensity, based on the relative distance;
[0063] The real-time task screen is a screen obtained by capturing a dynamic task scene, and the real-time task screen may be a two-dimensional screen or a three-dimensional screen.
[0064] Step S400, obtaining feedback from the biochip in response to stimulation and constructing it into an initial feedback vector, calculating a significance parameter for each channel in the initial feedback vector, selecting each channel based on the significance parameter of each channel, and constructing a channel vector based on the selected channels;
[0065] During implementation, to ensure consistency in the input feature dimensions, empty channels (zero vectors) are used to fill the gap when the number of valid channels is less than 100. If the number exceeds 100, the top 100 channels are retained by significance score to construct a channel vector. The resulting channel number ranking vector not only reflects the intensity of a neuron's discharge response to a specific stimulus but also captures its selective spatial distribution. This vector exhibits good distinguishability and can be used for subsequent classification, decoding, and behavioral control strategies.
[0066] Step S500: input the channel vector into a pre-trained classification model, and output an action through the classification model.
[0067] Specifically, the Bagging Tree algorithm is used as the classification model. This method is particularly suitable for neural data classification tasks with limited samples and high feature dimensions. Through three rounds of training, such as Figure 8 、 9 As shown in Figures 10, 11, and 12, the system establishes a label mapping relationship between each stimulation pattern and the corresponding behavioral instruction. After training, the classification model is solidified and used in the real-time recognition stage. The neural features corresponding to each stimulation response are input into the classifier, and the corresponding action is output.
[0068] Figure 8 and 9 The results of the heatmaps and average correlation coefficient changes of the discharge activity correlation between electrode channels in the four training stages (n=8) are shown respectively. Overall, the level of co-discharge between neurons showed a continuous upward trend during the training process. Before training, the correlation between channels was already at a high level, reflecting the existence of a certain degree of synchronization and activity redundancy in biological neural networks. As the external stimulation training progressed, the average correlation coefficient increased after each round of training, indicating that the co-discharge relationship between neurons continued to strengthen. This trend shows that in the process of adapting to repeated stimulation, the network's internal connections are gradually optimized and reorganized, forming a tighter and more efficient co-discharge pattern. The continuous enhancement of discharge correlation not only reflects the reconstruction of neural activity patterns, but also suggests that training promotes the improvement of network functional integration capabilities. The results reveal the activity plasticity of in vitro biological neural networks under external intervention and the potential for functional remodeling driven by training.
[0069] Figure 10 The changes in the global efficiency (Global Efficiency) of the neural network before training and after three rounds of training are shown (n=8). The experimental results show that the global efficiency has improved to varying degrees after each round of training, showing a continuous upward trend. As an important topological indicator for measuring the information integration ability of a neural network, global efficiency reflects the inverse value of the average shortest path between nodes in the network. The higher the value, the more efficient the information is transmitted throughout the network. This result shows that under the action of repeated stimulation training, the dispersed neurons within the network gradually establish a closer connection structure, thereby enhancing the communication efficiency and collaborative processing capabilities between regions. The gradual improvement in global efficiency also reflects that under the guidance of external stimulation, the overall information integration ability and dynamic response performance of the neural network are continuously optimized, reflecting good functional plasticity.
[0070] Figure 11The study presents the changing trends in the modularity coefficient of the neural network before and after three rounds of training (n=8). The experimental results show that the modularity coefficient gradually decreases with the number of training rounds, indicating that the network transitions from an initial state of distinct functional partitioning and local module collaboration to a more integrated and globally collaborative structure. The modularity coefficient reflects the degree to which, after the network is divided into modules (communities), the connections within the modules are dense and the connections between modules are sparse. High modularity means that each subgroup in the network is relatively independent, and information exchange occurs primarily locally; low modularity, on the other hand, indicates a greater tendency towards global collaboration and information integration. The gradual decrease in modularity indicates that, under the stimulation of repeated training, the neural network breaks down existing local information silos, promotes collaborative interaction between different functional areas, and thus supports a more efficient and unified information processing mechanism. This further confirms the process of functional structural reorganization and improvement of the network's integration capabilities during training, reflecting its strong plasticity.
[0071] Figure 12 The classification accuracy of the neural network is evaluated over four training phases using ten-fold cross-validation. The results show that the network's classification accuracy shows a significant improvement with continued training: from 66.5% in the untrained phase to 84.5% after the third training phase, reflecting the network's enhanced ability to recognize and respond to external stimulus patterns. This trend suggests that external stimulus training, particularly when introduced with perturbations, effectively promotes the expression of plasticity and functional reconfiguration in the neural network. At the structural level, the aforementioned analysis has revealed that training leads to a more integrated network connectivity pattern (increased global efficiency and decreased modularity) and enhanced functional synergy between channels (enhanced correlation). These changes provide a more favorable foundation for information transmission and computational support for the neural network. Functionally, the improved classification accuracy indicates not only structural optimization of the network but also enhanced information representation and output discriminability at the task level. Neuronal firing patterns under different stimulus conditions become more stable and distinguishable, demonstrating an overall evolution of the network from "spontaneous response" to "task-driven response."
[0072] Joint analysis: Combined results from the discharge correlation heatmap, global efficiency, modularity coefficient, and electrode degree distribution indicate that the in vitro biological neural network exhibits significant structural remodeling and functional plasticity during training. Initially, the network exhibits strong internal coordination and synchronization, with high redundancy in inter-channel discharge activity and significant modularity, indicating a primary state of integration with distinct localization and compartmentalization. After the first round of training, the correlation increases slightly, the modularity coefficient decreases slightly, and the global efficiency begins to rise. Simultaneously, the degrees of some electrodes increase, indicating a preliminary decorrelation of responses between neurons and a remodeling of information transmission pathways. This phase can be considered the initial stage of functional decoupling and remodeling of the network. With increasing training rounds, the correlation coefficient increases again, the global efficiency continues to grow, the modularity coefficient decreases further, and the electrode degree increases overall, indicating that the network undergoes a dynamic adjustment process from desynchronization to coordination and remodeling. More functional connections are established between neurons, breaking down the previously isolated modular structure and endowing the network with stronger information integration capabilities and global coordination properties.
[0073] This approach not only demonstrates the technical feasibility of using biological neural networks for complex behavior recognition and control tasks, but also embodies the advantages of the deep integration of "artificial systems + biological networks." As a physical reservoir, BNNs possess high parallelism, low power consumption, and structural self-organization. Without the need for deep training of network weights, they can effectively perform recognition and control with small samples, demonstrating the potential application value of natural nervous systems in intelligent control. At the same time, the modulation stimulation mechanism introduced by the system further enhances the learning ability and task adaptability of BNNs, enabling them to achieve enhanced plasticity and experience accumulation through key behavioral feedback, and initially possess the ability for adaptive regulation based on biofeedback. In the future, this approach is expected to be extended to more complex task environments, such as multi-objective coordinated control, bionic robot navigation, and human-computer interaction systems, promoting the development of brain-like intelligence towards practical application.
[0074] Using the above scheme, this approach induces in vitro biological neural networks (BNNs) to respond specifically to external inputs, achieving dynamic control of the neural state space and providing stable, distinguishable, high-dimensional state vectors for intelligent control systems. In this approach, BNNs are modeled as a physically implemented reservoir computing system (RC). Reservoir computing is a structure for time series data processing. Its core mechanism is to map inputs to a high-dimensional state space through a reservoir with nonlinear dynamic characteristics, while the output is extracted by a linearly trainable readout layer. Traditional RC systems often rely on recurrent neural networks (RNNs) as reservoirs. In vitro BNNs, due to their natural biophysical properties, are ideal vehicles for constructing physical reservoirs. In vitro BNNs possess two key characteristics: nonlinear computational capability and network plasticity. Neurons integrate spatial and temporal information through a complex network of synaptic connections, demonstrating highly nonlinear signal processing capabilities. At the same time, the neural state will undergo temporary changes after receiving input and gradually return to the ground state in the absence of continuous stimulation, reflecting typical short-term memory characteristics. Ultimately, the significance parameters of each channel can be calculated through feedback from the biochip, and the channel can be selected to construct a channel vector, which can locate representative neurons and ensure control accuracy.
[0075] In some embodiments of the present invention, in the step of obtaining the feedback of the biochip to a stimulus and constructing it into an initial feedback vector, and calculating a significance parameter for each channel in the initial feedback vector:
[0076] constructing the channel number of each channel in which a Spike response occurs within a first preset time period after the biochip is stimulated as an initial feedback vector;
[0077] For each channel in the initial feedback vector, a significance parameter is calculated based on the number of spike responses of the channel in the initial feedback vector and the number of spike responses of the channel in a second preset time period before the channel is stimulated.
[0078] In practice, the MEA platform collects BNN firing data in real time within a 100-200ms time window after stimulation is applied to the biochip. Feature extraction of the captured neural activity is then performed using a dynamic channel screening method based on a combination of response significance and spatial sparsity. This method first counts the baseline firing rate within 100ms before stimulation and the response firing rate within 100-200ms after stimulation for each channel, and calculates the response significance parameter.
[0079] In some embodiments of the present invention, in the step of calculating the significance parameter based on the number of spike responses of the channel in the initial feedback vector and the number of spike responses of the channel in the second preset time period before stimulation, the significance parameter is calculated using the following formula:
[0080]
[0081] Among them, R i represents the number of Spike responses of channel i in the initial feedback vector; B i represents the number of Spike responses of channel i in the second preset time period before being stimulated, S i represents the significance parameter of channel i.
[0082] In some embodiments of the present invention, in the step of selecting each channel based on the significance parameter of each channel and constructing a channel vector based on the selected channel, the significance parameter is compared with a preset threshold to determine the selected channel, and the channel number of the selected channel is used as the value of each dimension in the channel vector to obtain the channel vector.
[0083] All channels with significance scores exceeding a set threshold (e.g., 1.0) are retained to form a candidate set, and a spatial sparsity constraint is introduced: the MEA chip is divided into equidistant spatial grids, or the K-means clustering algorithm is used for channel partitioning, and the channel with the highest response score is selected in each sub-region to ensure that the final channel set is balanced in spatial distribution.
[0084] In some embodiments of the present invention, the steps of extracting and culturing neurons from an organism in vitro and constructing a biochip using a microelectrode array chip include:
[0085] The fetal rat cortical tissue was peeled off and cut into small pieces, washed in DMEM culture medium, and transferred to a digestion solution containing 0.125% trypsin and deoxyribonuclease I for digestion at 37°C for 20 minutes.
[0086] After sufficient dissociation by pipetting, a single-cell suspension was obtained, and the cells were separated by centrifugation at 80 × g for 5 minutes, and the supernatant was removed;
[0087] The obtained cells were resuspended in complete culture medium containing Neurobasal-Plus medium, B27-Plus, GlutaMax, and penicillin-streptomycin double antibody;
[0088] Specifically, fetal cortical tissue from Beijing Speifu pregnant mice at 15.5 days of gestation was used. To ensure cell activity and tissue integrity, the entire tissue acquisition process was performed on ice. After the peeled cortical tissue was cut into small pieces, it was placed in Zhongke Maichen DMEM culture medium and gently washed twice, and then transferred to a digestion solution containing 0.125% trypsin (#CC017) and deoxyribonuclease I (#D8071) and digested at 37°C for 20 minutes. After full dissociation by pipetting, a single cell suspension was obtained, and the cells were separated by centrifugation at 80×g for 5 minutes, and the supernatant was removed. The obtained cells were resuspended in complete culture medium containing Neurobasal-Plus medium (Gibco, #A3582901), B27-Plus (#A3582801), GlutaMax (#35050061) and penicillin-streptomycin double antibody (#CC004).
[0089] Using a multi-electrode array chip, cells were seeded and a biological neural network cultured in vitro was constructed. A freshly prepared 1% Terg-a-zyme solution was used to clean the chip surface at room temperature for 2 hours. The chip was then rinsed three times with deionized water and thoroughly dried. The dried chip was sterilized in 70% ethanol for 12 to 16 hours, rinsed three times with sterile deionized water, and placed in a 90mm sterile culture dish. A high-humidity chamber was constructed with a 30mm wet dish. 0.6mL of complete culture medium was added to each chip for pre-culture, and the chip was placed in a 37°C, 5% CO2 incubator for two days to improve surface adaptability.
[0090] After the pre-culture, the chip surface was double-coated in sequence. First, 0.1 mg / mL poly-D-lysine (PDL, Sigma-Aldrich, #P6407) was used for incubation at 37°C for 3 hours, then rinsed three times with sterile water and dried; then 0.02 mg / mL laminin was used for incubation at 37°C for 1 hour as the second layer of coating. After incubation, no washing was done and cells were directly inoculated to obtain a biochip.
[0091] In some embodiments of the present invention, during the cell seeding process, 50 μL of neuronal cell suspension is dripped onto the central electrode area of the chip. The suspension concentration is set between 12,000–16,000 cells / μL depending on the cell type. After seeding, the chip is covered with a sterilized cover and placed in a 37°C incubator for 2 hours to promote cell attachment. Subsequently, 0.6 mL of complete culture medium is slowly added along the side of the chip. The culture environment is controlled at 37°C, 5% CO2, and relative humidity greater than 95%.
[0092] After cells have completed attachment, they enter the routine culture phase: starting from the first day, the culture medium is changed three times a week, each time replacing 50% of the original volume. Neural electrical activity is recorded between 4 and 24 hours after the culture medium is changed, and the time interval between the change and the recording is kept consistent as much as possible. As the culture time progresses, neurons gradually grow, differentiate, and form a stable synaptic connection network. An in vitro neural network with bioelectrical activity is formed on the chip, providing a basis for subsequent electrical stimulation, training, and control experiments.
[0093] Specifically, the MaxOne multi-electrode array (MEA) chip of MaxWell Biosystems was used to inoculate neuronal cells and construct a biological neural network cultured in vitro. A freshly prepared 1% Terg-a-zyme solution was used to treat the chip surface at room temperature for 2 hours to clean the chip surface, and then rinsed three times with deionized water and dried thoroughly. The dried chip was sterilized in 70% ethanol for 12 to 16 hours, and then rinsed three times with sterile deionized water, placed in a 90mm sterile culture dish, and a high humidity chamber was constructed with a 30mm wet dish. 0.6mL of complete culture medium was added to each chip for pre-culture, and placed in a 37°C, 5% CO2 incubator for two days to improve surface adaptability. After the pre-culture is completed, the chip surface is double-coated in sequence. The chip was first coated with 0.1 mg / mL poly-D-lysine (PDL, Sigma-Aldrich, #P6407) for 3 hours at 37°C, then rinsed three times with sterile water and dried. A second coating layer was then applied with 0.02 mg / mL laminin (Sigma-Aldrich, #L2020) for 1 hour at 37°C. Cells were seeded directly without rinsing. Cell seeding was performed using the "DotPlating" method, with 50 μL of neuronal cell suspension dripped onto the central electrode area of the chip. The concentration of the suspension ranged from 12,000 to 16,000 cells / μL depending on the cell type. After seeding, the chip was covered with a sterile cover and incubated in a 37°C incubator for 2 hours to promote cell attachment. Subsequently, 0.6 mL of complete culture medium was slowly added along the side of the chip. The culture environment was maintained at 37°C, 5% CO2, and a relative humidity greater than 95%. After cells have adhered, they enter the routine culture phase: starting on day 1, the culture medium is changed three times a week, with 50% of the original volume replaced each time. Neural electrical activity is recorded between 4 and 24 hours after the culture medium is changed, with the interval between changes and recordings being kept consistent as much as possible. As the culture progresses, neurons gradually grow and differentiate, forming a stable network of synaptic connections. This results in an in vitro neural network with bioelectrical activity on the chip, providing the foundation for subsequent electrical stimulation, training, and control experiments.
[0094] In some embodiments of the present invention, in the step of selecting the electrode regions of the biochip based on the firing rates obtained by scanning the channels of the biochip:
[0095] Scan and record the overall activity of the biochip and calculate the spontaneous discharge rate of each channel;
[0096] According to the discharge activity, the first preset number of channels with the highest discharge rate are selected as candidate recording channels, and a two-dimensional spatial discharge heat map is constructed;
[0097] Specifically, the MaxLab Live system was used to scan and record the biochip's overall activity and calculate the spontaneous discharge rate of each electrode. Based on the discharge activity, the top 1024 channels with the highest discharge rates were selected as candidate recording channels, and a two-dimensional spatial discharge heat map was constructed.
[0098] A sliding window-based two-dimensional firing density search method was used to divide the two-dimensional spatial discharge heat map into non-overlapping windows of fixed size, and the average discharge rate of the channel was calculated by traversing each window.
[0099] Specifically, to systematically explore discharge structures, the system introduces a two-dimensional discharge density search method based on a sliding window. The entire electrode array is divided into fixed-size 3×3 non-overlapping windows. The average discharge rate of the electrodes within each window is calculated to estimate the local discharge density.
[0100] All windows are sorted in descending order according to the average discharge rate, and the window area ranked in the first interval is selected as the candidate stimulation area, and two candidate stimulation areas with a distance greater than the first distance are selected as electrode area 1 and electrode area 2 respectively.
[0101] The first distance is 1.0 mm.
[0102] like Figure 4 As shown, further, in order to construct richer input coding, 2 to 5 local high-density areas are screened from the candidate areas as stimulus coding areas according to task requirements, and the specific number can be dynamically set according to the complexity of the task.
[0103] In some embodiments of the present invention, the step of selecting the electrode area of the biochip based on the firing rate scanned from the channel of the biochip also includes selecting electrode area three and electrode area four from the candidate stimulation areas, wherein electrode area three is used to apply stimulation based on a special category of target bodies; and electrode area four is used to apply stimulation based on multiple target bodies.
[0104] In some embodiments of the present invention, the step of selecting the electrode area of the biochip based on the firing rate scanned from the channel of the biochip also includes selecting electrode area five from the candidate stimulation areas, and the electrode area five is used to apply stimulation after the body movement is successful.
[0105] In some embodiments of the present invention, six stimulation modes (divided by region and amplitude setting) are mapped according to the relative position between the target object and the main body in the scene, acting on BNNs and inducing them to produce specific discharge activities; then, the discharge data is collected within the preset response window, and the sorted active channel numbers are extracted as feature vectors; the feature vectors are input into the trained Bagging Tree model for classification and recognition, and the output category labels are converted into behavioral control instructions after mapping to guide the main body to complete the action.
[0106] The control process forms a complete closed-loop neural signal pathway: input encoding → neural response acquisition → feature extraction → pattern recognition → behavioral output. From real-time changes in environmental inputs, to the neural network's internal dynamic responses, to executable control at the system's behavioral level, this forms a closed-loop "perception-computation-execution" flow for the brain-inspired system's response to the external world.
[0107] like Figure 5 and 6 As shown, in some embodiments of the present invention, the method further comprises training the biochip, comprising:
[0108] Symmetrical bidirectional pulses with the same preset width and multiple preset pulse amplitudes are applied to electrode area 1 and electrode area 2 respectively, forming multiple stimulation patterns. Each pattern is repeated at equal time intervals. Multiple stimulation patterns are performed sequentially to form one round of training.
[0109] A perturbation stimulation mechanism is introduced in each round of training. A preset number of random repeated pulses are added between each mode training. The pulse amplitude is within a preset range, the interval is the same, and the stimulation form is a symmetrical biphasic pulse.
[0110] Figure 6 The researchers demonstrated differences in the spontaneous firing activity of neural networks under the two training conditions. Without the addition of perturbation stimuli, the network's spontaneous activity exhibited numerous synchronized network bursts (SNBs), whereby multiple electrode channels discharged densely within similar time periods, exhibiting a highly synchronized activation pattern. However, after the introduction of perturbation stimuli, SNBs were significantly reduced, the firing rhythm became more dispersed, and neuronal activity became less synchronized. This suggests that perturbation training helps break up the synchronous firing circuits formed by the original synaptic connections, making the network dynamically more stable, reducing structural redundancy, and providing a more stable dynamic foundation for subsequent task learning.
[0111] Figure 7 The degree distribution histograms of electrode channels in four training stages are shown. In the untrained stage, the connectivity of most electrodes is low, indicating that the functional connections between neurons have not been fully established and the network structure is relatively sparse. As the number of training rounds increases, the electrode degree shows an overall upward trend, and the number of high-degree nodes gradually increases, reflecting the increasing density and complexity of network connections. This change indicates that external electrical stimulation training promotes the establishment of new functional connections between neurons, and the network structure gradually evolves from loose to integrated. The increase in electrode degree indicates that more neurons are involved in collaborative activities, further supporting the structural plasticity and functional reconstruction capabilities of neural networks driven by training.
[0112] In summary, the present invention constructs a closed-loop control system based on in vitro cultured biological neural networks (BNNs), such as Figure 2 As shown, the system consists of four major functional modules: First, the in vitro biological neural network module: The core computing unit of the system is the BNNs cultured on the MEA. These neurons are derived from the cerebral cortex of pregnant mouse embryos. During the culture process, they spontaneously form complex connectivity structures and possess inherent nonlinear dynamics and network plasticity. Second, the MEA platform module: This module enables high-density signal acquisition and precise electrical stimulation of the BNNs, serving as a bridge between the neural network and control tasks. Third, the recognition algorithm module: This module uses a bagging tree ensemble learning model to classify and identify the aforementioned feature vectors, mapping different neural response patterns into corresponding discrete control instructions. Fourth, the scenario module: This module encodes the relative position information of the target object and the main body in the environment, generates corresponding electrical stimulation patterns, and transmits them to the MEA platform to stimulate the BNNs. Furthermore, the control instructions output by the recognition algorithm module are used to drive the main body to complete movement operations, thereby achieving real-time closed-loop control of the neural network response and behavioral output.
[0113] During the experiment, this scheme set the scene as a simplified flight shooting game. The game interface is divided horizontally into four parallel tracks, and the main aircraft (main body) and enemy aircraft (target body) are restricted to movement within the tracks. Enemy aircraft are randomly generated in the upper area and move forward, without attack capabilities. The main aircraft is located in the lower area and can move left or right in units of tracks based on the recognition results. The control range includes seven possible behaviors: moving left by 3 grids, 2 grids, 1 grid, not moving, and moving right by 1 grid, 2 grids, and 3 grids. This setting simplifies the complex control problem into finite state control, which facilitates the effective mapping of neural network signals.
[0114] The design of the input stimulation is determined by the position of the enemy aircraft relative to the main aircraft. When the enemy aircraft is generated on the left side of the main aircraft, electrical stimulation is applied to "area one" of the BNNs; if the enemy aircraft is located on the right side, "area two" is stimulated. In addition, the stimulation amplitude is further set according to the horizontal distance between the main and enemy aircraft: a distance of 3 grids, 2 grids, and 1 grid corresponds to a symmetrical biphasic electrical stimulation of 500mV, 300mV, and 100mV respectively; if the enemy aircraft coincides with the main aircraft (the distance is 0 grids), no stimulation is applied, indicating that the target has been aligned and no adjustment is required. In this way, 6 input stimulation modes are constructed to fully cover the six types of movement instructions required by the main aircraft.
[0115] An embodiment of the present invention also provides an intelligent control system based on an in vitro cultured biological neuronal network, the system including a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the system implements the steps implemented by the method described above.
[0116] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the aforementioned intelligent control method based on an in vitro cultured biological neural network. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the art.
[0117] It should be understood by those skilled in the art that the various exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is specifically performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.
[0118] It should be understood that the present invention is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, a detailed description of known methods is omitted. In the above embodiments, several specific steps are described and illustrated as examples. However, the method of the present invention is not limited to the specific steps described and illustrated. Those skilled in the art may make various changes, modifications, and additions, or change the order of the steps after understanding the spirit of the present invention.
[0119] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or replace features of other embodiments.
[0120] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.
Claims
1. An intelligent control method based on in vitro cultured biological neuron networks, characterized in that: The steps of the method include: Extract and culture neurons from organisms in vitro, and construct biochips using microelectrode array chips; Scanning the biochip, and selecting an electrode region of the biochip based on a firing rate obtained by scanning a channel of the biochip, wherein the electrode region of the biochip includes an electrode region 1 and an electrode region 2; Acquiring a real-time task screen to be determined, determining a relative distance based on the position of the subject and the target in the real-time task screen to be determined, and selecting electrode region 1 or electrode region 2 for applying stimulation to the biochip and a stimulation intensity based on the relative distance; Acquiring feedback from the biochip to stimulation and constructing it into an initial feedback vector, calculating a significance parameter for each channel in the initial feedback vector, selecting each channel based on the significance parameter of each channel, and constructing a channel vector based on the selected channels; The channel vector is input into a pre-trained classification model, and the action is output by the classification model.
2. The intelligent control method based on in vitro cultured biological neuron network according to claim 1, characterized in that: In the step of obtaining feedback of the biochip to stimulation and constructing it into an initial feedback vector, and calculating a significance parameter for each channel in the initial feedback vector: constructing the channel number of each channel in which a Spike response occurs within a first preset time period after the biochip is stimulated as an initial feedback vector; For each channel in the initial feedback vector, a significance parameter is calculated based on the number of spike responses of the channel in the initial feedback vector and the number of spike responses of the channel in a second preset time period before the channel is stimulated.
3. The intelligent control method based on in vitro cultured biological neuron network according to claim 2, characterized in that: In the step of calculating the significance parameter based on the number of spike responses of the channel in the initial feedback vector and the number of spike responses of the channel in the second preset time period before stimulation, the significance parameter is calculated using the following formula: Among them, R i represents the number of Spike responses of channel i in the initial feedback vector; B i represents the number of Spike responses of channel i in the second preset time period before being stimulated, S i represents the significance parameter of channel i.
4. The intelligent control method based on in vitro cultured biological neuron network according to any one of claims 1 to 3, characterized in that: In the step of selecting each channel based on the significance parameter of each channel and constructing a channel vector based on the selected channel, the significance parameter is compared with a preset threshold to determine the selected channel, and the channel number of the selected channel is used as the value of each dimension in the channel vector to obtain the channel vector.
5. The intelligent control method based on in vitro cultured biological neuron network according to claim 1, characterized in that: The steps of extracting and culturing neurons from an organism in vitro and constructing a biochip using a microelectrode array chip include: The fetal rat cortical tissue was peeled off and cut into small pieces, washed in DMEM culture medium, and transferred to a digestion solution containing 0.125% trypsin and deoxyribonuclease I for digestion at 37°C for 20 minutes. After sufficient dissociation by pipetting, a single-cell suspension was obtained, and the cells were separated by centrifugation at 80 × g for 5 minutes, and the supernatant was removed; The obtained cells were resuspended in complete culture medium containing Neurobasal-Plus medium, B27-Plus, GlutaMax, and penicillin-streptomycin double antibody; Using a multi-electrode array chip, cells were seeded and a biological neural network cultured in vitro was constructed. A freshly prepared 1% Terg-a-zyme solution was used to clean the chip surface at room temperature for 2 hours. The chip was then rinsed three times with deionized water and thoroughly dried. The dried chip was sterilized in 70% ethanol for 12 to 16 hours, rinsed three times with sterile deionized water, placed in a 90mm sterile culture dish, and a high-humidity chamber was constructed with a 30mm wet dish. 0.6mL of complete culture medium was added to each chip for pre-culture, and the chip was placed in a 37°C, 5% CO2 incubator for two days to improve surface adaptability. After the pre-culture is completed, the chip surface is double-coated in sequence. First, 0.1 mg / mL poly-D-lysine is used for incubation at 37°C for 3 hours, and then it is rinsed three times with sterile water and dried. Then, 0.02 mg / mL Laminin is used for incubation at 37°C for 1 hour as the second layer of coating. After incubation, no washing is done and cells are directly inoculated to obtain a biochip.
6. The intelligent control method based on in vitro cultured biological neuron network according to claim 1, characterized in that: In the step of selecting the electrode region of the biochip based on the firing rate obtained by scanning the channels of the biochip: Scan and record the overall activity of the biochip and calculate the spontaneous discharge rate of each channel; According to the discharge activity, the first preset number of channels with the highest discharge rate are selected as candidate recording channels, and a two-dimensional spatial discharge heat map is constructed; A sliding window-based two-dimensional firing density search method was used to divide the two-dimensional spatial discharge heat map into non-overlapping windows of fixed size, and the average discharge rate of the channel was calculated by traversing each window. All windows are sorted in descending order according to the average discharge rate, and the window area ranked in the first interval is selected as the candidate stimulation area, and two candidate stimulation areas with a distance greater than the first distance are selected as electrode area 1 and electrode area 2 respectively.
7. The intelligent control method based on in vitro cultured biological neuron network according to claim 6, characterized in that: The step of selecting the electrode area of the biochip based on the firing rate scanned from the channel of the biochip also includes selecting electrode area three and electrode area four from the candidate stimulation areas, wherein electrode area three is used to apply stimulation based on a special category of target bodies; and electrode area four is used to apply stimulation based on multiple target bodies.
8. The intelligent control method based on in vitro cultured biological neuron network according to claim 6, characterized in that: The step of selecting the electrode area of the biochip based on the firing rate obtained by scanning the channels of the biochip also includes selecting electrode area five from the candidate stimulation areas, and the electrode area five is used to apply stimulation after the body movement is successful.
9. The intelligent control method based on in vitro cultured biological neuron network according to claim 1, characterized in that: The method further includes training the biochip, the steps comprising: Symmetrical bidirectional pulses with the same preset width and multiple preset pulse amplitudes are applied to electrode area 1 and electrode area 2 respectively, forming multiple stimulation patterns. Each pattern is repeated at equal time intervals. Multiple stimulation patterns are performed sequentially to form one round of training. A perturbation stimulation mechanism is introduced in each round of training. A preset number of random repeated pulses are added between each mode training. The pulse amplitude is within a preset range, the interval is the same, and the stimulation form is a symmetrical biphasic pulse.
10. An intelligent control system based on in vitro cultured biological neuron networks, characterized in that: The system includes a computer device, which includes a processor and a memory. The memory stores computer instructions. The processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the system implements the steps implemented by the method according to any one of claims 1 to 9.
Citation Information
Patent Citations
Neural microphysiological systems and methods of using the same
CN107106606A
Cognitive computing methods and systems based on biological neural networks
CN112673382A
Regulation and control method, device and equipment for in-vitro culture of neuron population and storage medium
CN116611472A
Closed-loop training device and method for controlling dolly to avoid obstacles by using in-vitro neural chip
CN119558362A
Bioelectrical stimulation presentation method and electrical stimulation device
JP2017209322A
