Wearable electrophysiological microneedle sensor, manufacturing method, and early identification method for plant stress

Abstract

A wearable electrophysiological microneedle sensor, a manufacturing method, and an early identification method for plant stress. The microneedle sensor comprises a gel substrate and microneedle modules. Each microneedle module comprises a microneedle array, an adhesive layer, a conductive layer, and a silver wire, wherein the microneedle array is sequentially coated with the adhesive layer and the conductive layer, and the conductive layer is electrically connected to the silver wire. Each microneedle array comprises a microneedle base and microneedle bodies arranged on the microneedle base in an array, and platinum black is deposited on the conductive layer of each microneedle body. Electrophysiological signals of plants are collected by the microneedle sensor, the electrophysiological signals within a preset monitoring period are preprocessed, the preprocessed electrophysiological signals are input to a plant stress identification model, and a plant stress identification result is output. The microneedle sensor can stably and with high fidelity monitor changes in electrophysiological signals of plants caused by instantaneous and long-term stress of the plants, and can identify plant stress with high accuracy in combination with a machine learning model.

Description

Wearable electrophysiological microneedle sensor, manufacturing method and early identification method of plant stress TECHNICAL FIELD

[0001] The present application belongs to the technical field of plant electrophysiological signal monitoring, and particularly relates to a wearable electrophysiological microneedle sensor, a manufacturing method and an early identification method of plant stress. BACKGROUND

[0002] When plants are subjected to stress, they perceive and transmit information about the stress to induce adaptive responses of the system. During the transmission process, the stress information is converted into various signals (force, electrical and chemical signals) to help plants coordinate the development or defense processes of the whole body. Among them, plant electrophysiological signals are important indicators of fast long-distance signal pathways and plant physiology, which regulate physiological activities such as root elongation, photosynthesis, stress response and pathogen defense. By understanding the differences in the generation and conduction of electrical signals in different crop varieties, important information can be provided for crop breeding, which helps to cultivate new varieties that are more suitable for specific environments. In addition, studying the generation and conduction mechanism of plant electrophysiological signals helps to reveal the internal mechanism of plants in coping with stress, regulating growth and development, etc., providing more methods and strategies for optimizing crop production. Therefore, plant electrophysiological research not only provides scientific basis for crop breeding and cultivation, but also helps agricultural workers to prevent plant diseases in advance, provides more technical support for agricultural production, thereby improving the efficiency and quality of agricultural production, and ultimately promoting the sustainable development of agriculture.

[0003] Currently, plant electrophysiological signal monitoring methods can be divided into two categories: non-invasive and invasive. Non-invasive methods, such as gel electrodes and microelectrode arrays, can measure plant surface electrical signals. However, the effectiveness of this method is limited by the plant cuticle barrier, which may shield the electrical signal changes and affect the fidelity of the signal due to its insulating properties. Invasive methods, such as patch clamp technology, are widely used to study ion channels, but are limited to in vitro studies and have low flux detection capability. Solid electrodes, such as standard Ag / AgCl electrodes or other inorganic metal electrodes, are commonly used for plant electrophysiological signal monitoring, but it is difficult to establish long-term stable contact with plant tissues. In addition, due to the large amount of plant electrical signal data and the complexity of the data, the lack of specialized large-scale analysis models hinders signal decoding and physiological function correlation. Revealing the dynamic change rule of plant electrical signals is a key step in understanding the plant electrophysiological signal conduction mechanism and stress defense mechanism. Therefore, it is necessary to develop a high-fidelity signal acquisition, long-term stable wearable electrophysiological microneedle sensor to monitor and identify early plant stress. SUMMARY

[0004] In view of the problems of low signal fidelity, difficulty in long-term stable monitoring and lack of data analysis model in current in-situ monitoring of plant electrophysiological signals, the application provides a wearable electrophysiological microneedle sensor, a manufacturing method and a plant stress early identification method. Compared with traditional gel electrodes, the sensor can obtain high-fidelity electrophysiological signals under plant stress, and based on the obtained electrophysiological signals, a machine learning model can be used to identify early plant stress with high accuracy.

[0005] The technical solutions adopted by the application are as follows:

[0006] (I) Wearable electrophysiological microneedle sensor

[0007] The wearable electrophysiological microneedle sensor comprises a gel substrate and a plurality of microneedle modules; the microneedle modules are laid on the surface of the gel substrate and connected with the gel substrate. When the number of microneedle modules is greater than one, the microneedle modules are uniformly arranged on the surface of the gel substrate; the microneedle module comprises a microneedle array, an adhesive layer, a conductive layer and a silver wire; the microneedle array is arranged on the gel substrate and connected with the gel substrate, the upper part of the microneedle array is covered with the adhesive layer, the upper part of the adhesive layer is covered with the conductive layer, the silver wire is bonded on the microneedle array, and the silver wire is electrically connected with the conductive layer; the microneedle array comprises a microneedle base and at least one microneedle body, when the number of microneedle bodies is greater than one, the microneedle body array is arranged on the microneedle base, and platinum black is deposited on the upper part of the conductive layer of the microneedle body.

[0008] As an optional embodiment of the application, the wearable electrophysiological microneedle sensor comprises eight microneedle modules, and the eight microneedle modules are arranged in a 2x4 matrix on the upper surface of the gel substrate; in each microneedle module, four microneedle bodies are arranged on the upper surface of the microneedle base, and the four microneedle bodies are uniformly arranged in a square matrix; a through hole is vertically arranged on the microneedle base, a silver wire is arranged in the through hole, the silver wire is bonded with the microneedle base through conductive glue, and the silver wire is electrically connected with the conductive layer and an external electrophysiological data acquisition device.

[0009] Specifically, the microneedle body is a conical needle body, and the ratio between the height and the bottom diameter of the microneedle body is 2-3:1.

[0010] Specifically, the adhesive layer is a chromium layer, the conductive layer is a gold layer, and the adhesive layer and the conductive layer are formed by magnetron sputtering technology.

[0011] Specifically, the microneedle array is integrally manufactured by a numerical control machine tool processing technology from polymethyl methacrylate material.

[0012] Specifically, the silver wires on all the microneedle modules are combined by twisting to form a unified electrical signal output path.

[0013] (ii) A manufacturing method of the wearable electro-physiological microneedle sensor

[0014] The manufacturing method comprises the following steps:

[0015] S1) Using numerical control machine tool processing technology to process polymethyl methacrylate material to obtain at least one microneedle array.

[0016] S2) Using magnetron sputtering technology to sequentially deposit an adhesion layer and a conductive layer on the surface of each microneedle array to obtain at least one primary processing microneedle module;

[0017] The magnetron sputtering conditions in the step S2) are specifically: the magnetron sputtering power is 50-500 W, the chamber vacuum degree is 0-2 mTorr, and the sample rotation speed is 50-200 rpm.

[0018] S3) After filling the gel substrate mold with the acrylate precursor solution, all the primary processing microneedle modules are placed above the acrylate precursor solution, and after ultraviolet curing, a primary processing microneedle sensor is obtained.

[0019] The ultraviolet curing process in the step S3) is specifically: using an ultraviolet curing lamp to irradiate the acrylate precursor solution in the gel substrate mold; the ultraviolet light wavelength of the ultraviolet curing lamp is 365-380 nm, and the irradiation time is 30-50 s. The acrylate can be at least one of epoxy acrylate, polyurethane acrylate, polyester acrylate and methacrylate.

[0020] S4) Using conductive glue to adhere silver wires to the microneedle base in each primary processing microneedle module of the primary processing microneedle sensor obtained in step S3) and heat curing, so that in each primary processing microneedle module of the primary processing microneedle sensor, the conductive layer of the microneedle base is electrically connected to the corresponding silver wire itself.

[0021] Subsequently, the microneedle bodies on all the primary processing microneedle modules of the primary processing microneedle sensor are immersed in an electrodeposition solution, and after deposition treatment on the surface of the microneedle bodies by chronopotentiometry, they are blown dry, so that platinum black is deposited on the conductive layer on the surface of all the microneedle bodies, obtaining a wearable electro-physiological microneedle sensor. The step S4) comprises the following steps:

[0022] S4.1) Using epoxy conductive glue to adhere at least one silver wire to the microneedle base in each primary processing microneedle module of the primary processing microneedle sensor, and then heating and curing at an ambient temperature of 60-80℃ for 2 h, so that at least one silver wire is adhered to the microneedle base in each primary processing microneedle module.

[0023] S4.2) Immersing the microneedle bodies on all the primary processing microneedle modules of the primary processing microneedle sensor in an electrodeposition solution, taking the microneedle bodies as working electrodes, a platinum wire electrode as a counter electrode, and a silver / silver chloride electrode as a reference electrode, depositing platinum black on the surface of the microneedle bodies by chronoamperometry, and obtaining the wearable electrophysiological microneedle sensor after drying with an inert gas;

[0024] The electrodeposition solution is a mixed solution of chloroplatinic acid and lead acetate, and the concentration ratio of chloroplatinic acid to lead acetate in the electrodeposition solution is 100:1; the conditions of the deposition process are specifically as follows: a working constant voltage of -0.1 V, a sampling interval of 0.1 s, a working time length of 60 s, and a sensitivity of 10 -3 A / V.

[0025] (Three) Application of the wearable electrophysiological microneedle sensor

[0026] The wearable electrophysiological microneedle sensor is used for identifying plant stress, especially early identification.

[0027] (Four) Plant stress early identification method based on the wearable electrophysiological microneedle sensor

[0028] The plant stress early identification method comprises the following steps:

[0029] First, the wearable electrophysiological microneedle sensor is attached to the stem of the plant, the silver wire in the wearable electrophysiological microneedle sensor is electrically connected to the input end of the electrophysiological data acquisition device, and the output end of the electrophysiological data acquisition device is connected to the upper computer.

[0030] Then, the wearable electrophysiological microneedle sensor is used to continuously collect the electrophysiological signals of the plant in real time, and the collected electrophysiological signals are transmitted to the upper computer through the electrophysiological data acquisition device, the electrophysiological signals of the plant in a preset monitoring period are summarized, and the original time sequence data is formed;

[0031] Finally, after batch average processing of the original time sequence data, the batch average processed time sequence data is obtained; the batch average processed time sequence data is divided and processed by using a static window division method, a plurality of time sequence segments are generated, time domain features and frequency domain features are extracted from each time sequence segment, feature data of each time sequence segment is obtained, input features are obtained after summarizing the feature data of all time sequence segments; the input features are input into a pre-trained plant stress identification model, and the plant stress identification model processes the input features and outputs a plant stress identification result, the plant stress identification result being a plant stress type.

[0032] The range of the preset monitoring period is 3-7 days.

[0033] The time domain features include mean, standard deviation, maximum, minimum, median, 25% quantile, 75% quantile, peak-to-peak value, wave factor, peak factor and pulse factor; and the frequency domain features include power spectrum entropy, center of gravity frequency, mean square frequency, frequency variance and frequency standard deviation.

[0034] The plant stress types include long-term stress types such as drought stress, saline-alkali stress and nutrient deficiency. When plants are subjected to the above long-term stress, the plants do not show obvious symptoms that can be observed by naked eyes in a short period of time. However, the plant electrical signals can rapidly reflect the reaction of the plants to the environmental stress and provide timely information, because the plant electrical signals can rapidly propagate in the plant body. Therefore, the plant stress can be rapidly identified by machine learning processing of the electrical signals.

[0035] Specifically, when the plants are subjected to long-term stress such as mild drought stress, saline-alkali stress or nutrient deficiency, the plant electrical signals change within 3-7 days, and the symptoms that can be observed by naked eyes such as yellowing leaves begin to appear after about 7-10 days. Therefore, the plant stress can be identified in an early stage by identifying the plant physiological electrical signals within 3-7 days.

[0036] The training process of the plant stress identification model includes the following steps:

[0037] D1) a plurality of plants are divided into N+1 groups, and any one group of plants is cultured under normal culture conditions, and the remaining N groups of plants are cultured under N different stress culture conditions;

[0038] Each plant is continuously monitored for 7 days using a wearable electrophysiological microneedle sensor to obtain the original electrical signal time series data of each plant. The stress corresponding to the stress culture condition of each plant is taken as the corresponding label of each plant. After the original electrical signal time series data of each plant is paired with the corresponding label, a sample pair corresponding to each plant is obtained. Then, the sample pairs corresponding to all plants are collected to obtain an original data set.

[0039] D2) input features are extracted from the original electrical signal time series data in the sample pair corresponding to each plant. The input features are re-paired with the labels in the sample pair corresponding to each plant to obtain a feature sample pair corresponding to each plant. Then, the feature sample pairs corresponding to all plants are collected to obtain a feature data set.

[0040] The process of extracting input features from the original electrical signal time series data specifically comprises: after batch average processing of the original electrical signal time series data, time series data after batch average processing is obtained; the time series data after batch average processing is divided and processed by using a static window division method to generate a plurality of time series segments, time domain features and frequency domain features are extracted from each time series segment to obtain feature data of each time series segment, and the feature data of all time series segments are collected to obtain input features;

[0041] D3) constructing a plant stress recognition model based on an extreme gradient boosting model, using the feature data set obtained in step D2) to optimize the hyperparameters of the plant stress recognition model by combining a genetic algorithm, obtaining optimal hyperparameters, and configuring the plant stress recognition model with the optimal hyperparameters to obtain a trained plant stress recognition model; the hyperparameters include a learning rate, a number of trees, a maximum tree depth, and a regularization coefficient.

[0042] The plant stress early recognition method further comprises the following step: when the amplitude difference between the electrical physiological signals at two adjacent time points is greater than a preset threshold, outputting a plant stress recognition result, and the plant stress recognition result is that the plant is subjected to transient stress. This step is mainly used for monitoring whether the plant is subjected to mechanical stress, insect bites, flame burning and other different types of transient stress.

[0043] The microneedle sensor of the application can be stably fixed on the plant surface for a long time, can effectively obtain high-fidelity electrical physiological signals of the plant under transient and long-term stress, and can recognize plant stress with high accuracy based on the obtained electrical physiological signals by means of a machine learning model.

[0044] The application has the following beneficial effects:

[0045] 1. The microneedle sensor in the application can break through the stratum corneum to obtain electrical physiological signals in the plant body, the nanomaterial deposited on the surface of the microneedle body can accelerate the charge transfer rate and reduce the charge transfer impedance of the electrode-tissue interface, thereby improving the signal quality of the monitoring.

[0046] 2. The microneedle sensor in the application has excellent electrical physiological signal acquisition capability in the process of transient stress monitoring and has long-term stability in long-term stress monitoring.

[0047] 3、The plant adversity stress recognition model is obtained by combining and training the phase time sequence data of the plant electrophysiological signal obtained based on the microneedle sensor in the application and a machine learning model, and the plant adversity stress recognition model has excellent reliability and accuracy in identifying plant adversity stress, especially early plant adversity stress. Through accurate identification and management of plant adversity stress, crop yield and quality can be improved, and resource waste and environmental impact can be reduced. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a schematic diagram of the microneedle sensor in the application; wherein (a) is a structural schematic diagram of the microneedle sensor; (b) is a scanning electron microscope image of the microneedle body;

[0049] Figure 2 is a scanning electron microscope image and an energy spectrum diagram of the microneedle sensor in the application; wherein (a) is a scanning electron microscope image of the microneedle sensor obtained in Comparative Example 1, (b) is a scanning electron microscope image of the microneedle sensor obtained in Example 1, (c) is a platinum element energy spectrum diagram of the microneedle sensor obtained in Example 1, and (d) is a gold element energy spectrum diagram of the microneedle sensor obtained in Example 1;

[0050] Figure 3 is a microneedle impedance diagram of the microneedle sensor in the application;

[0051] Figure 4 is a schematic diagram of the plant transient stress monitoring results obtained in Example 2 of the application; wherein (a) is a plant transient stress electrophysiological signal diagram, and (b) is a plant transient stress electrophysiological signal quality diagram;

[0052] Figure 5 is a transient stress electrophysiological signal spectrum diagram monitored by the microneedle sensor and the gel electrode in Example 2 of the application;

[0053] Figure 6 is a time domain characteristic diagram of the electrophysiological signal under long-term stress monitored by the microneedle sensor in the application; wherein (a) is a time sequence curve of the mean value, (b) is a time sequence curve of the peak-to-peak value, and (c) is a time sequence curve of the standard deviation;

[0054] Figure 7 is a frequency domain characteristic diagram of the electrophysiological signal under long-term stress monitored by the microneedle sensor in the application; wherein (a) is a column chart of the power change over time under different stress conditions, (b) is a column chart of the center of gravity frequency change over time under different stress conditions, and (c) is a column chart of the power spectrum entropy change over time under different stress conditions;

[0055] Figure 8 is a monitoring effect diagram of the plant adversity stress recognition model in the application. DETAILED DESCRIPTION

[0056] The application will be further described in detail below in combination with the drawings and examples, and it should be pointed out that the following examples are intended to facilitate the understanding of the application and do not limit the application in any way.

[0057] The terms involved in the present application are further explained as follows:

[0058] The "stress" in the present application includes, but is not limited to, drought stress, saline-alkali stress, nutrient deficiency and heavy metal stress, etc.

[0059] The "transient stress" in the present application includes, but is not limited to, transient mechanical stress, insect bites, flame burning, etc.

[0060] The "early stage" in the present application mainly refers to the initial stage of the occurrence of stress, when the plant has not yet shown significant physiological or morphological changes.

[0061] The "early identification" in the present application refers to: at the initial stage of the occurrence of stress, before the plant has shown significant physiological or morphological changes, the plant is predicted and identified whether it has been affected by stress and the type of stress by the electrophysiological signal of the plant.

[0062] In the description of the present application, unless otherwise explicitly stated, the side of the microneedle sensor with the microneedle body is defined as the upper side.

[0063] The present application provides a wearable electrophysiological microneedle sensor, comprising a gel substrate and at least one microneedle module; the microneedle module is laid on the surface of the gel substrate and connected with the gel substrate. Wherein, the gel substrate is a flexible insulating material formed by light curing technology, by arranging several separated microneedle modules on the surface of the gel substrate in intervals, and adopting the bionic design of insect morphology, the adaptability of the wearable electrophysiological microneedle sensor to different plants and different parts of the plants is enhanced.

[0064] Specifically, the microneedle module comprises a microneedle array, an adhesive layer, a conductive layer and a silver wire; the microneedle array is located on the gel substrate, the microneedle array is covered with the adhesive layer on the top, the adhesive layer is covered with the conductive layer on the top, and the silver wire is arranged on the microneedle array and electrically connected with the conductive layer. Wherein, the adhesive layer is a chromium (Cr) layer, the conductive layer is an aurum (Au) layer, and the adhesive layer and the conductive layer are both formed by magnetron sputtering technology.

[0065] Specifically, the microneedle array comprises a microneedle base and at least one microneedle body, and the microneedle bodies are arranged in an array on the microneedle base.

[0066] The microneedle array is integrally manufactured by machining technology from polymethyl methacrylate (PMMA) material. The microneedle array is integrally manufactured by the process, so that the microneedle base and the microneedle body form a seamless whole structure. The microneedle array is the main structure of the microneedle module, which not only improves the stability and durability of the microneedle sensor, but also ensures the reliability and consistency of the microneedle sensor during use.

[0067] The microneedle body is a conical needle body, and the ratio between the height and the bottom diameter of the microneedle body is 2-3:1, so that the microneedle body can break through the stratum corneum to obtain the electrophysiological signal in the plant body.

[0068] The conductive layer of each microneedle body is deposited with platinum black. The platinum black deposited on the surface of the microneedle body can accelerate the charge transfer rate, reduce the charge transfer impedance of the electrode-tissue interface, and thus improve the signal quality of the monitoring.

[0069] Specifically, a through hole is formed in the microneedle base, and a silver wire is arranged in the through hole. The silver wire is bonded to the microneedle base by conductive adhesive, preferably epoxy conductive adhesive. The silver wire is electrically connected to the conductive layer and the external electrophysiological data acquisition device.

[0070] Further, the silver wires on all microneedle modules are combined by twisting to form a unified electrical signal output path.

[0071] Further, the present application provides an alternative embodiment, wherein the wearable electrophysiological microneedle sensor comprises eight microneedle modules, and the eight microneedle modules are arranged in a 2x4 matrix on the upper surface of the gel substrate to form an elliptical shape. The microneedle array in each microneedle module comprises four microneedle bodies arranged in a matrix on the upper surface of the microneedle base. The microneedle base has a through hole formed therein, and the through hole is vertically formed and has a silver wire arranged therein. The silver wire is bonded to the microneedle base by conductive adhesive, and the silver wire is electrically connected to the conductive layer and the external electrophysiological data acquisition device.

[0072] The specific embodiments of the present application are as follows:

[0073] Embodiment 1

[0074] In this embodiment, a wearable electrophysiological microneedle sensor is manufactured, and the structure of the wearable electrophysiological microneedle sensor is shown in FIG. 1.

[0075] As shown in FIG. 1(a), the wearable electrophysiological microneedle sensor comprises a gel substrate and eight separate microneedle modules, and the eight microneedle modules are arranged on the surface of the gel substrate and connected to the gel substrate. The eight microneedle modules are arranged in a 2(row)×4(column) matrix.

[0076] Each microneedle module comprises a microneedle array, an adhesive layer, a conductive layer and a silver wire. The microneedle array is arranged on and connected to the gel substrate, the microneedle array is covered with the adhesive layer, the adhesive layer is covered with the conductive layer, the microneedle array is provided with the silver wire, and the silver wire is electrically connected to the conductive layer. The adhesive layer is a chromium layer with a thickness of 15 nm, and the conductive layer is a gold layer with a thickness of 200 nm.

[0077] The microneedle array in each microneedle module is made of polymethyl methacrylate material, and comprises a microneedle substrate and four microneedle bodies. The microneedle substrate is provided with the microneedle bodies on the upper surface, and the lower surface of the microneedle substrate is arranged and connected to the gel substrate.

[0078] As shown in (b) of FIG. 1, the thickness of the microneedle substrate is 1 mm, and the height and bottom diameter of the microneedle body are about 800 μm and 300 μm, respectively. In the same microneedle module, the distance between the center lines of two adjacent microneedle bodies is 600 μm. A through hole is vertically provided on the side of the microneedle substrate away from the center, and the silver wire is bonded to the microneedle substrate through the conductive adhesive after passing through the through hole. The silver wire is electrically connected to the conductive layer and the external electrophysiological data acquisition device.

[0079] In this embodiment, the wearable electrophysiological microneedle sensor is manufactured by the following steps:

[0080] S1) Machining technology is used to process polymethyl methacrylate material to obtain eight microneedle arrays;

[0081] S2) A chromium layer (adhesive layer) and a gold layer (conductive layer) are sequentially deposited on the surface of each microneedle array by magnetron sputtering to obtain eight primary microneedle modules;

[0082] The thickness of the chromium layer is 15 nm, and the thickness of the gold layer is 200 nm.

[0083] The conditions of magnetron sputtering are as follows: the magnetron sputtering power is 100 W, the chamber vacuum degree is kept constant at 2 mTorr, and the sample rotation speed is kept constant at 120 rpm.

[0084] S3) A gel substrate mold was obtained by 3D printing technology using a transparent resin in advance, and the printing accuracy was set to 50 μm; then, after filling the pre-manufactured gel substrate mold with an acrylate precursor solution, all the primary processing microneedle modules were placed above the acrylate precursor solution according to the above arrangement (eight microneedle modules were arranged in a 2x4 matrix); the acrylate precursor solution was subjected to photopolymerization crosslinking reaction under the irradiation of ultraviolet light (UV) with a wavelength of 365 nm, and the photopolymerization crosslinking reaction time was 30 s; the final microneedle array with gel on the back was taken out of the gel substrate mold, which was used as a primary processing microneedle sensor;

[0085] wherein the lower surface of the primary processing microneedle module is in contact with the acrylate precursor solution;

[0086] wherein the acrylate precursor solution is mainly composed of an acrylate monomer and a photoinitiator (or photosensitizer). The photoinitiator (or photosensitizer) in the acrylate precursor solution absorbs ultraviolet light under the irradiation of ultraviolet light to produce active free radicals or cations, which initiate monomer polymerization and crosslinking chemical reaction, so that the acrylate precursor solution is converted from a liquid state to a flexible gel substrate in solid state within a few seconds.

[0087] wherein the acrylate can be at least one of epoxy acrylate, polyurethane acrylate, polyester acrylate and methacrylate.

[0088] S4) The primary processing microneedle sensor was adhered to the silver wire with epoxy conductive adhesive and then heated and cured in an oven at 80°C for 2 h; then, the microneedle body on the primary processing microneedle sensor was immersed in an electrodeposition solution, wherein the electrodeposition solution was composed of 1 g of hydrogen hexachloroplatinate (PtCl6H2), 0.01 g of lead acetate (Pb(C2HO2)2) and 100 mL of ultrapure water; the microneedle body was used as a working electrode, a platinum wire electrode was used as a counter electrode, and an Ag / AgCl electrode was used as a reference electrode; platinum black was deposited on the surface of the microneedle by chronocoulometry, and then dried with nitrogen; finally, a wearable electro-physiological microneedle sensor was obtained; wherein the working constant voltage of chronocoulometry was -0.1 V, the sampling interval was 0.1 s, the working time was 60 s, and the sensitivity was 10 -3 A / V.

[0089] Comparative Example 1

[0090] The microneedle sensor of this comparative example was manufactured according to the same process and conditions as steps S1) to S3) in Example 1. The microneedle sensor of this comparative example was the same as that of Example 1 in size and structure, except that the surface of the microneedle body was not deposited with platinum black.

[0091] The results of the morphology analysis of the microneedle sensors manufactured in Example 1 and Comparative Example 1 are as follows:

[0092] The micro-needle body surface micro-morphology was observed by scanning electron microscope (SEM), and the results are shown in (a)-(b) of FIG. 2. Compared with the bare micro-needle body (Comparative Example 1) without deposited platinum black, the micro-needle body (Example 1) with deposited platinum black has dense granular structures on the surface, which can act as charge transfer channels to accelerate the charge transfer rate and thus reduce the charge transfer impedance of the electrode-tissue interface.

[0093] Further, the micro-needle body of the micro-needle sensor obtained in Example 1 was subjected to surface analysis by X-ray energy spectrum analysis, and the results are shown in (c) of FIG. 2 and (d) of FIG. 2. The micro-needle with deposited platinum black has Pt and Au elements on the surface.

[0094] Meanwhile, it can be seen from the impedance test results shown in FIG. 3 that the impedance of the micro-needle with deposited platinum black (2584.22 ± 647.4 Ω) is one order of magnitude lower than that of the bare micro-needle (10165.31 ± 2259.63 Ω), which verifies the successful construction of the micro-needle conductive interface.

[0095] Example 2

[0096] In this example, a commercial gel electrode and the micro-needle sensor obtained in Example 1 were used to monitor the transient stress of plants, taking tomatoes as an example.

[0097] The process of monitoring the transient stress of plants is as follows: the gel electrode and the micro-needle sensor are placed equidistantly on both sides of the midrib of the tomato seedling stem, the plant is subjected to transient mechanical stress by cutting the plant leaves with scissors, and the plant electrophysiological signals are obtained by an electrophysiological data acquisition device. As shown in (a) of FIG. 4 and (b) of FIG. 4, the amplitude of the electrophysiological signals recorded by the micro-needle sensor is 11.7 ± 0.7 mV, and the signal-to-noise ratio is 20.5 ± 5.5 dB, while the amplitude of the electrophysiological signals recorded by the gel electrode is 3.5 ± 1 mV, and the signal-to-noise ratio is 13.7 ± 2.5 dB.

[0098] It can be seen that when the plant is subjected to transient mechanical stress, the amplitude of the electrophysiological signal is much larger than the fluctuation value, and therefore the transient stress state of the plant (i.e. whether the plant is subjected to mechanical stress, insect bites, flame burns, etc.) can be monitored or identified by obtaining the amplitude difference between the electrophysiological signals at two adjacent time points.

[0099] Further, the recorded cutting electrical signals were processed by fast Fourier transform (hereinafter referred to as FFT), and as shown in FIG. 5, the electrophysiological signals caused by the cutting of the tomato seedling leaves are concentrated around 5 Hz, and the amplitude of the electrical signals recorded by the micro-needle array is significantly larger than that of the gel electrode. These results show that compared with the commercial gel electrode, the electrophysiological micro-needle sensor can capture high-fidelity and stable plant electrophysiological signals.

[0100] Example 3

[0101] This embodiment uses the microneedle sensor obtained in Experimental Example 1 for long-term stress monitoring of tomato seedlings (drought stress and saline-alkali stress). The tomato seedlings are continuously watered / normally watered / salted water (0.1 mol / L) for 7 days, and the electrical signals of the tomato seedlings during normal / drought stress / saline-alkali stress are recorded (7 days). The amplitude and frequency of the electrical signals recorded under different conditions all show slight changes, and the plant electrophysiological sensor can sensitively detect these changes.

[0102] As can be seen from FIG. 6, although the mean values of the electrical signals under different conditions have no significant difference (FIG. 6(a)), the peak-to-peak values (FIG. 6(b)) and standard deviations (FIG. 6(c)) of the electrical signals both show obvious fluctuations in the drought and salt stress groups. In particular, after the sixth day, these values show obvious trends: the peak-to-peak values and standard deviations of the salt stress group show an upward trend, while those of the drought stress group show a downward trend. Drought stress causes plant cells to lose water, reduces the change in ion concentration inside the cells, and thus reduces the change in the amplitude of the electrical signals. However, salt stress affects the ion balance in plant cells, causing the ion concentration to fluctuate more, thus increasing the change in the amplitude of the electrical signals. The fluctuations in the time-domain characteristics of the electrical signals of the control group are relatively small, indicating that the electrical signals have certain correlation with the growth status of the tomato seedlings.

[0103] Due to the non-stationary nature of plant electrical signals, time-domain analysis alone is limited. Therefore, the signal is converted to the frequency domain for further analysis by using FFT. Power spectrum analysis shows that the frequency of the electrical signals of the tomato seedlings is mainly distributed below 1 Hz, and the average power spectrum peak values of the salt stress and drought stress groups both exceed those of the control group (FIG. 7(a)). The changes in the center of gravity frequency (SGF) of the electrical signals of the tomato seedlings shown in FIG. 7(b) and the power spectrum entropy (PSE) shown in FIG. 7(c) demonstrate the time sequence changes in the power spectrum of the electrical signals of the tomato seedlings under different conditions.

[0104] As can be seen, the center of gravity frequency of the normal group (watered normally for 7 days) first increases, then decreases, and finally tends to be stable, the center of gravity frequency of the saline-alkali stress group (watered with salt water for 7 days) shows a gradually decreasing trend, and the center of gravity frequency of the drought group shows a trend of first decreasing, then increasing, and finally decreasing, indicating that the power spectrum distribution of the electrical signals expands to the low frequency band. Similarly, the change patterns of the power spectrum entropy of the three groups are similar to those of the center of gravity frequency. In addition, the power spectrum entropy of the drought and salt stress groups both finally show a downward trend, indicating that the signal complexity decreases and the cell activity is inhibited, which is consistent with the phenomenon of wilting and yellowing of the leaves of the tomato seedlings under stress. Thus, it can be concluded that the wearable microneedle array provided by the present application can not only be used for transient stress monitoring of plants, but also be used for long-term stress monitoring of plants.

[0105] Example 4

[0106] The embodiment is based on an XGBoost model to construct a plant stress identification model, and based on a genetic algorithm to iteratively optimize the hyperparameters of the plant stress identification model, thereby obtaining a plant stress identification model for early identification of plant stress.

[0107] The specific process of the embodiment is as follows:

[0108] The training process of the plant stress identification model includes the following steps:

[0109] D1) Twelve plants are divided into three groups: a normal watering group, a drought stress group, and a saline-alkali stress group, and the three groups of plants are cultured under three different culture conditions; the culture conditions of the three groups of plants are as follows:

[0110] Normal watering group: water regularly every day to keep the soil moist;

[0111] Drought stress group (stress culture condition): no watering for 7 consecutive days;

[0112] Salt-alkali stress group (stress culture condition): water regularly every day, using salt water with a concentration of 0.1 mol / L;

[0113] Each plant is continuously monitored for 7 days using a wearable electrophysiological microneedle sensor, and the original electrical signal time series data of each plant is obtained. The stress corresponding to the stress culture condition of each plant is used as its own corresponding label: the label of the normal watering group is normal, the label of the drought stress group is drought stress, and the label of the salt-alkali stress group is salt-alkali stress. The stress corresponding to the stress culture condition of each plant is used as its own corresponding label, and the original electrical signal time series data of each plant is paired with its own corresponding label to obtain its own corresponding sample pair, and then the sample pairs corresponding to the twelve plants are summarized to obtain the original data set.

[0114] D2) The input features are extracted from the original electrical signal time series data in the sample pair corresponding to each plant, and the input features and the labels in the sample pair form the feature sample pair corresponding to the plant; then, the feature sample pairs corresponding to the twelve plants are summarized to obtain the feature data set. The feature data set is divided into a training set and a test set according to a ratio of 3:1.

[0115] The process of extracting the input features from the original electrical signal time series data is specifically as follows: the original electrical signal time series data is divided into multiple batches, and each batch contains 5 data points. Then, the average value of all data points in each batch is calculated, and the average values corresponding to all batches are summarized to obtain the time series data after batch average processing. The time series data after batch average processing is divided into multiple time series segments by using a static window division method according to the manner that each window contains 200 data points, time domain features and frequency domain features are extracted from each time series segment, the time domain features and the frequency domain features of each time series segment form the corresponding feature data of the time series segment, the feature data corresponding to all time series segments are summarized to obtain the input features corresponding to the original electrical signal time series data.

[0116] The time domain features include average value, standard deviation, maximum value, minimum value, median value, 25% quantile, 75% quantile, peak-to-peak value, waveform factor, peak factor and pulse factor, etc. The frequency domain features include power spectrum entropy, center of gravity frequency, mean square frequency, frequency variance and frequency standard deviation, etc.

[0117] D3) Based on the Extreme Gradient Boosting (XGBoost) model, a plant stress recognition model is constructed, the feature data set obtained in step D2) is used, and a genetic algorithm is used to optimize the hyperparameters of the plant stress recognition model, to obtain the optimal hyperparameters, and the plant stress recognition model is configured with the optimal hyperparameters to obtain the optimal plant stress recognition model; wherein the hyperparameters include learning rate, number of trees, maximum tree depth and regularization coefficient.

[0118] The process of using the feature data set obtained in step D2) to optimize the hyperparameters of the plant stress recognition model by using a genetic algorithm is specifically as follows:

[0119] D3.1) initialize to generate an initial population; the initial population includes a plurality of individuals, and each individual includes four to-be-optimized hyperparameters, which are learning rate, number of trees, maximum tree depth and regularization coefficient, respectively;

[0120] Then, the initial population is taken as the current population, and step D3.2) is entered;

[0121] D3.2) Obtain the fitness value of each individual in the current population respectively. Specifically, for each individual in the current population, use the hyperparameter configuration contained in the individual to plant stress identification model, train the plant stress identification model using the training set, obtain the trained plant stress identification model, evaluate the trained plant stress identification model using the test set, obtain the accuracy of the trained plant stress identification model, and use the accuracy as the fitness value of the individual; wherein the accuracy is the proportion of the number of samples correctly predicted by the classification model to the total number of samples, which is calculated by the following formula:

[0122] ACCURACY = (TP + TN) / (TP + TN + FP + FN)

[0123] Wherein, ACCURACY is the accuracy of the trained plant stress identification model; TP represents the number of samples that are actually positive and correctly predicted as positive; TN represents the number of samples that are actually negative and correctly predicted as negative; FP represents the number of samples that are actually negative and incorrectly predicted as positive; FN represents the number of samples that are actually positive and incorrectly predicted as negative.

[0124] D3.3) Determine whether the maximum number of iterations is reached, if the maximum number of iterations is not reached, go to step D3.4); if the maximum number of iterations is reached, the iteration is ended, and the hyperparameters contained in the individual with the highest fitness value are used as the optimal hyperparameters; in this embodiment, the maximum number of iterations is 5 times.

[0125] D3.4) Select elite individuals from the current population according to fitness, generate a new population by crossing and mutating the elite individuals; then, use the new population as the current population, and return to step D3.2) to obtain the fitness value of each individual in the current population; in this embodiment, the crossover probability is 0.5 and the mutation probability is 0.2.

[0126] D4) attaching the wearable electrophysiological microneedle sensor on the stem of the plant, electrically connecting the silver wire in the wearable electrophysiological microneedle sensor with the input end of the electrophysiological data acquisition device, connecting the output end of the electrophysiological data acquisition device with the host computer, continuously collecting the electrical signals of the plant using the wearable electrophysiological microneedle sensor and transmitting the collected electrical signals to the host computer through the electrophysiological data acquisition device, collecting the electrical signals of the plant in the last 3 days to obtain original time series data, performing batch average processing on the original time series data in the host computer to obtain time series data after batch average processing, dividing and processing the time series data after batch average processing using a static window division method to generate a plurality of time series segments, extracting time domain features and frequency domain features from each time series segment, the time domain features and the frequency domain features of each time series segment forming feature data corresponding to the time series segment, and the feature data corresponding to all time series segments forming input features. The input features are input into the optimal plant stress identification model obtained in step D3), and the optimal plant stress identification model outputs a plant stress identification result after processing the input features. In this embodiment, the plant stress identification result includes normal, drought stress and saline-alkali stress.

[0127] As shown in the confusion matrix in FIG. 8, the total recognition accuracy of the model obtained in this embodiment is 99.29%, among which the recognition accuracy of the normal control group is 98.98%, the accuracy of the saline-alkali group is 99.17%, and the accuracy of the drought group is 99.68%, which illustrates the superior recognition performance of the model. The plant electrophysiological decoding model based on machine learning has good recognition accuracy and scalability, and is expected to play an important role in plant physiological state monitoring, stress diagnosis and intelligent management of agricultural production.

[0128] In summary, the wearable microneedle sensor of the present application can obtain plant electrophysiological signals with long-term stability and high fidelity, which is of great significance for plant stress monitoring and dynamic analysis and management of plant growth and development.

[0129] The above specific embodiments further illustrate the purpose, technical solutions and advantages of the present application. It should be understood that the above description is only for specific embodiments of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A wearable electrophysiological microneedle sensor, characterized by: The wearable electrophysiological microneedle sensor comprises a gel substrate and a plurality of microneedle modules; the microneedle modules are laid on the surface of the gel substrate and connected with the gel substrate, the microneedle modules are uniformly arranged on the surface of the gel substrate; the microneedle module comprises a microneedle array, an adhesive layer, a conductive layer and a silver wire; the microneedle array is arranged on the gel substrate, the microneedle array is covered with the adhesive layer, the adhesive layer is covered with the conductive layer, and the silver wire is electrically connected with the conductive layer; the microneedle array comprises a microneedle base and at least one microneedle body, the microneedle body array is arranged on the microneedle base, and platinum black is deposited on the conductive layer of the microneedle body.

2. The wearable electro-physiologic microneedle sensor of claim 1, wherein: The wearable electrophysiological microneedle sensor comprises eight microneedle modules, and the eight microneedle modules are arranged on the upper surface of the gel substrate in the form of a 2*4 matrix; in each microneedle module, the upper surface of the microneedle base is arranged with four microneedle bodies, and the four microneedle bodies are uniformly arranged in the form of a square matrix; a through hole is vertically arranged on the microneedle base, and a silver wire is arranged in the through hole and bonded with the microneedle base; the silver wire is electrically connected with the conductive layer and an external electrophysiological data acquisition device.

3. The wearable electro-physiologic microneedle sensor of claim 1, wherein: The microneedle array is integrally manufactured from polymethyl methacrylate material, the microneedle body is a conical needle body, the ratio of the height of the microneedle body to the diameter of the bottom end is 2-3:1; the adhesive layer is a chromium layer, the conductive layer is a gold layer, and the adhesive layer and the conductive layer are formed by magnetron sputtering technology.

4. A method of manufacturing a wearable electrophysiological microneedle sensor according to any one of claims 1 to 3, characterized in that: The manufacturing method comprises the following steps: S1) using numerical control machine tool processing technology to process polymethyl methacrylate material to obtain at least one microneedle array; S2) using magnetron sputtering technology to sequentially deposit an adhesive layer and a conductive layer on the surface of each microneedle array to obtain at least one preliminary processing microneedle module; S3) filling the gel substrate mold with an acrylic precursor solution, placing all the preliminary processing microneedle modules above the acrylic precursor solution, and then performing ultraviolet curing to obtain a preliminary processing microneedle sensor; S4) using conductive glue to bond a silver wire on each microneedle base and heat curing, then immersing all the microneedle bodies in an electrodeposition solution, and then performing deposition treatment on the surface of the microneedle body by chronopotentiometry and blowing dry to obtain a wearable electrophysiological microneedle sensor.

5. The method of manufacturing a wearable electrophysiological microneedle sensor of claim 4, wherein: The magnetron sputtering conditions in step S2) are specifically: the magnetron sputtering power is 50-500 W, the chamber vacuum degree is 0-2 mTorr, and the sample rotation speed is 50-200 rpm; the ultraviolet curing process in step S3) is specifically: using an ultraviolet curing lamp to irradiate the acrylic ester precursor solution in the gel substrate mold; the ultraviolet light wavelength of the ultraviolet curing lamp is 365-380 nm, and the irradiation time is 30-50 s.

6. The method of manufacturing a wearable electrophysiological microneedle sensor of claim 4, wherein: The step S4) comprises the following steps: S4.1) using epoxy conductive glue to bond at least one silver wire on the microneedle base in each preliminary processing microneedle module, and then heating and curing at an environmental temperature of 60-80℃ for 2 h; S4.2) Immersing the microneedle bodies on all the primary processing microneedle modules in an electrodeposition solution, taking the microneedle bodies as working electrodes, a platinum wire electrode as a counter electrode, and a silver / silver chloride electrode as a reference electrode, depositing platinum black on the surface of the microneedle bodies by chronoamperometry, and obtaining wearable electro-physiological microneedle sensors after drying with inert gas; the electrodeposition solution is a mixed solution of chloroplatinic acid and lead acetate, and the concentration ratio of chloroplatinic acid to lead acetate in the electrodeposition solution is 100:1; the conditions of the deposition process are specifically as follows: a working constant voltage of-0.1 V, a sampling interval of 0.1 s, a working time length of 60 s, and a sensitivity of 10 -3 A / V.

7. Use of a wearable electrophysiological microneedle sensor according to any one of claims 1 to 3 or a wearable electrophysiological microneedle sensor obtained by the manufacturing method according to any one of claims 4 to 6. Identification of plant stress.

8. A method for early identification of plant stress based on the wearable electrophysiological microneedle sensor according to any one of claims 1-3 or the wearable electrophysiological microneedle sensor obtained by the manufacturing method according to any one of claims 4-6. Comprise the following steps: Firstly, the wearable electrophysiological microneedle sensor is attached to the stem of the plant, the silver wire in the wearable electrophysiological microneedle sensor is electrically connected with the input end of the electrophysiological data acquisition device, and the output end of the electrophysiological data acquisition device is connected with the upper computer; Then, the wearable electrophysiological microneedle sensor is used to collect the electrophysiological signals of the plant in real time, and the collected electrophysiological signals are transmitted to the upper computer through the electrophysiological data acquisition device, the electrophysiological signals of the plant in a preset monitoring period are summarized, and original time sequence data is formed; Finally, after batch average processing of the original time sequence data, time sequence data after batch average processing is obtained; The time sequence data after batch average processing is divided and processed by using a static window division method, a plurality of time sequence segments are generated, time domain features and frequency domain features are extracted from each time sequence segment, feature data of each time sequence segment is obtained, input features are obtained by summarizing the feature data of all time sequence segments; the input features are input into the pre-trained plant stress identification model, and the plant stress identification model processes the input features and outputs a plant stress identification result, wherein the plant stress identification result is a plant stress type; The preset monitoring period ranges from 3 to 7 days; the time domain features include mean, standard deviation, maximum, minimum, median, 25% quantile, 75% quantile, peak-to-peak value, waveform factor, peak factor and pulse factor; and the frequency domain features include power spectrum entropy, center of gravity frequency, mean square frequency, frequency variance and frequency standard deviation.

9. The method for early recognition of plant stress according to claim 8, characterized in that: The training process of the plant stress identification model includes the following steps: D1) a plurality of plants are divided into N+1 groups, any one group of plants is cultured under normal culture conditions, and the remaining N groups of plants are cultured under N different stress culture conditions; Each plant is continuously monitored for 7 days using the wearable electrophysiological microneedle sensor to obtain original electrophysiological signal time sequence data of each plant, the stress corresponding to each plant is taken as a label corresponding to the plant, and after pairing the original electrophysiological signal time sequence data of each plant with the label corresponding to the plant, a sample pair corresponding to the plant is obtained, and then sample pairs corresponding to all plants are summarized to obtain an original data set; D2) input features are extracted from the original electrophysiological signal time sequence data in each sample pair corresponding to a plant, the input features are re-paired with the label in the sample pair corresponding to the plant to obtain a feature sample pair corresponding to the plant; then, the feature sample pairs corresponding to all plants are summarized to obtain a feature data set; The process of extracting input features from the original electrophysiological signal time sequence data is as follows: after batch average processing of the original electrophysiological signal time sequence data, time sequence data after batch average processing is obtained; the time sequence data after batch average processing is divided and processed by using a static window division method, a plurality of time sequence segments are generated, time domain features and frequency domain features are extracted from each time sequence segment, feature data of each time sequence segment is obtained, and input features are obtained by summarizing the feature data of all time sequence segments; D3) constructing a plant stress identification model based on an extreme gradient boosting model, using the feature data set obtained in step D2) to optimize the hyperparameters of the plant stress identification model by a genetic algorithm, obtaining optimal hyperparameters, and configuring the plant stress identification model with the optimal hyperparameters to obtain a trained plant stress identification model; the hyperparameters include a learning rate, a number of trees, a maximum tree depth, and a regularization coefficient.

10. The method of claim 8, wherein the method is for early recognition of abiotic stress in plants. The identification method further comprises the following step: when the amplitude difference between the electrophysiological signals at two adjacent time points is greater than a preset threshold, outputting a plant stress identification result, wherein the plant stress identification result is that the plant is subjected to transient stress.

Images