Method and system for stably monitoring action potential in cells based on graphene microelectrode array

Abstract

The invention discloses an intracellular action potential stable monitoring method and system based on a graphene microelectrode array. The method comprises the following steps: recording a cell external field potential signal through a graphene microelectrode; preprocessing the extracellular field potential signal, and inputting the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is used for representing the relation between the extracellular field potential signal and the intracellular action potential signal; and on the basis of the continuous intracellular action potential signals, continuously monitoring the intracellular action potential. The intracellular action potential signal is predicted through the extracellular field potential signal, so that the problem of cell damage caused by intracellular records is relieved, stable monitoring is realized, and the monitoring accuracy is improved. The method can be widely applied to the technical field of computers.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and system for monitoring intracellular action potential stability based on a graphene microelectrode array. Background Art

[0002] Recording intracellular action potentials in cardiomyocytes is of great significance for studying cardiac electrophysiology, cardiomyopathy, and drug development. Patch clamp electrodes are the gold standard for recording intracellular action potentials. However, patch clamp techniques rely on precise instrumentation and operator experience, and require gigaohm-level seals during recording, which can easily lead to mechanical damage or perforation of the cell membrane, compromising cell viability and long-term recording stability. Fluorescence signals from voltage-sensitive dyes / proteins typically exhibit weak fluctuations and low signal-to-noise ratios. Prolonged high-frequency laser excitation can easily cause photobleaching and damage to cell viability, limiting the duration of continuous monitoring. Cells typically die within a few hours after intracellular recordings. Using vertical nanoelectrode arrays combined with perforation methods relies on the electrode tip penetrating the cell membrane, potentially disrupting membrane integrity, leading to cellular ion leakage or apoptosis, particularly affecting fragile cells such as cardiomyocytes. Electroporation and photoporation methods based on short pulses of voltage or light typically induce reversible perforation of the cell membrane. The pores on the cell membrane heal within a short healing period (~10 minutes), gradually converting intracellular signals into extracellular signals. This limits long-term detection of intracellular action potentials. Summary of the Invention

[0003] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0004] To this end, the object of the present invention is to provide an efficient method and system for monitoring the stability of intracellular action potential based on graphene microelectrode arrays.

[0005] In order to achieve the above technical objectives, one aspect of an embodiment of the present invention provides a method for stable monitoring of intracellular action potentials based on a graphene microelectrode array, comprising the following steps: recording an extracellular field potential signal through a graphene microelectrode; preprocessing the extracellular field potential signal, and inputting the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is used to characterize the connection between the extracellular field potential signal and the intracellular action potential signal; based on the continuous intracellular action potential signal, the intracellular action potential is continuously monitored. The present application predicts the intracellular action potential signal through the extracellular field potential signal, which is conducive to alleviating the cell damage problem caused by intracellular recording, achieving stable monitoring, and helping to improve monitoring accuracy.

[0006] In some embodiments, the signal prediction model is trained by the following steps:

[0007] Record extracellular field potential signal samples and intracellular action potential signal samples through graphene microelectrode electroporation;

[0008] Classifying, downsampling, segmenting, and normalizing the extracellular field potential signal samples, and inputting the processed extracellular field potential signal samples into the signal prediction model to obtain predicted intracellular action potential signal results;

[0009] A loss function is constructed according to the intracellular action potential signal result and the intracellular action potential signal sample, and the signal prediction model is trained according to the loss function to obtain a trained signal prediction model.

[0010] In some embodiments, the signal prediction model is a multilayer perceptron neural network, which includes an input layer, a hidden layer, and an output layer; the signal prediction model is trained by the following steps:

[0011] Determining a first input signal sample and a first output signal sample of the hidden layer according to a first weight between neurons of the input layer and neurons of the hidden layer;

[0012] determining a second input signal sample and a second output signal sample of the output layer according to a second weight between neurons of the hidden layer and neurons of the output layer;

[0013] Based on the stochastic gradient descent method, the neuron threshold, the first weight and the second weight are iteratively updated until the loss value reaches a preset loss, thereby obtaining a trained signal prediction model; wherein the preset loss is related to the set learning rate.

[0014] In some embodiments, the loss function is established by the following steps:

[0015] Determine the loss function through the mean square error;

[0016] Alternatively, determine a loss function based on the relationship between the total variability, the variability explained by the regression model, and the residual variability.

[0017] In some embodiments, recording extracellular field potential signals by graphene microelectrodes comprises:

[0018] Taking a substrate, and depositing a metal layer on the substrate;

[0019] Placing the substrate in a vacuum chamber, heating it to a preset temperature, and introducing hydrogen to pre-treat the surface of the substrate by plasma;

[0020] introducing hydrogen and methane to generate graphene on the surface of the substrate at a preset plasma power;

[0021] Cooling the graphene and metal layers, etching the graphene and metal layers based on a preset microelectrode array pattern, cutting the graphene and metal layers, and coating the non-recording area with PDMS to obtain a graphene microelectrode array chip.

[0022] The graphene microelectrode array chip is fixed to a PCB to record extracellular field potential signals.

[0023] In some embodiments, the method comprises:

[0024] Display signal recording interface;

[0025] In response to a trigger operation on a first control received on the signal recording interface, recording a cell signal based on a current recording mode; wherein the cell signal includes an extracellular field potential signal sample or an intracellular action potential signal sample, and the first control is used to start signal recording;

[0026] The cell signals are marked according to preset rules, and the cell signals are grouped and displayed, and the cell signals marked as normal are stored.

[0027] In some embodiments, the method further comprises:

[0028] Displaying the grouping results of cell signals on the signal recording interface;

[0029] In the signal recording interface, in response to a triggering operation on a first signal, a group of the first signal is adjusted.

[0030] On the other hand, an embodiment of the present invention provides a system for monitoring intracellular action potential stability based on a graphene microelectrode array, comprising:

[0031] The first module is used to record extracellular field potential signals through graphene microelectrodes;

[0032] A second module is configured to preprocess the extracellular field potential signal and input the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is configured to characterize the relationship between the extracellular field potential signal and the intracellular action potential signal;

[0033] The third module is used to continuously monitor the intracellular action potential based on the continuous intracellular action potential signal.

[0034] On the other hand, an embodiment of the present invention provides a device for monitoring intracellular action potential stability based on a graphene microelectrode array, comprising:

[0035] at least one processor;

[0036] at least one memory for storing at least one program;

[0037] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned method for monitoring intracellular action potential stability based on graphene microelectrode array.

[0038] On the other hand, an embodiment of the present invention provides a storage medium storing a program executable by a processor, which is used to implement the above-mentioned method for monitoring intracellular action potential stability based on a graphene microelectrode array when executed by the processor.

[0039] The embodiments of the present application include at least the following beneficial effects: the method provided by the embodiments of the present invention includes: recording an extracellular field potential signal through a graphene microelectrode; preprocessing the extracellular field potential signal, and inputting the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is used to characterize the connection between the extracellular field potential signal and the intracellular action potential signal; based on the continuous intracellular action potential signal, the intracellular action potential is continuously monitored. The present application predicts the intracellular action potential signal through the extracellular field potential signal, which is conducive to alleviating the cell damage problem caused by intracellular recording, achieving stable monitoring, and helping to improve monitoring accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following introduction is made to the drawings of the embodiments of the present invention or the related technical solutions in the prior art. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0041] Figure 1 A schematic flow chart of an embodiment of a method for monitoring intracellular action potential stability based on a graphene microelectrode array provided by the present invention;

[0042] Figure 2 A schematic structural diagram of an embodiment of a graphene microelectrode array provided by the present invention;

[0043] Figure 3 A schematic diagram of a graphene electrode pattern according to an embodiment of the present invention;

[0044] Figure 4 A schematic diagram of an interface of an embodiment of a signal recording interface provided by the present invention;

[0045] Figure 5 A schematic diagram of another embodiment of the signal recording interface provided by the present invention;

[0046] Figure 6 A signal diagram of an embodiment of extracellular and intracellular electrophysiological signals provided by the present invention;

[0047] Figure 7 A signal schematic diagram of an embodiment of the present invention for continuously recording extracellular and intracellular electrophysiological signals using a graphene microelectrode array;

[0048] Figure 8 A schematic diagram comparing the extracellular and intracellular electrophysiological signals continuously recorded by the graphene microelectrode array provided by the present invention with the intracellular action potential signals predicted by the signal prediction model;

[0049] Figure 9 This is a schematic diagram of the loss value curve of the intracellular action potential signal derived from the signal prediction model provided by the present invention. DETAILED DESCRIPTION

[0050] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention. The step numbers in the following embodiments are provided for ease of explanation only and do not limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0051] Recording the intracellular action potential of cardiomyocytes is of great significance for studying cardiac electrophysiology, cardiomyopathy, and drug development. Patch clamp electrodes are the gold standard for recording intracellular action potentials. They are made by drawing borosilicate glass capillaries into conical microtubes. The electrode is filled with an electrolyte solution with a composition similar to that of the intracellular fluid. The tail of the electrode is connected to the patch clamp amplifier through a silver wire or a chloride-plated silver wire. During recording, the tip of the electrode is adsorbed on the surface of the cell membrane by negative pressure (20-30cm H2O), forming a gigaohm (10-100GΩ) high-resistance seal (Giga-seal), isolating an area of ​​about 1μm. 2 The patch of membrane. Breaking the patch clamp by negative pressure can record intracellular action potential (AP) with high signal-to-noise ratio, time resolution and signal fidelity. In recent years, voltage-sensitive dyes and calcium indicators have become methods for recording action potentials. Voltage-sensitive dyes / proteins (such as Di-4-ANEPPS, GCaMP) reflect membrane potential or Ca through fluorescence changes. 2+The voltage dye is directly embedded in the cell membrane, and the fluorescence intensity changes with the membrane potential (ΔF / F0 is about 0.1-1%). Combined with high-speed imaging (kHz level), fast signals can be captured. 2+ Indicators (such as Fluo-4 and GCaMP6) rely on action potential-triggered Ca 2+ Influx of cells results in a greater fluorescence enhancement (ΔF / F0 approximately 5-10%), but with lower temporal resolution (10-100 Hz). Both methods are recorded using optical systems (e.g., laser excitation, sCMOS camera). The voltage method is suitable for high-resolution single-cell dynamics. The calcium method is suitable for in vivo monitoring, but the detection process requires a trade-off between directness, resolution, and invasiveness. Vertical nanoelectrode arrays combined with perforation methods have become a cutting-edge technology for high-throughput, high-temporal and high-resolution intracellular electrophysiological recordings. Vertical nanoelectrode arrays typically use silicon, glass, or flexible polymers (e.g., polyimide) as a substrate. Metals (e.g., gold, platinum) or semiconductors (e.g., silicon) are then deposited on the substrate surface, and finally, the electrode surface is covered with an insulating material. Neurons or cardiomyocytes are cultured on the surface of the vertical nanoelectrode array, and the cell membrane either adheres naturally or is forced into contact with the electrode by centrifugal force or fluid pressure. The electrode relies on the nanometer-scale sharpness of the electrode tip to directly penetrate the cell membrane or temporarily open the cell membrane by applying a short pulse of voltage and light.

[0052] Patch clamp techniques rely on precision instruments and operator experience, and require gigaohm-level seals during recording, which can easily lead to mechanical damage or perforation of the cell membrane, compromising cell activity and long-term recording stability. Fluorescence signals from voltage-sensitive dyes / proteins typically exhibit weak fluctuations and low signal-to-noise ratios. Prolonged high-frequency laser excitation can easily induce photobleaching and damage to cell activity, limiting the duration of continuous monitoring. Cells typically die within a few hours after intracellular recordings. Vertical nanoelectrode arrays combined with perforation methods rely on the electrode tip penetrating the cell membrane, potentially disrupting membrane integrity, leading to cellular ion leakage or apoptosis, particularly affecting fragile cells such as cardiomyocytes. Electroporation and photoporation methods based on short pulse voltage or light typically induce reversible perforation of the cell membrane. The pores on the cell membrane gradually convert intracellular signals to extracellular signals within a short healing period (~10 minutes), limiting the long-term detection of intracellular action potentials. Therefore, a method is needed to improve the long-term detection capability of nanoelectrode arrays for intracellular signals.

[0053] The following describes in detail the method and system for monitoring the stability of intracellular action potential based on a graphene microelectrode array proposed in an embodiment of the present invention with reference to the accompanying drawings. First, the method for monitoring the stability of intracellular action potential based on a graphene microelectrode array proposed in an embodiment of the present invention will be described with reference to the accompanying drawings.

[0054] In an embodiment of the present invention, a method for monitoring the stability of intracellular action potential based on a graphene microelectrode array is provided. The method for monitoring the stability of intracellular action potential based on a graphene microelectrode array in an embodiment of the present invention can be applied to a terminal, a server, or software running in a terminal or a server. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The method for monitoring the stability of intracellular action potential based on a graphene microelectrode array in an embodiment of the present invention mainly includes the following steps:

[0055] S100: Recording extracellular field potential signals via graphene microelectrodes;

[0056] S200: preprocessing the extracellular field potential signal, and inputting the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is used to characterize the relationship between the extracellular field potential signal and the intracellular action potential signal;

[0057] S300: Continuously monitoring the intracellular action potential based on the continuous intracellular action potential signal.

[0058] The present invention proposes a multilayer perceptron neural network (MLP) based on the error back propagation (BP) algorithm to enhance the stability of graphene microelectrode array (MEA) monitoring intracellular action potential ( Figure 1 This application firstly uses a 32-channel graphene MEA electroporation method to efficiently record high-quality extracellular and intracellular signal data sets ( Figure 2 ). The present application then classifies, downsamples, segments, and normalizes the signals in the dataset for training a multilayer perceptron neural network. The multilayer perceptron neural network consists of an input layer, a hidden layer, and an output layer. The trained neural network is able to establish a connection between extracellular and intracellular signals. When the trained multilayer perceptron neural network is input with the extracellular signals recorded by the graphene MEA, the intracellular signals can be derived, thereby achieving long-term and stable monitoring of intracellular APs by the graphene MEA.

[0059] In some embodiments, the signal prediction model is trained by the following steps:

[0060] Record extracellular field potential signal samples and intracellular action potential signal samples through graphene microelectrode electroporation;

[0061] Classifying, downsampling, segmenting, and normalizing the extracellular field potential signal samples, and inputting the processed extracellular field potential signal samples into the signal prediction model to obtain predicted intracellular action potential signal results;

[0062] A loss function is constructed according to the intracellular action potential signal result and the intracellular action potential signal sample, and the signal prediction model is trained according to the loss function to obtain a trained signal prediction model.

[0063] It is understood that the intracellular action potential signal samples in this application correspond to the extracellular field potential signal samples, and the intracellular action potential signal samples can be regarded as the real results corresponding to the extracellular field potential signal samples. The intracellular action potential signal samples in this application can be recorded by temporarily perforating the cells using electroporation technology.

[0064] In some embodiments, the signal prediction model is a multilayer perceptron neural network, which includes an input layer, a hidden layer, and an output layer; the signal prediction model is trained by the following steps:

[0065] Determining a first input signal sample and a first output signal sample of the hidden layer according to a first weight between neurons of the input layer and neurons of the hidden layer;

[0066] determining a second input signal sample and a second output signal sample of the output layer according to a second weight between neurons of the hidden layer and neurons of the output layer;

[0067] Based on the stochastic gradient descent method, the neuron threshold, the first weight and the second weight are iteratively updated until the loss value reaches a preset loss, thereby obtaining a trained signal prediction model; wherein the preset loss is related to the set learning rate.

[0068] Reference Figure 1 As shown in Figure 1, the multilayer perceptron neural network model architecture consists of three hidden layers, each containing 1024 neurons, and is optimized using the stochastic gradient descent (SGD) algorithm. This method updates weights and biases by calculating gradients using random samples during each iteration. To speed up training, the learning rate is set to 0.9. Once the loss value reaches 3×10 -3 , training was terminated to prevent overfitting. The dataset was randomly split into a training subset (1219 samples, ~90%) and a test subset (126 samples, ~10%). The performance of the model was tested after reaching the expected loss value.

[0069] In some embodiments, the loss function is established by the following steps:

[0070] Determine the loss function through the mean square error;

[0071] Alternatively, determine a loss function based on the relationship between the total variability, the variability explained by the regression model, and the residual variability.

[0072] In some embodiments, recording extracellular field potential signals by graphene microelectrodes comprises:

[0073] Taking a substrate, and depositing a metal layer on the substrate;

[0074] Placing the substrate in a vacuum chamber, heating it to a preset temperature, and introducing hydrogen to pre-treat the surface of the substrate by plasma;

[0075] introducing hydrogen and methane to generate graphene on the surface of the substrate at a preset plasma power;

[0076] Cooling the graphene and metal layers, etching the graphene and metal layers based on a preset microelectrode array pattern, cutting the graphene and metal layers, and coating the non-recording area with PDMS to obtain a graphene microelectrode array chip.

[0077] The graphene microelectrode array chip is fixed to a PCB to record extracellular field potential signals.

[0078] See also Figure 2 The relationship between graphene and substrate shown, and Figure 3 Preset microelectrode array pattern shown.

[0079] In some embodiments, the method comprises:

[0080] Display signal recording interface;

[0081] In response to a trigger operation on a first control received on the signal recording interface, recording a cell signal based on a current recording mode; wherein the cell signal includes an extracellular field potential signal sample or an intracellular action potential signal sample, and the first control is used to start signal recording;

[0082] The cell signals are marked according to preset rules, and the cell signals are grouped and displayed, and the cell signals marked as normal are stored.

[0083] See also Figure 4 and Figure 5 In the signal recording interface of an embodiment shown, the first control may be Figure 5In the start control, before signal recording, select the recording mode and file storage path, and trigger the first control to execute signal acquisition and recording. It is understandable that the trigger operation in this application can be a click operation, a drag operation, a touch operation, a selection operation, etc., and this application does not limit the specific form of the trigger operation. The preset rules in this application are used to determine the quality of the collected signal. The preset rules can be to compare the preset parameters of the collected signal with the corresponding preset parameter thresholds to determine the labeling results of the cell signal, and then group the cell signals according to the labeling results. The preset parameters can be the signal-to-noise ratio of the signal, whether the signal has a clear peak, etc. This application can also adjust the preset rules, and users can set the preset rules according to actual needs. The cell signal in this application can be an extracellular field potential signal sample or an intracellular action potential signal sample. The quality of the signal sample is controlled through the signal recording interface, and the signal prediction model is then trained using the adjusted signal samples to improve the accuracy of the model training. On the other hand, the cell signal in this application can also be an extracellular field potential signal. The signal quality of the collected extracellular field potential signal is controlled to improve the prediction accuracy of the intracellular action potential signal.

[0084] In some embodiments, the method further comprises:

[0085] Displaying the grouping results of cell signals on the signal recording interface;

[0086] In the signal recording interface, in response to a triggering operation on a first signal, a group of the first signal is adjusted.

[0087] Reference Figure 4 As shown, if there is a deviation in the system's classification of cell signals, the user can adjust the group of the first signal through a trigger operation.

[0088] The following specific examples are used to describe in detail the method for monitoring intracellular action potential stability provided by the present application:

[0089] This application develops a machine learning method based on a multilayer perceptron (MLP) model to derive intracellular signals from extracellular recordings of graphene MEA. The graphene MEA-coupled MLP (graphene MEA-MLP) can achieve long-term monitoring of intracellular APs.

[0090] The multilayer perceptron (MLP) model derived over 1,200 sets of extracellular signals into intracellular signals, and the derived results were strongly correlated with the recorded intracellular APs, with a high degree of overlap. The graphene MEA-MLP was able to record intracellular action potentials for up to nine days. Furthermore, the graphene MEA-MLP was able to accurately and noninvasively monitor the induction effects of drugs (quinidine and E-4031) on cardiomyocytes.

[0091] It should be noted that on the basis of graphene MEA, carbon nanotube (CNT) or gold nanowire composite structure can also be introduced to improve conductivity and signal-to-noise ratio (SNR). This application can also use polyimide (PI) or PDMS flexible substrate to adapt to the dynamic deformation of living tissue and reduce signal drift caused by mechanical stress. This application can also add pH sensor, temperature sensor or oxygen partial pressure sensor to monitor the cell microenvironment in real time and correct signal deviation caused by metabolic changes.

[0092] First, the accompanying drawings related to this application are introduced. Figure 1 Schematic diagram of the graphene MEA combined with the MLP model. Extracellular signals recorded by the graphene MEA and intracellular signals recorded by electroporation were normalized and used as training input. The MLP model was constructed using the back-propagation (BP) algorithm. The model was optimized using the mean square error (MSE) as the loss function. Figure 2 (A) is a schematic diagram of a graphene MEA. The bottom of the graphene is tightly connected, ensuring a stable electrical pathway and providing a large contact area for cardiomyocytes. Figure 2 (B) is a schematic diagram of a cardiomyocyte coupled to a single microelectrode. The graphene microelectrode records the extracellular field potential (FP) and intracellular AP signals before and after electroporation. Figure 3 (A) CAD design for laser-etched graphene electrode patterns. Dimensions in the figure are in μm. Figure 3 (B) Optical image of a single graphene electrode array obtained after cutting. Figure 6 To continuously record extracellular and intracellular electrophysiological signals on 32 channels of the graphene MEA chip. Figure 7 (A, B) Typical extracellular FP signals and intracellular AP signals of cardiomyocytes recorded by graphene MEA over 5-9 days. Figure 7 Middle (C, D) are superpositions of single contraction-relaxation processes in extracellular FP and AP of cardiomyocytes recorded by the graphene MEA containing 32 channels over 5-9 days. Figure 8 Middle (A) Representative results showing extracellular FPs recorded by graphene MEA (blue), intracellular APs derived from the graphene MEA-MLP model (red), and intracellular APs recorded by graphene MEA electroporation (black). Figure 8 Middle (B) Derivation of intracellular APs based on 5 consecutive days of measurement using the graphene MEA-MLP model. Representative extracellular FPs recorded by the graphene MEA (blue), intracellular APs recorded by electroporation of the graphene MEA (black), and intracellular APs derived from the graphene MEA-MLP model (red). Figure 9 (A) MSE and R of intracellular AP waveforms derived from the MLP model2 High R 2 and low MSE values ​​indicate high accuracy of the derived waveform. Figure 9 Middle (BD) Depolarization time, repolarization time, and APD of intracellular APs recorded by graphene MEA electroporation and derived from the graphene MEA-MLP model. 10-90 quantification. Figure 9 Middle (E) shows the intracellular AP waveforms recorded by graphene MEA electroporation after treatment with different quinidines (0, 10, 20, and 50 μM), as well as representative waveforms derived from graphene MEA-MLP modeling. Figure 9 Medium (FH) APD 50 , depolarization time, and repolarization time change rate analysis. After treatment with quinidine. Figure 9 Middle (I) Intracellular AP waveforms recorded by graphene MEA electroporation after treatment with E-4031 (0, 10, 50, and 100 nM), as well as representative waveforms from graphene MEA-MLP modeling. Figure 9 Middle (J) is APD 50 Rate of change analysis. Figure 9 (K) is the depolarization time. Figure 9 Middle (L) is the repolarization time after E-4031 treatment.

[0093] First, this application provides graphene microelectrode array fabrication.

[0094] In this work, a 4-inch quartz glass substrate was used as the base material. Initially, the quartz glass substrate was cleaned with anhydrous ethanol. A 30 nm Au layer (Au purity 99.999%) was then uniformly deposited on the substrate surface using a magnetron sputtering device. The Au-coated quartz glass substrate was then transferred to a radio frequency microwave plasma enhanced chemical vapor deposition device. The vacuum chamber was heated to 950°C at a pressure of 25 Pa, and hydrogen (H2) was introduced at a flow rate of 20 sccm. The substrate surface was pretreated using 1500W plasma for 15 minutes. Subsequently, hydrogen and methane (CH4) were introduced into the vacuum chamber at flow rates of 10 sccm and 20 sccm, respectively. Graphene was then grown on the substrate surface at a plasma power of 1500W for 20 minutes. After cooling, the excess graphene and Au connecting layer ( Figure 3 A). The 4-inch square glass substrate was then cut into 25 uninsulated 2×2 cm graphene MEA chips ( Figure 3(B) The non-recording surface of the graphene microelectrode was coated with a layer of uncured PDMS (Dow Corning Sylgard 184, PDMS:curing agent = 10:1) as an insulating layer and heat-cured at 80°C for 2 hours to obtain the final 32-channel graphene MEA chip. During assembly, PDMS was used to secure the graphene MEA chip to a custom PCB. Conductive silver adhesive (Electrolube, Ashby de la Zouch, UK) was used to connect the 32 microelectrodes to the PCB interface. A glass ring with an inner diameter of 1 cm and a height of 1.5 cm was then attached to the center of the device using PDMS to serve as a cell culture chamber. A reference electrode was secured to the plastic cover and electrically connected to the PCB. Finally, a four-row pin header was soldered onto the PCB to match the interface of the electrophysiological signal acquisition system. This graphene MEA device utilizes the same reference electrode for electrophysiological signal recording and nanoelectroporation.

[0095] Second, the recording of extracellular FP and intracellular AP electrophysiological signal data sets.

[0096] After the myocardial cells contracted synchronously, electrophysiological signals were collected from the myocardial cells using a homemade electrophysiological signal acquisition system ( Figure 6 ). The electrophysiological signal acquisition system consists of the following main components: a signal processing module, a customized data collector, an electroporation module, and a graphical user interface (computer software) for data visualization. For primary and secondary amplification, this application uses OPA2140 and OPA227 operational amplifiers (Texas Instruments). The OPA2140 combines low noise, stable unity gain, and low power consumption, while the OPA227 has the characteristics of low noise, high bandwidth, high precision, and compact size. Both amplifiers support ±5V power supply and operate stably in an environment of 37°C, ensuring the reliable performance of the experimental device of this application. The data recorded by the 32-channel graphene MEA is visualized and stored in real time by an application developed based on LabVIEW 2018 software. During the visualization process, signals (such as high-quality waveforms and low-quality waveforms) are marked according to preset rules (such as noise and amplitude) to generate labeled data sets for model training. Waveforms with extremely poor signal quality may not be stored. By controlling data and quality at the signal source stage, the efficiency of obtaining high-quality data sets can be improved. Electrophysiological signals were sampled at 20 kHz and filtered using a 1 Hz to 7.5 kHz bandpass filter. The electrophysiological signal acquisition system was used to record extracellular FPs. For intracellular AP recording, the graphene MEA device applied 3 V, 20 Hz, 200 μs square wave pulses for a total duration of 1 second. The electrophysiological signal acquisition system was used to record intracellular APs.

[0097] Third, preprocessing of training set.

[0098] After obtaining the extracellular FP and intracellular AP electrophysiological signal data sets, the extracellular and intracellular signals with low noise and few miscellaneous peaks were screened out according to the noise level and the number of miscellaneous peaks ( Figure 7 ). The extracellular signals and intracellular signals in each group of extracellular-intracellular signals come from the same channel of a certain experiment, while different groups of extracellular-intracellular signals come from different experiments (normal culture, E-4031 and quinidine treatment). The length of each segment of the extracellular and intracellular electrophysiological signals has the same ratio, and the number of signal points is made to 6000 points by downsampling or interpolation calculation. After data extraction, this application uses a standardized method to normalize each group of data. According to the above conditions, 2500 groups of extracellular-intracellular signal data were screened out, and 2000 groups of data were randomly selected as training sets for training the network, and the remaining 500 groups were used as test sets for testing the network. During the training process, the extracellular signal is the input information of the network, and the intracellular signal is the learning target of the network.

[0099] Fourth, intracellular signal error back propagation neural network.

[0100] BP neural network consists of input layer, hidden layer and output layer. Next, we take the three-layer neural network as an example. Assume that the training set D = {(X1, Y1), (X2, Y2), ..., (X p ,Y p )},X i ∈R l , Y i ∈R n Indicates that the network input parameters have l characteristic attributes and output n-dimensional column vector. For the convenience of discussion, assume that the network input layer has l neurons, the hidden layer has m neurons, the output layer has n neurons, w is the weight between neurons, θ j is the threshold of the hidden layer neurons, θ k is the threshold of the output layer neurons, the activation function is the Sigmoid function, and its expression is Eq.

[0101]

[0102] The input of the jth neuron in the hidden layer is:

[0103]

[0104] where w ij is the weight between the i-th neuron in the input layer and the j-th neuron in the hidden layer, x i is the i-th feature attribute of the input layer.

[0105] The input of the kth neuron in the output layer is:

[0106]

[0107] where w jk is the weight between the jth neuron in the hidden layer and the kth neuron in the output layer, b j is the output of the jth neuron in the hidden layer.

[0108] For the training sample (x q ,y q ), the output layer of the network is Similarly, the hidden layer output can be obtained in:

[0109]

[0110] When the training sample (x q ,y q ) At the first iteration, the mean square error on the network is:

[0111]

[0112] The BP neural network is based on the gradient descent method, which updates the parameters in the negative gradient direction of the target. For the error E of formula (5), q , when the learning rate is lr, we can get:

[0113]

[0114] By the chain rule we can get:

[0115]

[0116] From formula (3), we can get:

[0117]

[0118] The derivative of the Sigmoid function is:

[0119] f ′ (x)=f(x)(1-f(x)) (9)

[0120] From formula (4) and (5), we can get:

[0121]

[0122] Substituting equations (7), (8) and (10) into equation (6), we obtain:

[0123] Δw jk =lr·g k b j (11)

[0124] Similarly, the threshold update amount of the output layer neurons is:

[0125] Δθk =-lr·g k (12)

[0126] Similarly, the weight update formula between the input layer and the hidden layer and the threshold update formula of the hidden layer neurons are:

[0127] Δw ij =lr·e j ·x i (13)

[0128] Δθ j =lr·e j (14)

[0129] In formulas (13) and (14), e j The calculation method of g is the same as that of g in formula (10) k Similar, but e j The quantitative relationship in is between the input layer and the hidden layer.

[0130] The learning rate, lr, determines the magnitude of weight and threshold updates during training. Excessively large learning rates can prevent the network from converging, while excessively small ones can reduce training efficiency and slow network convergence. Therefore, an appropriate learning rate should be selected based on the actual situation. In the above reasoning process, the goal of the BP algorithm is to minimize the average error of all samples in the training set D, that is, to minimize the value of Equation (15).

[0131]

[0132] Fifth, evaluation of neural network training performance.

[0133] The extracted data was split into a ratio of approximately 9:1, with 1218 randomly selected as the training set, and the remaining 127 datasets used as the test set. To quantitatively evaluate the performance of the neural network, MSE and R2 were selected as evaluation metrics. MSE measures the error by calculating the average of the squares of the differences between the predicted and actual values. Its expression is as follows:

[0134]

[0135] Where n is the total number of data points, is the yth i The actual recorded intracellular action potential value. is the fourth predicted intracellular action potential value. R 2 It reflects the degree of fit between the model and the data, which is calculated by the relationship between the total variability, the variability explained by the regression model, and the residual variability. Its expression is as follows:

[0136]

[0137] in, is the arithmetic mean of the AP values ​​actually recorded in cells i.

[0138] Sixth, intracellular AP prediction.

[0139] First, the model was initialized, that is, the neural network was allowed to load the trained weight file. The extracellular signal data of the control group (normal culture) and drug group (E-4031 and quinidine) were extracted in the same way as the pre-training group and the experimental group, and normalized ( Figure 8 The present application inputs the processed extracellular signal into the model to predict the intracellular signal and calculates the depolarization time, repolarization time and APD10-90 of the predicted intracellular signal ( Figure 9 )

[0140] Regarding the isolation and culture of neonatal rat cardiomyocytes:

[0141] All animal experiments were performed in accordance with the relevant animal experimental procedures. Neonatal rat ventricular myocytes (NRVMs) were isolated from the hearts of Sprague-Dawley rats aged 1-3 days after birth. The specific procedures were as follows: the hearts were removed from the neonatal rats, placed in pre-chilled phosphate-buffered saline, and rinsed, and blood clots and other non-cardiac tissues were carefully removed. The hearts were then transferred to D-Hanks' solution and cut with ophthalmic scissors. Then, digestion was repeated in D-Hanks' solution containing 0.07% trypsin (Gibco) and 0.05% type II collagenase (Gibco) until the tissue was completely digested. Before each use, the graphene MEA device was disinfected with 75% (v / v) ethanol and irradiated under UV light overnight. Then, the microelectrode surface was etched with oxygen plasma (Drytek) for 120 seconds and incubated with 50 μg / ml fibronectin solution for 4 hours to promote cell attachment. Isolated cardiomyocytes were suspended in 5 ml of culture medium and seeded onto the graphene MEA at a density of approximately 1.5 × 10⁴ cells / mm⁻. The culture medium consisted of Dulbecco's Modified Eagle Medium (DMEM, Vivacell) supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin / streptomycin (Gibco). The cells were cultured in a 37°C, 5% CO₂ incubator, with the culture medium replaced every two days.

[0142] Regarding drug treatment of cardiomyocytes:

[0143] Cardiac electrophysiology can be modulated by drugs. Drugs act as ion channel blockers, affecting the influx and / or backflux of ions across the cell membrane, thereby altering cellular electrophysiological properties. Cardiac AP waveforms reflect ion channel activity. In this experiment, cells were treated with E-4031 (MCE, No. 113559-13-0, Chinese Academy of Sciences) and quinidine (McLean, Chinese Academy of Sciences; CAS 56-54-2). Compound E-4031 is a Class III antiarrhythmic drug that prolongs cardiac APs by blocking potassium channels. Quinidine, a class I antiarrhythmic drug, slows the influx of sodium ions by blocking sodium channels, thereby reducing the rate of cardiomyocyte depolarization. Both drugs (E-4031 and quinidine) were initially dissolved in dimethyl sulfoxide (DMSO) to the desired concentrations. Before drug treatment, untreated cardiomyocytes were electroporated using a graphene MEA, and AP waveforms were recorded as blank controls. For drug treatment, cardiomyocytes were treated with increasing doses of 10, 20, and 50 μM quinidine and 10, 50, and 100 nM E-4031. Five minutes after drug injection, AP recordings were observed to be affected. After recordings were completed, the drug solution was replaced with a typical complete medium solution to restore the cells' normal electrophysiological properties. After treatment with different drug concentrations, cells were allowed to recover in complete medium for at least 30 minutes before further experiments. Experiments with different drug concentrations were completed within 24 hours.

[0144] Implementation results:

[0145] Graphene MEA-MLP model derives intracellular AP waveforms from extracellular FP signals

[0146] This application has developed a graphene MEA system that exhibits excellent efficiency and stability in high-quality recording of extracellular and intracellular signals. This system facilitates a large number of experiments and enables the collection of large data sets. Leveraging these capabilities, this application has expanded the research of this application by culturing cells in different states on the graphene MEA platform and obtained various types of extracellular and intracellular electrical signal data sets through microelectrode electroporation. This application has developed a machine learning method based on a multilayer perceptron (MLP) model to infer intracellular signals from extracellular recordings, thereby enabling indirect and long-term monitoring of intracellular APs ( Figure 1 This application established a collection of 1,345 extracellular / intracellular data sets as training and testing data sets for the MLP machine learning model to establish the correlation between highly correlated intracellular and extracellular signals. The extracellular signals recorded by the non-invasive graphene MEA method were then converted into corresponding intracellular signals using the MLP model. Figure 8Figures AC show the results of a representative test set, showing the extracellular FPs used for derivation (blue) and the intracellular APs derived by the graphene MEA-MLP model (red). For comparison, the AP waveforms derived by the graphene MEA-MLP model were superimposed on the intracellular AP waveforms recorded by the graphene MEA-electroporation (black). In addition, in a series (~10s) of extracellular FP waveforms recorded by the graphene MEA, each extracellular FP was individually derived as an intracellular AP waveform and then concatenated to form a series of continuous intracellular AP waves ( Figure 8 To further evaluate the accuracy of the derivation, this application calculated waveform metrics such as mean square error (MSE) and R squared (R 2 ).like Figure 9 As shown in Figure A, the MSE is 0.0032 ± 0.0021, indicating that the derived intracellular waveform has high accuracy. 2 The value of 0.9519 ± 0.0313 confirmed a strong correlation between the derived and recorded intracellular APs. This application also evaluated intracellular AP parameters, including depolarization time, repolarization time, and APD10-90. The depolarization time of intracellular APs recorded by graphene MEA electroporation and derived by the graphene MEAMLP model was 37.02 ± 16.02 ms and 38.31 ± 11.06 ms, respectively ( Figure 9 B), while the repolarization time was 176.34±23.53ms and 184.31±22.61ms ( Figure 9 (C). Figure 9 As shown in Figure 3D, the maximum absolute difference in APD10-90 is only 3.33 ms, further emphasizing the high fidelity of the derived waveform parameters. These results demonstrate that the graphene MEA-MLP model accurately reconstructs the characteristics of the intracellular AP waveform and the extracellular FP.

[0147] Graphene MEA-MLP model to deduce intracellular AP waveforms from extracellular FP signals after drug treatment

[0148] The graphene MEA-MLP model further demonstrated the ability to derive intracellular APs in cells treated with drugs, which induce subtle waveform changes. Here, the present application investigated the effects of varying concentrations of quinidine and E-4031 on APs within cardiomyocytes over a 24-hour period. Quinidine can reduce the rate of depolarization in cardiomyocytes by blocking sodium channels and slowing the inward flow of sodium ions. At higher concentrations, it also blocks rectifier potassium channels, reducing potassium ion efflux and prolonging repolarization time. Figure 9As shown in Figure E, extracellular FPs were recorded from cardiomyocytes treated with 10, 50, and 100 μM quinidine, and intracellular APs were also physically recorded by graphene MEA electroporation for comparison. Intracellular APs recorded by both the graphene MEA-MLP and graphene MEA electroporation models showed a significant prolongation of the APD50, which was attributed to a slower depolarization phase. Specifically, with increasing quinidine concentrations, the APD50 values ​​of the recorded and derived intracellular APs increased by 75.81% and 94.77%, respectively ( Figure 9 The depolarization phase of AP was prolonged by 121.34% and 167.20% ( Figure 9 G, H), while the repolarization time remained essentially unchanged. However, at higher quinidine concentrations, a slight increase in repolarization time was observed. Meanwhile, compound E-4031 is a Class III antiarrhythmic drug that prolongs cardiac AP by blocking potassium channels. The present application observed that for the intracellular AP derived from the graphene MEA-MLP model and the waveform recorded by graphene MEA electroporation, the depolarization time was gradually prolonged after treatment with 10nM, 50nM and 100nM of E-4031 ( Figure 9 The APD50 derived from the graphene MEA-MLP increased from 45.66±6.10ms to 116.85±5.38ms, equivalent to an increase of 155.91%, and similar results were observed in the intracellular AP recorded by graphene MEA electroporation ( Figure 9 J). The graphene MEA-MLP-derived intracellular AP showed negligible changes during the depolarization phase, while the repolarization time increased by 163.67% ( Figure 9 These results demonstrate that this method can accurately and noninvasively monitor the effects of drugs on cardiomyocytes. This application is primarily enabled by the multi-channel graphene MEA and its accompanying high-throughput recording instrument, which effectively collects high-throughput extracellular and intracellular signal pairs.

[0149] In summary, the method provided in the embodiment of the present application includes: recording an extracellular field potential signal through a graphene microelectrode; preprocessing the extracellular field potential signal, and inputting the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein, the signal prediction model is used to characterize the connection between the extracellular field potential signal and the intracellular action potential signal; based on the continuous intracellular action potential signal, the intracellular action potential is continuously monitored. The present application predicts the intracellular action potential signal through the extracellular field potential signal, which is conducive to alleviating the cell damage problem caused by intracellular recording, achieving stable monitoring, and helping to improve monitoring accuracy.

[0150] Secondly, an embodiment of the present invention proposes a system for monitoring intracellular action potential stability based on a graphene microelectrode array, which specifically includes:

[0151] The first module is used to record extracellular field potential signals through graphene microelectrodes;

[0152] A second module is configured to preprocess the extracellular field potential signal and input the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is configured to characterize the relationship between the extracellular field potential signal and the intracellular action potential signal;

[0153] The third module is used to continuously monitor the intracellular action potential based on the continuous intracellular action potential signal.

[0154] It can be seen that the contents of the above method embodiments are all applicable to the present system embodiments. The functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0155] On the other hand, an embodiment of the present invention provides a device for monitoring intracellular action potential stability based on a graphene microelectrode array, comprising:

[0156] at least one processor;

[0157] at least one memory for storing at least one program;

[0158] When the at least one program is executed by the at least one processor, the at least one processor implements the method for monitoring intracellular action potential stability based on the graphene microelectrode array.

[0159] Similarly, the contents of the above method embodiments are applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0160] An embodiment of the present invention further provides a computer-readable storage medium storing a program executable by a processor. When the program is executed by the processor, it is used to execute the above-mentioned method for monitoring intracellular action potential stability based on a graphene microelectrode array.

[0161] Similarly, the contents of the above method embodiments are applicable to the present storage medium embodiment. The functions specifically implemented by the present storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0162] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0163] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the ordinary skill of an engineer. Therefore, a person skilled in the art will be able to implement the present invention set forth in the claims using ordinary skill without undue experimentation. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0164] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several programs for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0165] The logic and / or steps represented in a flowchart or otherwise described herein, for example, may be considered as an ordered list of executable programs for implementing the logical functions, and may be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can retrieve and execute a program from a program execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, a program execution system, apparatus, or device.

[0166] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0167] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0168] In the above description of this specification, reference to the terms "one embodiment / example," "another embodiment / example," or "certain embodiments / examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0169] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

[0170] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A method for monitoring intracellular action potential stability based on a graphene microelectrode array, characterized in that: The following steps are involved: The extracellular field potential signals were recorded by graphene microelectrodes; Preprocessing the extracellular field potential signal and inputting the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is used to characterize the relationship between the extracellular field potential signal and the intracellular action potential signal; Based on the continuous intracellular action potential signal, the intracellular action potential is continuously monitored.

2. The method for monitoring intracellular action potential stability based on graphene microelectrode array according to claim 1, characterized in that: The signal prediction model is trained by the following steps: Record extracellular field potential signal samples and intracellular action potential signal samples through graphene microelectrode electroporation; Classifying, downsampling, segmenting, and normalizing the extracellular field potential signal samples, and inputting the processed extracellular field potential signal samples into the signal prediction model to obtain predicted intracellular action potential signal results; A loss function is constructed according to the intracellular action potential signal result and the intracellular action potential signal sample, and the signal prediction model is trained according to the loss function to obtain a trained signal prediction model.

3. The method for monitoring intracellular action potential stability based on graphene microelectrode array according to claim 1, characterized in that: The signal prediction model is a multilayer perceptron neural network, which includes an input layer, a hidden layer, and an output layer. The signal prediction model is trained by the following steps: Determining a first input signal sample and a first output signal sample of the hidden layer according to a first weight between neurons of the input layer and neurons of the hidden layer; determining a second input signal sample and a second output signal sample of the output layer according to a second weight between neurons of the hidden layer and neurons of the output layer; Based on the stochastic gradient descent method, the neuron threshold, the first weight and the second weight are iteratively updated until the loss value reaches a preset loss, thereby obtaining a trained signal prediction model; wherein the preset loss is related to the set learning rate.

4. The method for monitoring intracellular action potential stability based on graphene microelectrode array according to claim 2, characterized in that: The loss function is established by the following steps: Determine the loss function through the mean square error; Alternatively, determine a loss function based on the relationship between the total variability, the variability explained by the regression model, and the residual variability.

5. The method for monitoring intracellular action potential stability based on graphene microelectrode array according to claim 1, characterized in that: The method of recording extracellular field potential signals by using a graphene microelectrode comprises: Taking a substrate, and depositing a metal layer on the substrate; Placing the substrate in a vacuum chamber, heating it to a preset temperature, and introducing hydrogen to pre-treat the surface of the substrate by plasma; introducing hydrogen and methane to generate graphene on the surface of the substrate at a preset plasma power; Cooling the graphene and metal layers, etching the graphene and metal layers based on a preset microelectrode array pattern, cutting the graphene and metal layers, and coating the non-recording area with PDMS to obtain a graphene microelectrode array chip. The graphene microelectrode array chip is fixed to a PCB to record extracellular field potential signals.

6. The method for monitoring intracellular action potential stability based on graphene microelectrode array according to claim 1, characterized in that: The method comprises: Display signal recording interface; In response to a trigger operation on a first control received on the signal recording interface, recording a cell signal based on a current recording mode; wherein the cell signal includes an extracellular field potential signal sample or an intracellular action potential signal sample, and the first control is used to start signal recording; The cell signals are marked according to preset rules, and the cell signals are grouped and displayed, and the cell signals marked as normal are stored.

7. The method for monitoring intracellular action potential stability based on graphene microelectrode array according to claim 6, characterized in that: The method further comprises: Displaying the grouping results of cell signals on the signal recording interface; In the signal recording interface, in response to a triggering operation on a first signal, a group of the first signal is adjusted.

8. A graphene microelectrode array-based intracellular action potential stabilization monitoring system, characterized in that: include: The first module is used to record extracellular field potential signals through graphene microelectrodes; A second module is configured to preprocess the extracellular field potential signal and input the preprocessed extracellular field potential signal into a signal prediction model to obtain an intracellular action potential signal; wherein the signal prediction model is configured to characterize the relationship between the extracellular field potential signal and the intracellular action potential signal; The third module is used to continuously monitor the intracellular action potential based on the continuous intracellular action potential signal.

9. A device for monitoring intracellular action potential stability based on a graphene microelectrode array, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method for monitoring intracellular action potential stabilization based on a graphene microelectrode array as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to implement the method for monitoring intracellular action potential stabilization based on a graphene microelectrode array as claimed in any one of claims 1 to 7 when executed by the processor.

Images