AI-enabled real-time electrochemical detection system

Abstract

The invention discloses an AI-enabled real-time electrochemical detection system, which is composed of a real-time electrochemical detection module and an embedded AI data processing and analysis module, the real-time electrochemical detection module comprises an electrochemical electrode system, a sample pool and an electrochemical workstation, the surface of a working electrode is modified with a probe of a specific recognition target, and when the probe is combined with the target, the probe of the specific recognition target can be detected by the embedded AI data processing and analysis module. The working electrode interface charge transfer impedance changes in real time to generate a dynamic electric signal, and the electrochemical workstation provides electric input and collects the signal; the embedded AI data processing and analysis module is used for denoising the dynamic electric signals through an adaptive filtering algorithm based on wavelet transform, extracting time-frequency domain joint features through a 1D-CNN and LSTM mixed deep learning model, and establishing a nonlinear mapping relation between the signals and target concentration. Through AI enabling electrochemical detection, the defects of low time efficiency, large error and low information utilization rate of the traditional technology are comprehensively overcome, and an innovative solution is provided for high-precision, rapid and anti-interference target substance detection.

Description

Technical Field

[0001] This invention belongs to the field of intelligent electrochemical detection technology, and in particular relates to a dynamic analysis system that combines artificial intelligence (AI) with real-time electrochemical detection. Background Technology

[0002] Electrochemical detection technology, with its high integration, ease of operation, and simple deployment, has been widely used in biomedical diagnostics, environmental monitoring, food safety, and industrial analysis. Traditional electrochemical detection methods, such as cyclic voltammetry (CV), differential pulse voltammetry (DPV), and electrochemical impedance spectroscopy (EIS), typically employ an endpoint detection mode, where the target concentration is calculated by measuring the steady-state signal after the molecular recognition event on the working electrode surface reaches reaction equilibrium. However, this static detection mode has inherent limitations in the detection of enzyme-free and non-electroactive substances.

[0003] 1. Low time efficiency: Endpoint detection requires waiting for redox reaction equilibrium, which takes 30 minutes to several hours, and cannot meet the needs of point-of-care testing (POCT) or rapid on-site testing. For example, in clinical diagnosis, traditional electrochemical immunosensors require 1-2 hours of incubation time to detect tumor markers, which limits their application in emergency departments, primary care, or telemedicine.

[0004] 2. Large random errors: Single endpoint measurements are easily affected by factors such as uneven electrode surface modification, temperature fluctuations, solution pH changes, or electronic instrument noise, resulting in poor repeatability of test results. In complex samples, matrix effects may mask the target signal, causing false positive or false negative results.

[0005] 3. Low information utilization: Only static signals after the reaction is completed are collected, ignoring dynamic information during the target-probe molecule binding process, such as binding rate constant and diffusion mass transfer characteristics. Existing technologies cannot effectively extract and utilize these kinetic parameters containing key information.

[0006] Although some studies have attempted to improve this, such as using high-frequency sampling to record signal changes and combining microfluidic chips to achieve continuous monitoring, there are still problems such as insufficient anti-interference ability and complex signal analysis. The application of introducing machine learning and deep learning into electrochemical data analysis is also mostly limited to the post-processing of endpoint detection data.

[0007] Therefore, we propose an AI-enabled real-time electrochemical detection system to address the aforementioned issues. Summary of the Invention

[0008] The purpose of this invention is to provide an AI-enabled real-time electrochemical detection system to solve the problems of long detection time, low efficiency, poor accuracy and repeatability of existing electrochemical detection technologies, which cannot meet the needs of instant detection and rapid on-site detection.

[0009] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0010] This invention relates to an AI-enabled real-time electrochemical detection system, comprising:

[0011] Real-time electrochemical detection module: includes an electrochemical electrode system, a sample cell, and an electrochemical workstation;

[0012] Embedded AI data processing and analysis module: includes filtering algorithms and deep learning models;

[0013] in,

[0014] The working electrode surface of the electrochemical electrode system is modified with a probe that specifically recognizes the target substance; when the probe binds to the target substance, the charge transfer impedance at the working electrode interface changes in real time, generating a dynamic electrical signal; the electrochemical workstation provides electrical input to the electrode system and collects the dynamic electrical signal.

[0015] The embedded AI data processing and analysis module performs the following operations sequentially on the dynamic electrical signals:

[0016] (a) Smoothing and denoising preprocessing is performed using a filtering algorithm;

[0017] (b) Extract time-frequency domain joint features through a deep learning model to establish a nonlinear mapping relationship between signal changes and target concentration.

[0018] In one embodiment, the real-time electrochemical detection module employs differential pulse voltammetry (DPV), square wave pulse voltammetry (SWV), or cyclic voltammetry scanning (CV).

[0019] In one embodiment, the filtering algorithm is an adaptive filtering algorithm based on wavelet transform, which improves the signal-to-noise ratio by ≥20dB.

[0020] In one embodiment, the deep learning model is a hybrid architecture of 1D-CNN and LSTM, comprising:

[0021] The 1D-CNN module extracts local temporal features (such as peak value and slope) of dynamic signals;

[0022] The bidirectional LSTM module contains 128 hidden units and introduces an attention mechanism to weight key time points;

[0023] The output features of 1D-CNN and the temporal features of LSTM are concatenated and then fused through a fully connected layer;

[0024] The output layer uses the Sigmoid activation function and optimizes concentration prediction with the Huber loss function.

[0025] In one embodiment, the dynamic signal is the curve of the change in peak current of DPV over time (ΔI-t), which is obtained by the embedded module extracting the peak current in real time from multiple DPV scan curves and subtracting it from the peak current of the initial curve.

[0026] The hybrid deep learning architecture of 1D-CNN and LSTM enables rapid prediction of sample concentration based on the variation law and trend of ΔI in the initial stage of electrochemical measurement, thereby improving detection speed.

[0027] In one embodiment, the electrochemical electrode system includes:

[0028] A circular gold working electrode with a surface-modified aptamer probe containing thiol groups;

[0029] Arc-shaped silver paste reference electrode;

[0030] Arc-shaped gold counter electrode;

[0031] Each electrode is connected to the electrochemical workstation via gold / silver wires and interfaces.

[0032] In one embodiment, the electrochemical electrode system is a rod-shaped electrode, and the sample cell is a beaker; the electrode system is connected to the electrochemical workstation via a wire with an electrode clamp, forming a circuit within the beaker.

[0033] In one embodiment, the electrochemical electrode system is a micro / nano electrode chip system, and the sample pool is a micro / nano liquid sample tank integrated on the chip surface;

[0034] The electrode chip system is connected to a portable electrochemical workstation via a micro-nano fabrication interface, and the embedded AI data processing and analysis module is built into the portable workstation.

[0035] In one embodiment, the probe includes an aptamer, an antibody, or a DNA probe; the target substance includes at least one of the following: nucleic acid molecules, tumor markers, or other enzyme-free, non-electroactive substances.

[0036] The real-time electrochemical detection method based on the system includes the following steps:

[0037] (a) Immerse the working electrode of the modified probe into a sample solution containing the target substance;

[0038] (b) Dynamic electrical signals are acquired in real time by applying a DPV scanning signal through an electrochemical workstation;

[0039] (c) The embedded AI module filters and denoises the dynamic electrical signal and extracts time-frequency domain joint features through a deep learning model;

[0040] (d) Output the quantitative concentration of the target substance based on the nonlinear mapping relationship.

[0041] This invention deeply integrates real-time electrochemical detection technology with artificial intelligence technology, and has the following significant advantages compared with traditional electrochemical detection technology:

[0042] 1. Significantly improves testing efficiency

[0043] Breaking through the limitations of traditional endpoint detection which requires waiting for reaction equilibrium (30 minutes to several hours), this method achieves rapid detection through real-time dynamic signal acquisition and AI real-time analysis, combined with differential pulse voltammetry (DPV) with ≥31 cyclic scans. This meets the needs of point-of-care testing (POCT) and rapid on-site testing, and is especially suitable for time-sensitive scenarios such as emergency rooms and primary healthcare.

[0044] 2. Significantly improves detection accuracy and repeatability.

[0045] An adaptive filtering algorithm based on wavelet transform is adopted to improve the signal-to-noise ratio by ≥20dB, effectively reducing interference from electronic instrument noise, temperature fluctuations, etc.

[0046] The hybrid deep learning model of 1D-CNN and bidirectional LSTM extracts joint time-frequency features to establish a nonlinear mapping relationship between signal and target concentration. Combined with Huber loss function to optimize prediction, it reduces random errors in single measurements, improves the repeatability of results, and reduces the risk of false positives / false negatives.

[0047] 3. Fully explore dynamic information and improve information utilization.

[0048] Abandoning the traditional static endpoint detection mode, this method captures dynamic electrical signals (such as the peak current change curve of DPV) during the target-probe binding process. It extracts local temporal features through 1D-CNN and mines temporal dependencies through bidirectional LSTM (containing 128 hidden units and attention mechanism). It effectively utilizes dynamic information such as binding rate constant and diffusion mass transfer characteristics, breaking through the limitation of traditional techniques that only utilize steady-state signals.

[0049] 4. Enhance anti-interference capability and detection sensitivity

[0050] The working electrode surface is modified with specific probes (aptamers, antibodies, etc.), and combined with AI models for precise feature extraction, significantly improving the ability to identify low-concentration target substances in complex matrices (such as biological samples and environmental samples), and is suitable for high-sensitivity detection of various substances such as proteins, nucleic acids, and tumor markers.

[0051] 5. Expands application scenarios and has the potential for portability.

[0052] The system is compatible with various electrode systems, such as rod electrodes and micro / nano electrode chips, and can be integrated into portable electrochemical workstations. The embedded AI module supports real-time analysis and is suitable for multiple fields such as biomedical diagnostics, environmental monitoring, and food safety, especially meeting the needs of rapid on-site detection.

[0053] In summary, this invention, through AI-enabled electrochemical detection, comprehensively overcomes the shortcomings of traditional technologies, such as low time efficiency, large errors, and low information utilization, providing an innovative solution for high-precision, rapid, and interference-resistant target substance detection.

[0054] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0055] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 A schematic diagram of a three-electrode system for micro / nano fabrication (working electrode modified with an aptamer);

[0057] Figure 2 The real-time detection process for the DPV method (oscilloscope peak current changes with antigen binding);

[0058] Figure 3 A schematic diagram illustrating data processing for a deep learning model with a hybrid architecture of 1D-CNN and LSTM. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] In the description of this invention, it should be understood that the terms "upper," "middle," "outer," "inner," etc., which indicate orientation or positional relationship, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting this invention.

[0061] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0062] Example 1

[0063] Please see Figures 1-3 As shown, this invention is an AI-enabled real-time electrochemical detection system, comprising:

[0064] Real-time electrochemical detection module: includes an electrochemical electrode system, a sample cell, and an electrochemical workstation;

[0065] Embedded AI data processing and analysis module: includes filtering algorithms and deep learning models;

[0066] in,

[0067] The working electrode surface of the electrochemical electrode system is modified with a probe that specifically recognizes the target substance; when the probe binds to the target substance, the charge transfer impedance at the working electrode interface changes in real time, generating a dynamic electrical signal; the electrochemical workstation provides electrical input to the electrode system and collects the dynamic electrical signal.

[0068] The embedded AI data processing and analysis module executes the following steps sequentially on the dynamic electrical signals:

[0069] (a) Smoothing and denoising preprocessing is performed using a filtering algorithm;

[0070] (b) Extract time-frequency domain joint features through a deep learning model to establish a nonlinear mapping relationship between signal changes and target concentration.

[0071] like Figure 2 As shown, the real-time electrochemical detection module uses differential pulse voltammetry (DPV), and its operating parameters include:

[0072] Base potential range: -0.2V to -0.8V;

[0073] Pulse width: 0.1V;

[0074] Step width: 0.2V;

[0075] Delay: 10s;

[0076] The number of cyclic scans is ≥31.

[0077] Furthermore, the filtering algorithm is an adaptive filtering algorithm based on wavelet transform, which improves the signal-to-noise ratio by ≥20dB.

[0078] like Figure 3 As shown, the deep learning model is a hybrid architecture of 1D-CNN (one-dimensional convolutional neural network) and LSTM (long short-term memory network), including:

[0079] The 1D-CNN module extracts local temporal features of dynamic signals;

[0080] The bidirectional LSTM module contains 128 hidden units and introduces an attention mechanism to weight key time points;

[0081] The output features of 1D-CNN and the temporal features of LSTM are concatenated and then fused through a fully connected layer;

[0082] The output layer uses the Sigmoid activation function and optimizes concentration prediction with the Huber loss function.

[0083] Furthermore, the dynamic signal is the curve of the change in peak current of DPV over time (ΔI-t), which is obtained by the embedded module extracting the peak current in real time from multiple DPV scan curves and subtracting it from the peak current of the initial curve; the deep learning hybrid architecture of 1D-CNN and LSTM realizes rapid prediction of sample concentration based on the change law and trend of ΔI in the initial stage of electrochemical measurement, thereby improving the detection speed.

[0084] Furthermore, the electrochemical electrode system includes:

[0085] The circular gold working electrode is surface-modified with thiol-containing aptamer probes, which can form Au-S bonds with the gold working electrode, fix it on the surface of the gold working electrode, and specifically bind to the target nucleic acid molecules such as DNA and RNA according to the base complementary pairing principle, so as to achieve highly specific sensing.

[0086] Arc-shaped silver paste reference electrode;

[0087] Arc-shaped gold counter electrode;

[0088] Each electrode is connected to the electrochemical workstation via gold / silver wires and interfaces.

[0089] Furthermore, the electrochemical electrode system is a rod-shaped electrode, and the electrochemical detection module also includes a beaker as a sample pool, in which the target sample solution is stored. The electrochemical electrode system is connected to the electrochemical workstation via a wire with an electrode clamp at one end, and forms a circuit in the sample solution in the beaker.

[0090] like Figure 1As shown, the electrochemical electrode system is a micro / nano electrode chip system, and the sample cell is a micro / nano liquid sample tank integrated on the chip surface. The liquid sample tank is manufactured on the surface of the electrode chip system and contains the electrode chip system. The liquid sample stored in the liquid sample tank can completely cover the surface of the contained electrode chip system.

[0091] The electrode chip system is connected to a portable electrochemical workstation via a micro-nano fabrication interface, and an embedded AI data processing and analysis module is built into the portable workstation.

[0092] Furthermore, the probes include aptamers, antibodies, or DNA probes; the target substances include at least one of the following: nucleic acid molecules, tumor markers, and other enzyme-free, non-electroactive substances.

[0093] This embodiment describes a real-time electrochemical detection method based on an electrochemical detection system, including the following steps:

[0094] (a) Immerse the working electrode of the modified probe into a sample solution containing the target substance;

[0095] (b) Dynamic electrical signals are acquired in real time by applying a DPV scanning signal through an electrochemical workstation;

[0096] (c) The embedded AI module filters and denoises the dynamic electrical signal and extracts time-frequency domain joint features through a deep learning model;

[0097] (d) Output the quantitative concentration of the target substance based on the nonlinear mapping relationship.

[0098] Example 2

[0099] This embodiment illustrates the working principle of an AI-enabled real-time electrochemical detection system:

[0100] Signal generation: The target substance (such as nucleic acid, tumor marker) binds to the probe on the working electrode surface, causing a change in the interfacial charge transfer impedance, generating a dynamic electrical signal (such as the DPV peak current curve);

[0101] Signal processing:

[0102] (a) Smoothing and denoising preprocessing is performed using a filtering algorithm;

[0103] (b) Extract time-frequency domain joint features through a deep learning model to establish a nonlinear mapping relationship between signal changes and target concentration.

[0104] Output: Real-time output of target quantitative concentration.

[0105] Example 3

[0106] like Figure 1As shown, a specific embodiment of a real-time electrochemical detection module design for a micro / nano-fabricated electrochemical three-electrode system is provided:

[0107] In practical applications, the electrochemical electrode system, electrode shape design, number of working electrodes, and preparation method can be reasonably selected according to the application scenario and the requirements of the target to be tested.

[0108] In this exemplary embodiment, the working electrode is designed as a circular gold electrode, the reference electrode as an arc-shaped silver paste electrode, and the counter electrode as an arc-shaped gold electrode. The working electrode and the counter electrode are connected to the downstream electrochemical workstation via gold wires and a square gold electrode interface; the reference electrode is connected to the downstream electrochemical workstation via silver wires and a square silver electrode interface. In this exemplary embodiment, an aptamer with a designed thiol group at one end is modified on the surface of the working electrode. This aptamer can form an Au-S bond with the gold working electrode, is fixed on the surface of the gold working electrode, and can specifically bind to the target nucleic acid molecules such as DNA and RNA according to the base complementary pairing principle, thereby achieving highly specific sensing.

[0109] Example 4

[0110] like Figure 2 As shown, a specific embodiment of real-time electrochemical detection using differential pulse voltammetry is provided:

[0111] In practical applications, methods such as differential pulse voltammetry, cyclic voltammetry, amperometric voltammetry, and electrochemical impedance spectroscopy can be selected appropriately based on the application scenario and the type of target to be tested.

[0112] In this exemplary embodiment, a conventional large-scale electrochemical workstation (2) is used to provide electrical signals and collect feedback detection signals for the electrochemical electrode system. In the DPV detection of this exemplary embodiment, a base potential (-0.2 to 0.8 V) with a step-increasing value is applied to the working electrode, with a pulse width of 0.1 V, a step width of 0.2 V, a delay of 10 s, and 31 scan cycles. In this exemplary embodiment, the antigen and antibody are used as the target and probe (1), respectively. As shown in the figure, as the antibody on the surface of the working electrode binds to the antigen in the sample, the peak current in the DPV curve increases (the absolute value decreases). Based on this, 31 DPV scan curves (5) at this antigen concentration are obtained. The embedded AI data processing and analysis module preprocesses the 31 DPV scan curves received by the electrochemical workstation to extract the peak value and obtain the peak current change curve (4).

[0113] Example 5

[0114] like Figure 3As shown, a specific embodiment is provided for data processing and analysis of electrochemical dynamic signals using an adaptive filtering algorithm based on wavelet transform and a deep learning model with a hybrid architecture of 1D-CNN (one-dimensional convolutional neural network) + LSTM (long short-term memory network):

[0115] In practical applications, filtering algorithms and deep learning models can be selected appropriately based on requirements such as signal type and detection method.

[0116] In this exemplary embodiment, each peak curve extracted from the 31 DPV curves corresponds to a sample concentration. An adaptive filtering algorithm smooths and denoises the peak curves, distinguishing between effective signals and noise, improving the signal-to-noise ratio by ≥20dB. The 1D-CNN module is responsible for extracting local features from the original time-domain signal, effectively capturing local fluctuation features in the signal, such as peak values ​​and slope changes. The LSTM module focuses on modeling temporal features, employing a bidirectional LSTM structure containing 128 hidden units. It introduces an attention mechanism to automatically weight key time nodes, processes the feature sequences extracted by the 1D-CNN, and establishes long-term dependencies, making it particularly suitable for analyzing dynamic response processes in electrochemical signals. The high-dimensional features output by the 1D-CNN are concatenated with the temporal features extracted by the LSTM, and feature fusion is achieved through a fully connected layer. The final output layer uses the Sigmoid activation function and the Huber loss function to balance the influence of outliers, achieving accurate prediction of sample concentration.

[0117] It needs to be further explained that:

[0118] The electrochemical electrode system in the real-time electrochemical detection module includes, but is not limited to, two-electrode systems, three-electrode systems, four-electrode systems, and multi-electrode array systems.

[0119] The electrochemical electrode system in the real-time electrochemical detection module is prepared by methods including but not limited to laser induction, screen printing, radio frequency sputtering, and chemical vapor deposition; the working electrode materials include but are not limited to novel nanomaterials such as graphene, gold, and carbon paste.

[0120] AI-powered real-time electrochemical detection system can detect targets including but not limited to viral proteins, inflammatory factors, tumor markers and other proteins and antigens, nucleic acids such as DNA, RNA and microRNA, small molecule metabolites such as glucose, lactic acid and cholesterol, and neurotransmitters and hormones such as dopamine, adrenaline and serotonin.

[0121] The sample cell in the real-time electrochemical detection module includes, but is not limited to, containers such as beakers that are adapted to the volume of rod-shaped electrodes or screen-printed electrodes, and micro-liquid tanks or microchannels fabricated on the electrode surface that are adapted to micro-nano fabricated electrodes.

[0122] The electrochemical workstation in the real-time electrochemical detection module includes, but is not limited to, conventional large-scale electrochemical workstations and portable electrochemical workstations based on integrated circuits.

[0123] The filtering algorithms in the embedded AI data processing and analysis module include, but are not limited to, digital filtering algorithms such as moving average filtering, recursive least squares, median filtering, Savitzky-Golay, classical mode decomposition, and wavelet threshold denoising, as well as machine learning methods such as Kalman filtering and deep learning end-to-end denoising.

[0124] Deep learning models in embedded AI data processing and analysis modules include, but are not limited to, time-series feature analysis models such as LSTM and Transformer, frequency-domain feature analysis models such as Fourier transform and wavelet analysis, noise suppression and signal enhancement models such as generative adversarial networks and Kalman filtering, and multimodal data fusion models such as convolutional neural networks and graph neural networks.

[0125] In the description of this specification, references to terms such as "an embodiment," "example," and "specific example" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0126] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An AI-enabled real-time electrochemical detection system, characterized in that, include: Real-time electrochemical detection module: includes an electrochemical electrode system, a sample cell, and an electrochemical workstation; Embedded AI data processing and analysis module: includes filtering algorithms and deep learning models; in, The working electrode surface of the electrochemical electrode system is modified with a probe that specifically recognizes the target substance. When the probe binds to the target material, the charge transfer impedance at the working electrode interface changes in real time, generating a dynamic electrical signal; the electrochemical workstation provides electrical input to the electrode system and collects the dynamic electrical signal. The embedded AI data processing and analysis module performs the following operations sequentially on the dynamic electrical signals: (a) Smoothing and denoising preprocessing is performed using a filtering algorithm; (b) Extract time-frequency domain joint features through a deep learning model to establish a nonlinear mapping relationship between signal changes and target concentration.

2. The AI-enabled real-time electrochemical detection system according to claim 1, characterized in that, The real-time electrochemical detection module employs differential pulse voltammetry (DPV), square wave pulse voltammetry (SWV), or cyclic voltammetry scanning (CV).

3. The AI-enabled real-time electrochemical detection system according to claim 2, characterized in that, The filtering algorithm is an adaptive filtering algorithm based on wavelet transform, which improves the signal-to-noise ratio by ≥20dB.

4. The AI-enabled real-time electrochemical detection system according to claim 3, characterized in that, The deep learning model is a hybrid architecture of 1D-CNN and LSTM, including: The 1D-CNN module extracts local temporal features of dynamic signals; The bidirectional LSTM module contains 128 hidden units and introduces an attention mechanism to weight key time points; The output features of 1D-CNN are concatenated with the temporal features of LSTM and then fused through a fully connected layer; The output layer uses the Sigmoid activation function and optimizes concentration prediction with the Huber loss function.

5. The AI-enabled real-time electrochemical detection system according to claim 4, characterized in that, The dynamic signal is the curve of the change in peak current of DPV over time (ΔI-t), which is obtained by the embedded module extracting the peak current in real time from multiple DPV scan curves and subtracting it from the peak current of the initial curve. The hybrid deep learning architecture of 1D-CNN and LSTM enables rapid prediction of sample concentration based on the variation law and trend of ΔI in the initial stage of electrochemical measurement, thereby improving detection speed.

6. The AI-enabled real-time electrochemical detection system according to claim 1, characterized in that, The electrochemical electrode system includes: A circular gold working electrode with a surface-modified aptamer probe containing thiol groups; Arc-shaped silver paste reference electrode; Arc-shaped gold counter electrode; Each electrode is connected to the electrochemical workstation via gold / silver wires and interfaces.

7. The AI-enabled real-time electrochemical detection system according to claim 1, characterized in that, The electrochemical electrode system is a rod-shaped electrode, and the sample cell is a beaker; the electrode system is connected to the electrochemical workstation via wires with electrode clips, forming a circuit inside the beaker.

8. The AI-enabled real-time electrochemical detection system according to claim 1, characterized in that, The electrochemical electrode system is a micro / nano electrode chip system, and the sample cell is a micro / nano liquid sample tank integrated on the chip surface; The electrode chip system is connected to a portable electrochemical workstation via a micro-nano fabrication interface, and the embedded AI data processing and analysis module is built into the portable workstation.

9. The AI-enabled real-time electrochemical detection system according to claim 1, characterized in that, The probe includes an aptamer, antibody, or DNA probe; the target substance is an enzyme-free, non-electroactive substance.

10. A real-time electrochemical detection method based on the system according to any one of claims 1-9, characterized in that, Including the following steps: (a) Immerse the working electrode of the modified probe into a sample solution containing the target substance; (b) Dynamic electrical signals are acquired in real time by applying a DPV scanning signal through an electrochemical workstation; (c) The embedded AI module filters and denoises the dynamic electrical signal and extracts time-frequency domain joint features through a deep learning model; (d) Output the quantitative concentration of the target substance based on the nonlinear mapping relationship.

Images