Electrodes with ultra-low bioelectronic impedance based on molecular anchoring

Abstract

Disclosed herein is an ionically conductive gel comprising a polymeric binder material in an amount of from 5 to 20 wt% of the gel, a metal salt in an amount of from 2 to 10 wt% of the gel, and a balance of a hydrophilic polymeric material wherein the gel matrix is formed by intermolecular interaction between the polymeric binder material and the hydrophilic polymeric material, and the metal salt is distributed throughout the gel matrix. Also disclosed herein is a flexible electrode device comprising a flexible substrate material having a first surface and a second surface, a metal electrode sheet on the first surface of the substrate, and an ion conductive gel layer covering the metal electrode sheet, and a prosthetic control system comprising one or more electrode devices.

Description

Technical Field

[0001] The present invention relates generally to ion-conducting gels, and more particularly to ion-conducting gels comprising a polymer binder material and a metal salt, and is suitable for use in flexible electrode devices and prosthetic control systems. Background Art

[0002] The listing or discussion of a previously published document in this specification should not necessarily be taken as an admission that the document is part of the state of the art or is common general knowledge.

[0003] Myoelectric control is an advanced technology for amputees that involves detecting, identifying, and applying non-destructive myoelectric signals to control prostheses. However, due to skin fibrosis and atrophy caused by long-term inactivity in amputees, the ionic-electronic coupling efficiency at the skin-electrode interface is low, which may easily distort low-level myoelectric signals, especially those from <10% maximal voluntary contraction (MVC).

[0004] Disliking but relying on their electric prostheses is a typical feeling of amputees, but this dislike leads to an average rejection and abandonment rate of 35% for these prostheses (Biddiss, EA & Chau, TT Prosthet. Orthot. Int. 2007, 31, 236). In order to realize the ultimate dream of amputees: the integration of man and prosthesis, people have been actively exploring the design of a myoelectric prosthesis with capabilities and comfort comparable to human limbs (Craelius, W. Science, 2002, 295, 1018; Hochberg, LR et al., Nature, 2012, 485, 372). As a basic premise, distortion-free high-fidelity myoelectric signals can ensure accurate identification of phantom limb activity, thereby achieving precise prosthetic control. Myoelectric signals are electrical action potentials generated during muscle contraction. These signals contain neural information from the brain that controls specific movements, making them useful as a source of input for prosthetic control in amputees who have been trained to produce them (Kaur, M. et al., Adv. Mater., 2021, 33, 2002534; Craelius, W., Science, 2002, 295, 1018; Hochberg, LR et al., Nature, 2012, 485, 372). Myoelectric signals can be recorded by placing non-invasive electrodes on the skin of the residual limb. However, these signals vary greatly and are usually normalized to the maximum voluntary contraction (MVC) of the muscle. For fine and dexterous movements such as finger and wrist movements, muscle contractions are less than 10% MVC (Alenabi, T. et al., J. Shoulder Elbow Surg., 2013, 22, 1400; Martin, BJ et al., Hum. Factors, 1996, 38, 654). The challenge is that the myoelectric signals generated by these actions are often distorted and have a low signal-to-noise ratio (SNR) because traditional electrodes cannot effectively couple with the amputee's skin. Therefore, amputees are often forced to use stronger muscle contractions (more than 10% MVC) in order to trigger the recognition of weak but fine finger movement signals. This mismatch between the amputee's muscle contractions and the prosthetic movements creates a physiological conflict between the new reflex arc and the past muscle memory. It also creates a lot of incoordination and excessive friction at the prosthesis-residual limb interface, leading to discomfort, excessive sweating, skin allergies, and ultimately abandonment of myoelectric prostheses (Biddiss, EA & Chau, TT, Prosthet. Orthot. Int., 2007, 31, 236; Gajewski, D. & Granville, R., J. Am. Acad. Orthop. Surg., 2006, 14, S183).Improving the quality and availability of low-level EMG signals, especially those below 10% MVC, is critical to bridging this mismatch and allowing amputees to more precisely control their prostheses.

[0005] Detection of electromyographic signals involves the capacitive coupling of ion flux in the skin and electron flow in the gel electrode through two electrical double layers (EDL) at the skin-electrode interface (Figure 1) (Yang, C. & Suo, Z., Nat. Rev. Mater., 2018, 3, 125; Yuk, H. et al., Chem. Soc. Rev., 2019, 48, 1642; Cai, P. et al., Nat. Commun., 2020, 11, 2183). Traditionally, methods to enhance ion-electron coupling and reduce bioelectronic impedance have focused on making electrodes more flexible and stretchable (King, DR et al., Adv. Mater., 2014, 26, 4345; Miyamoto, A. et al., Nat. Nanotechnol., 2017, 12, 907; Pan, S. et al., Adv. Funct. Mater., 2020, 30, 1909540) or improving the interaction between electrodes and skin surface molecules (Pan, L. et al., Adv. Mater., 2020, 32, 2003723; Luo, Y. et al., Adv. Mater., 2021, 33, 2007848) to adapt to curved skin. However, these methods can only improve the detection of low-level signals in normal skin. Due to internal skin changes caused by amputation, such as fibrosis and atrophy, capacitive coupling on amputated skin is still weak. This leads to an imbalance in intracellular and extracellular calcium concentrations in the amputated skin due to long-term inactivity (Lannan, FM & Meyerle, JH, Cutis, 2019, 103, 86), which reduces the induced charge at the electrode metal layer during capacitive coupling (Parekh, AB, J. Physciol., 2003, 547, 333; Agrawal, A. et al., J. Cell Commun. Signal., 2018, 12, 645). This imbalance raises the level of interface impedance from a value close to the intrinsic impedance of the human body (6-10 kΩ) to a range from high kΩ to low MΩ, making it impossible to record myoelectric signals below 10% MVC with high quality (which is crucial to compensate for this mismatch). It is known that about 80% of the bioelectrical impedance (the so-called barrier effect) comes from the outermost layer of the skin, the stratum corneum (SC) (Wu, KS et al., Biomaterials, 2006, 27, 785; and Kabiri Ameri, S. et al., ACS Nano, 2017, 11, 7634). Summary of the invention

[0006] It has been unexpectedly found that the ion-conducting gel produced and used in the production of gel electrodes can be firmly anchored into the skin, thereby reducing the above-mentioned barrier effect and increasing the ion-electron coupling area. Therefore, the present invention relates to said ion-conducting gel, its use in gel electrodes and the use of said electrodes in prosthetic devices.

[0007] Aspects and embodiments of the invention will now be discussed by reference to the following numbered items.

[0008] 1. An ion conductive gel comprising:

[0009] a polymeric binding material in an amount of 5 to 20 wt % of the gel;

[0010] a metal salt in an amount of 2 to 10 wt % of the gel; and

[0011] The rest is a hydrophilic polymer material, wherein a gel matrix is ​​formed between the polymer binding material and the hydrophilic polymer material, and the metal salt is distributed throughout the gel matrix.

[0012] 2. The ion-conductive gel according to item 1, wherein the number average molecular weight of the hydrophilic polymer material is 200 to 15,000 Daltons.

[0013] 3. The ion conductive gel according to item 1 or item 2, wherein the hydrophilic polymer material is polyethylene glycol or a polyol.

[0014] 4. The ion-conductive gel according to item 3, wherein the hydrophilic polymer material is polyethylene glycol, optionally wherein the polyethylene glycol has a number average molecular weight of 200.

[0015] 5. An ion-conducting gel according to any of the preceding clauses, wherein the polymer binder material is selected from one or more of the group consisting of polysaccharides, cellulose hydrogels, protein hydrogels, nucleic acid gels, biohydrogels, and more particularly poly(N-vinyl caprolactam) (PVCL), polyvinylpyrrolidone (PVP), polydopamine and polymerized acrylics.

[0016] 6. The ion-conductive gel according to item 5, wherein the polymer binder material is selected from one or more of the group consisting of poly(N-vinyl caprolactam) (PVCL), polyvinyl pyrrolidone (PVP), polydopamine and polymerized acrylic.

[0017] 7. The ion-conductive gel according to item 6, wherein the polymer binder material is poly(N-vinyl caprolactam) (PVCL).

[0018] 8. An ion-conducting gel according to any one of the preceding clauses, wherein the polymer binder material comprises about 10 wt% of the total weight of the gel.

[0019] 9. The ion-conductive gel according to any one of the preceding clauses, wherein the metal salt is one or more selected from the group consisting of sodium salts, potassium salts, calcium salts and magnesium salts.

[0020] 10. The ion-conductive gel according to item 9, wherein the metal salt is sodium chloride.

[0021] 11. The ion-conductive gel according to any one of the preceding clauses, wherein the metal salt accounts for about 5 wt% of the total weight of the gel.

[0022] 12. A flexible electrode device, comprising:

[0023] a flexible substrate material having a first surface and a second surface;

[0024] a metal electrode sheet on a first surface of the substrate; and

[0025] An ion conductive gel layer covering the metal electrode sheet, wherein the ion conductive gel is as described in any one of items 1 to 11.

[0026] 13. The electrode device according to item 12, wherein the device further comprises two or more metal electrode sheets on the first surface of the substrate,

[0027] wherein each of the two or more metal electrode sheets is electrically connected to at least one conductive path,

[0028] wherein each of the two or more metal electrode sheets is covered by an ion conductive gel layer, and

[0029] Therein, each conductive path is covered by an insulating material.

[0030] 14. The electrode device according to item 12 or item 13, wherein the impedance of the electrode device is between 1 and 10 6 About 10 at Hz -3 Up to 10 5 Ω.

[0031] 15. The electrode device according to item 13, wherein the two or more metal electrode sheets are 2 to 512, such as 8 to 256, such as 15 to 128 metal electrode sheets.

[0032] 16. A prosthetic control system comprising one or more electrode devices according to items 12 to 15. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 depicts a schematic diagram showing capacitive coupling of (a) surface interfaces and (b) internal interfaces. For surfaces without anchoring, the induced charge at the metal layer depends on the transport efficiency of EDL-1 (interface of the metal layer and the ion-conductive gel layer) and EDL-2 (interface of the gel and the stratum corneum). The induced charge at the metal layer increases when anchoring is present. (c) depicts that the PVCL electrode anchored to the SC shows an impedance that is 93% lower than that of the alginate polyacrylamide (Alg-PAAm) electrode without anchoring. The error bars in (c) are the standard deviations of 10 independent experiments.

[0034] Figure 2 Molecular dynamics (MD) simulations of the interaction between the gel electrode and the stratum corneum are depicted. (a) Schematic diagram showing an ion-conducting gel electrode embedded in the stratum corneum, the most superficial layer of the skin (left). The simulated stratum corneum (modeled as a lipid bilayer using free fatty acids (FFA), long-chain ceramides (CER), and cholesterol (CHOL)) and the ion-gel electrode interface (center) show that hydrophobic molecules interact closely with the lipid matrix (right). Due to the molecular anchoring effect of the gel, the coupling area increases, resulting in higher coupling efficiency. (b) Changes in the potential mean force (PMF) with penetration depth. Here, we define z = 0 as the center of the lipid matrix. From the MD results, when the anchoring molecule is anchored to the lipid matrix, an energy barrier ΔG should be overcome. And this ΔG depends on the hydrophobicity of the molecule. (c) Molecular simulations show that among the five monomers (PVCL, polyvinyl pyrrolidone (PVP), acrylic acid, dopamine (DOPA), and alginate), poly (N-vinyl caprolactam) (PVCL) has the strongest interaction with the lipid matrix. To accelerate the structural evolution, simulations were performed using the V-rescale algorithm at 310 K, 340 K, and 360 K. The evolution at 310 K and 360 K at 100 ns is shown.

[0035] Figure 3 Depicted is a schematic diagram for measuring the impedance of electrodes on the skin. The distance between two electrodes (edges) is 1 cm and the size of the electrodes is 0.5 x 0.5 cm (Approval No: SIAT-IRB-180315-H0242).

[0036] Figure 4 Depicted is a graph showing the frequency response of commercial electrodes (from Heal Force Bio-Meditech Holding Limited) and PVCL electrodes on amputee skin from 1 Hz to 10 5 Plot of the system bioelectrical impedance at Hz. The impedance of the PVCL electrode at 1 Hz is almost 1 / 100 of that of the commercial acrylic-based electrode.

[0037] Figure 5 The molecular structures of the lipid matrix are depicted. (a) Computer generated structures of the lipids ceramide (CER), cholesterol (CHOL), and free fatty acids (FFA). (b) Simulation showing a monomer molecule anchored to the lipid matrix.

[0038] Figure 6 PMF and ΔG of five monomers when interacting with SC are depicted. (a)-(e): PMF calculation results of (a) PVCL, (b) PVP, (c) acrylic acid, (d) DOPA and (e) alginate monomers. (f) Among the five monomers, PVCL has the lowest energy barrier (6.57 kBT) when anchored in the lipid matrix. Alginate monomer has the highest energy barrier (ΔG = 40.67 kBT) and is not easily anchored to the lipid matrix.

[0039] Figure 7 The molecular structure evolution of the five monomers during the anchoring process is depicted. The monomer-lipid interaction is accelerated by simulation at high temperature (310K to 360K). Among the five monomers, PVCL is the most deeply anchored to the lipid matrix.

[0040] Figure 8 The time-dependent interactions of the five monomers with the SC are depicted. At 0 ns, no interaction occurs between the monomers and the SC. Except for alginate, all monomers gradually penetrate into the SC within 100 ns, with PVCL showing the fastest penetration. The simulations were performed at 360 K.

[0041] Fig. 9 The dynamic penetration process of five gels into bionic skin is depicted. (a)-(c) The order parameter S at (a) 310K, (b) 340K and (c) 360K z or details of changes in the lipid matrix over time. The inset of (a) depicts the factors that determine S z Obviously, at 310K, the S z are all about 0.65, which means that the permeation process is not a fast process. When the temperature is increased to 340K, as the time increases to 100ns, the S z This indicates that PVCL interacts more strongly with the SC matrix than the other four monomers. For alginate, even when the temperature is increased to 360 K, S zThe results are also always similar to those of the control. Therefore, it is difficult for alginate to penetrate into the SC. (d)-(e) Confocal microscopy images show that PVCL is anchored on the bionic skin at a depth of 3.07 μm (d) while no anchoring of alginate is seen (e). The bionic skin was stained with rhodamine 6G, and the gel was stained with Alexa 488. (f) Quantitative measurement results show that among the five monomers, PVCL has the deepest anchoring to the bionic skin. The error bars are standard deviations based on 10 independent samples.

[0042] Fig.10 The order parameters (S) of lipid matrices of different monomers at 310K, 340K, and 360K are plotted. z ). Figure 6 The PMF calculations in the literature show that PVCL has the lowest S at 340K and 360K. z , indicating that it has the greatest damage to the lipid structure. The error bar is the S in the last 20 ns of the simulation. z The standard deviation of the values.

[0043] Fig.11 Confocal microscopy images depicting the anchoring of PVP, acrylic, and DOPA ion-conducting gels to the SC matrix. The dashed line shows the depth of gel embedding into the SC (silk protein-based biomimetic skin). All gels except alginate were anchored to the SC matrix. PVCL was anchored the deepest at 3 μm.

[0044] Figure 12 depicts the electrical properties of five gel electrodes. (a) Photograph (left) and schematic (right) of a gel electrode consisting of a polydimethylsiloxane (PDMS) substrate and insulator and Cr / Ag / Au metal contact material. (b) Adhesion forces of different electrodes obtained by a standard 90° peel test using bionic skin. Consistent with the simulation study, the PVCL electrode showed the highest adhesion (~189 N / m). (c) Graph showing the adhesion of various electrodes on human skin at 1 Hz to 10 5Hz impedance. (d) Comparison of the impedance and adhesion of our PVCL electrode with different electrodes: 1: CNT / aPDMS (Lee, SM et al., Sci. Rep., 2014, 4, 6074); 2: PDMS_40NW (He, K. et al., ACS Nano, 2022, 16, 9691); 3: PDMS_40NW / Tape (Kim, J.-H. et al., Nano Lett., 2018, 18, 5431), 4: Au / Parylene (Nawrocki, RA et al., Adv. Funct. Mater., 2018, 28, 1803279), 5: Silk (Chen, G. et al., Adv. Mater., 2018, 30, 1800129), 6: Commercial electrode (from Heal Force Bio-Meditech Holding Limited), 7: Silk / Au (Chen, G. et al., Adv. Mater., 2018, 30, 1800129), 8: Fe@Sibione (Jang, KI et al., Adv. Funct. Mater., 2016, 26, 7281), 9: PDA-rGO-PAM (Han, L. et al., Small, 2017, 13, 1601916), 10: a4-PDMS_40NW (Kim, J.-H. et al., Nano Lett., 2018, 18, 5431), 11: Alg-PAAm (Pan, L. et al., Adv. Mater., 2020, 32, 2003723). The impedance decreases as the electrode adhesion increases (curve). The shaded area shows the intrinsic impedance of human skin. The system impedance of the PVCL electrode on the skin (total impedance of the skin, electrode and interface) (14 kΩ) is close to the impedance of human skin (6-10 kΩ). (e) The recording limit of the PVCL electrode on human skin (~1.5% MVC) is close to the lower limit of human muscle contraction (~1% MVC). When the muscle contraction is below 10% MVC, the DOPA, acrylic and alginate electrodes will experience signal loss (shaded area) (all data are recorded at the same amplification of 24, the Y axis is relative intensity and the curve is offset by about 5 mV). (f) When recording low-level (1.7% MVC) electromyographic signals on human skin, the SNR of the PVCL electrode is the best (>5) among the five electrodes. Traditional identification systems require SNR>5. The error bars are standard deviations based on 10 independent samples (all data are recorded at the same amplification of 24).

[0045] Fig.1390° peel tests of PVCL, PVP, alginate, acrylic acid, and DOPA on biomimetic skin are depicted, and show that the adhesion of PVCL is superior to the others. Error bars are standard deviations based on 10 independent samples.

[0046] Figure 14 depicts the 180° peel test of PVCL, PVP, alginate, acrylic acid, and DOPA. Similarly, among the five gels, PVCL has the strongest adhesion. This result is consistent with Fig. 9 b and Fig.13 The results of the 90° peel test shown are in good agreement.

[0047] Figure 15 depicts the impedance of electrodes based on five molecular anchors. (a) Five ion-conducting gels at 1 to 10 4 (b) The system bioelectrical impedance of the five electrodes. The error bars are the standard deviation of 20 samples of gel electrodes.

[0048] Figure 16 depicts dexterous myoelectric control. (a) Photograph of our myoelectric prosthetic control system on an amputated arm. Gel electrodes attached to residual active muscles are connected to a wireless data logging box via flexible cables. The size of the electrode array depends on the size of the muscle being measured. (b) Real-time pixel image of the myoelectric signal obtained from the extensor carpi ulnaris of an amputee using a PVCL electrode array with n = 16 channels (left; size 1 × 16 × 200 ms). The pixel image is input into a convolutional neural network (CNN)-based image recognition architecture (right), which extracts and processes features and classifies them according to the finger and wrist movements initiated by the amputated limb. (c) The myoelectric pixel image (left) is subjected to the mean average (MAV) attention block (center rectangle) shown in (b), returning a feature map (right) that can predict the real-time myoelectric map and theoretically reduce the recognition time from 200 ms to 20 ms. (d) The confusion matrix of G1 to G6 finger and wrist movements obtained using PVCL electrodes at low levels (≤20% MVC) of muscle contraction shows that the CNN algorithm is highly accurate (97.6%). The rows represent the actual classification of G1-G6, while the columns are the predicted classifications. The values ​​of the diagonal elements represent the degree to which the class is correctly predicted. Reset is the electromyographic signal without finger and wrist movement. (e) The recognition accuracy of CNN for low-level signals (≤20MVC; CNN-L) and high-level signals (≥20MVC; CNN-H). (F) The photo shows a myoelectrically controlled robot car arm (prosthesis) of a below-elbow amputee grabbing a candle (1.5 cm in diameter) and placing it in a hole (2 cm in diameter). The control signal source is the real-time electromyographic mapping generated by G1-G6.

[0049] Figure 17 depicts the electromyographic mapping of finger and wrist actions initiated by a severed limb. A photo of an amputee with our electromyographic control system attached to his amputated left arm (left column). The types of finger and wrist actions (G1-G6) initiated by the severed limb (red dashed line) are shown on the right arm as a reference. Pixel images and corresponding feature maps (right column) of myoelectric signals from 100% MVC, ≥20% MVC (high level), ≤20% MVC (low level) for each action used to drive the prosthesis (the arm of the robot car in our study) to release (G1), grasp (G2), move left (G3), move right (G4), move up (G5) and move down (G6). Use the electromyographic mapping of 100% MVC as the source to preliminarily extract the effective electrodes and weights of G1-G6 through the attention block. The feature map is the result of the extraction, which can be used as a preliminary judgment of the real-time electromyographic mapping. In theory, prejudgment can reduce the delay time of the recognition process from 200ms to 20ms. The high-level and low-level EMG maps are used as the source for the next fine-tuning step to further improve the classification accuracy. The grayscale of the EMG map is the relative strength of the muscle contraction.

[0050] Figure 18 depicts the (a) high level (≥20% MVC) and (b) low level (≤20% MVC) unit datasets of G1 used for training. The unit dataset is 2s in duration. To extract features, the dataset is divided into 10 packets, each of which is about 200ms.

[0051] Figure 19 depicts the (a) high level (≥20% MVC) and (b) low level (≤20% MVC) unit datasets of G2 used for training. The unit dataset is 2s in duration. To extract features, the dataset is divided into 10 packets, each of which is about 200ms.

[0052] Figure 20 depicts the (a) high level (≥20% MVC) and (b) low level (≤20% MVC) unit datasets of G3 used for training. The unit dataset is 2s in duration. To extract features, the dataset is divided into 10 packets, each of which is about 200ms.

[0053] Figure 21 depicts the (a) high level (≥20% MVC) and (b) low level (≤20% MVC) unit datasets of G4 used for training. The unit dataset is 2s in duration. To extract features, the dataset is divided into 10 packets, each of which is about 200ms.

[0054] Figure 22 depicts the (a) high level (≥20% MVC) and (b) low level (≤20% MVC) unit datasets of G5 used for training. The unit dataset is 2s in duration. To extract features, the dataset is divided into 10 packets, each of which is about 200ms.

[0055] Figure 23 depicts the (a) high level (≥20% MVC) and (b) low level (≤20% MVC) unit datasets of G6 used for training. The unit dataset is 2s in duration. To extract features, the dataset is divided into 10 packets, each about 200ms. DETAILED DESCRIPTION

[0056] As described above, it has been unexpectedly discovered that the generation and use of ion-conductive gels in the production of gel electrodes can be firmly anchored into the skin, thereby reducing the above-mentioned barrier effect and increasing the ion-electron coupling area.

[0057] Therefore, in a first aspect of the present invention, there is provided an ion-conducting gel comprising:

[0058] a polymeric binding material in an amount of 5 to 20 wt % of the gel;

[0059] a metal salt in an amount of 2 to 10 wt % of the gel; and

[0060] The rest is a hydrophilic polymer material, wherein a gel matrix is ​​formed between the polymer binding material and the hydrophilic polymer material, and the metal salt is distributed throughout the gel matrix.

[0061] In embodiments herein, the word "comprising" may be interpreted as requiring the features mentioned, but not limiting the presence of other features. Alternatively, the word "comprising" may also refer to situations where it is intended that only the listed components / features are present (e.g., the word "comprising" may be replaced by the phrase "consisting of" or "consisting essentially of". It is expressly contemplated that both the broader and narrower interpretations may apply to all aspects and embodiments of the present invention. In other words, the word "comprising" and its synonyms may be replaced by the phrase "consisting of" or the phrase "consisting essentially of" or their synonyms, and vice versa.

[0062] The phrase "consisting essentially of..." and its pseudonyms may be interpreted herein to mean a material in which small amounts of impurities may be present. For example, the material may be greater than or equal to 90% pure, such as greater than 95% pure, such as greater than 97% pure, such as greater than 99% pure, such as greater than 99.9% pure, such as greater than 99.99% pure, such as greater than 99.999% pure, such as 100% pure.

[0063] As used herein, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "a composition" includes mixtures of two or more such compositions, reference to "an oxygen carrier" includes mixtures of two or more such oxygen carriers, and so forth.

[0064] Without wishing to be bound by theory, it is believed that the gel matrix can be formed by intermolecular interactions between the polymer binder material and the hydrophilic polymer material. Such intermolecular interactions can be in the form of hydrogen bonds, van der Waals interactions, etc. As will be appreciated, the gel matrix can be formed by mixing the polymer binder material and the hydrophilic polymer material, and a gel can be formed even if one (or both) of the components is initially in the liquid phase at standard temperature and pressure.

[0065] In embodiments of the invention that may be referred to herein, the number average molecular weight of the hydrophilic polymer material may be 200 to 15,000 Daltons. For example, the number average molecular weight of the hydrophilic polymer material may be 200 Daltons. For the avoidance of doubt, the term "Dalton" is used herein to refer to an atomic mass unit.

[0066] Any suitable hydrophilic polymer material can be used in embodiments of the present invention. Examples of suitable hydrophilic polymer materials include, but are not limited to, polyethylene glycol and polyols. In a specific embodiment of the present invention, the hydrophilic polymer material can be polyethylene glycol. For example, the number average molecular weight that polyethylene glycol can have is 200.

[0067] Any suitable polymer binder can be used in embodiments of the present invention. Examples of suitable polymer binders include, but are not limited to, polysaccharides, cellulose hydrogels, protein hydrogels, nucleic acid gels, biohydrogels, and more particularly, poly(N-vinyl caprolactam) (PVCL), polyvinyl pyrrolidone (PVP), polydopamine, polymerized acrylic acid, and combinations thereof. In more particular embodiments of the present invention that may be mentioned herein, the polymer binder can be selected from one or more of the group consisting of poly(N-vinyl caprolactam) (PVCL), polyvinyl pyrrolidone (PVP), polydopamine, and polymerized acrylic acid. For example, in certain embodiments of the present invention that may be mentioned herein, the polymer binder can be poly(N-vinyl caprolactam) (PVCL).

[0068] In the embodiments disclosed herein, any suitable amount of polymeric binding material may be used, such as 2 to 10 wt%, such as 3 to 5 wt%, etc. In a particular embodiment of the invention, the polymeric binding material may comprise about 10 wt% of the total weight of the gel.

[0069] The metal salt used in the ion conductive gel can be any suitable metal salt. Examples of suitable metal salts include, but are not limited to, iron salts, or more particularly, sodium salts, potassium salts, calcium salts, magnesium salts and combinations thereof. Any suitable metal counter ions can be used, for example, the counter ions can be one or more of the group consisting of fluoride, iodide, chloride, bromide, phosphate, sulfate, etc. Specific salts that can be mentioned include potassium salts (or tissue salts). It is worth noting that potassium salts are generally considered to be present in the body, and therefore can be considered compatible with the intended use of the ion conductive gel. For example, iron phosphate, or more particularly, calcium fluoride, calcium phosphate, potassium phosphate, potassium sulfate and sodium salts. In a specific embodiment of the present invention, the metal salt can be sodium chloride.

[0070] As will be appreciated, the purpose of the metal salt in the ionically conductive gel is to enhance the conductivity of the gel. Therefore, other materials that may be substituted in place of the metal salt are expressly contemplated herein.

[0071] The metal salt may be present in any suitable amount in the ion conductive gel. For example, the metal salt may comprise 5 to 20 wt % of the total weight of the gel, such as 5 to 10 wt % of the total weight of the gel, such as about 5 wt % of the total weight of the gel.

[0072] It has been found that the ion-conductive gel disclosed herein is particularly suitable for forming a flexible electrode device that can be used in a prosthetic device. Therefore, in a second aspect of the present invention, a flexible electrode device is provided, comprising:

[0073] a flexible substrate material having a first surface and a second surface;

[0074] a metal electrode sheet on a first surface of the substrate; and

[0075] An ion conductive gel layer covering the metal electrode sheet, wherein the ion conductive gel is as described above.

[0076] Any suitable material can be used as the flexible substrate material. For example, the flexible substrate material can be a suitable elastomer, such as polydimethylsiloxane (PDMS). As will be understood, the flexible substrate material can also be used as an electrical insulator material.

[0077] The metal electrode sheet can be disposed on a surface (e.g., a first surface) of the flexible substrate material and can be formed of any suitable conductive material or combination of these materials. For example, the metal electrode sheet can be formed of layers of chromium, silver, and gold, wherein the chromium forms a base layer attached to the flexible substrate material and the gold forms a top layer.

[0078] In an embodiment of the second aspect of the invention, the metal electrode sheet may be connected to a conductive path (eg a wire such as a gold wire). In a further embodiment of the invention, the metal electrode sheet and the conductive path (when present) may be covered by an insulating material.

[0079] In an embodiment of the second aspect of the present invention, the device may further include two or more metal electrode sheets located on the first surface of the substrate,

[0080] wherein each of the two or more metal electrode sheets is electrically connected to at least one conductive path,

[0081] wherein each of the two or more metal electrode sheets is covered by an ion conductive gel layer, and

[0082] Each conductive path is covered by an insulating material. Any suitable number of metal electrode sheets may be used herein. For example, there may be 2 to 512, such as 8 to 256, such as 15 to 128 metal electrode sheets.

[0083] The insulating material used herein may be any suitable insulating material showing an appropriate degree of flexibility. For example, the insulating material may be PDMS.

[0084] Compared to conventional flexible electrode devices intended for use in prosthetic control systems and / or prosthetic devices, the flexible electrode devices disclosed herein can exhibit reduced impedance levels. For example, the impedance of the flexible electrode devices is between 1 and 10 6 At Hz, it can be about 10 -3 Up to 10 5 Ω. For example, the flexible electrode device has a 6 Hz can have an impedance of 0.1 to 50 kΩ, such as 1 to 10 6 About 20kΩ at Hz.

[0085] Fabrication of the flexible electrode device may be performed using any suitable technique known in the art using the materials disclosed herein. The following experimental section provides details of how such a flexible electrode device may be fabricated and applied to a prosthetic control system (and thus a prosthetic device).

[0086] In yet another aspect of the invention, a prosthetic control system is provided, comprising one or more electrode devices as previously described herein. The manufacture of the prosthetic control system can be carried out using any suitable technique known in the art using the materials disclosed herein. The following experimental section provides details of how such a control system can be manufactured and applied to a prosthetic device.

[0087] Compared with the traditional improvement of the physicochemical properties of the skin-electrode surface, we found that the use of a gel that can utilize molecular anchoring enables the gel electrode to be embedded in the stratum corneum, thereby weakening the barrier effect of the skin and increasing the higher efficiency of the ion-electron coupling interface. For example, as described in the following example, a gel electrode material using poly (N-vinyl caprolactam) as a polymer adhesive material can provide a gel electrode whose impedance to the skin of an amputee is only about 20 kΩ (which is about 1 / 100 of the impedance of a commercial electrode) and can detect an ultra-low myoelectric signal of 1.5% MVC, which is close to the human body limit. This molecular anchor-based electrode opens up new avenues for dexterous myoelectric control.

[0088] It is believed that the addition of polymer adhesive material allows the molecules to be anchored into the skin, providing the good effects mentioned above and in the examples. The polymer adhesive materials used herein can each have a high adhesion to the skin (e.g., 50 to 250 N / m, such as 75 to 200 N / m, such as 100 to 190 N / m, such as about 189 N / m in a standard 90° peel test using bionic skin).

[0089] Further aspects and embodiments of the invention are discussed in the lettered items below.

[0090] A) Conductive gel for molecular anchoring, comprising:

[0091] (a) 5% NaCl to improve ion conductivity electrode preparation; and

[0092] (b) PVCL, PVP, dopamine and acrylic acid were dissolved in PEG(200) at 10 wt%.

[0093] Alginate gels are prepared in one step by mixing water with sodium alginate, an ionic crosslinker, a covalent crosslinker, acrylamide, and a thermal initiator, as described in the Examples below.

[0094] B) A method for preparing an electrode with ultra-low bioelectronic impedance, comprising:

[0095] a) To generate the PDMS substrate, liquid PDMS was mixed with PMDS curing agent in a ratio of 10: 1. The mixture was then stirred by using a centrifugal mixer until it was homogeneous;

[0096] b) After the mixture is homogeneous, it is placed in a vacuum chamber to eliminate bubbles. A PDMS layer is then deposited on top of the fluorinated silicon wafer;

[0097] c) Fluorinated wafers were prepared by immersing silicon wafers in a 10 mM solution of fluoroalkylsilane (heptadecafluoro-1,1,2,2-tetrahydrodecyltriethoxysilane) in toluene for 30 min and then heat treating at 150 °C for 1 h;

[0098] d) PDMS deposition is done by spin coating at 1000 rpm for 40 seconds on top of the fluorinated wafer. The last step is to let it cure at 60 °C for 4 hours;

[0099] e) depositing a conductive layer of Cr / Ag / Au on top of the PDMS substrate by sputtering technique;

[0100] f) Before sputtering the gold layer, a 5 nm chromium layer was deposited on top of the PDMS to increase the adhesion between the PDMS and the gold layer for better contact;

[0101] g) sputtering a silver layer having a thickness of 400 nm by using a sputtering machine;

[0102] h) depositing a 20 nm layer of gold on the silver to prevent oxidation; and

[0103] i) On top of the gold layer, PDMS was deposited as an insulating layer to prevent the gold connected to the device from collecting any signal from the skin surface.

[0104] Further aspects and embodiments of the invention will now be discussed with reference to the following non-limiting examples.

[0105] Example

[0106] Material

[0107] All reagents were purchased from commercial sources and used without further purification.

[0108] Example 1. Molecular anchoring of poly(N-vinyl caprolactam) (PVCL) gel electrodes on the skin for ultra-low myoelectric signal monitoring

[0109] We designed hydrophobic poly(N-vinylcaprolactam) (PVCL) gel electrodes that are firmly anchored to the skin to reduce the barrier effect and increase the ion-electron coupling area, e.g. Figure 2 a and b.

[0110] Electrode preparation method: PDMS substrate preparation

[0111] To generate the PDMS substrate, liquid PDMS was mixed with PDMS curing agent in a ratio of 10:1. The mixture was then stirred by using a centrifugal mixer until it was homogeneous. After the mixture was homogeneous, it was placed in a vacuum chamber to eliminate bubbles. The PDMS layer was then deposited on top of the fluorinated silicon wafer. The fluorinated wafer was made by immersing the silicon wafer in a 10mM solution of fluoroalkylsilane (heptadecafluoro-1,1,2,2-tetrahydrodecyltriethoxysilane) in toluene for 30 minutes and then heat treating it at 150°C for 1 hour. The purpose of fluorination is to generate a hydrophobic surface. The deposition of PDMS was done by spin coating at 1000rpm for 40 seconds on top of the fluorinated wafer. The last step was to let it cure at 60°C for 4 hours.

[0112] Electrode preparation method: Deposition of Cr / Ag / Au and insulator

[0113] The deposition of the conductive layer Cr / Ag / Au on top of the PDMS substrate was done by sputtering technique (by the method disclosed in Pan, L. et.al., Adv. Mater., 2020, 32, 2003723). Very thin layers (on the scale of nm) are required to maintain its conductive compliance even when it is stretched.

[0114] Before sputtering the gold layer, a 5nm chromium layer is deposited on top of the PDMS to increase the adhesion between the PDMS and gold layers, thereby obtaining a better contact. Then, a silver layer with a thickness of 400nm is sputtered by using a sputtering machine. Finally, a 20nm gold layer is deposited on the silver layer to prevent oxidation. A mask is used during the sputtering process to control and pattern the deposition of gold in the desired areas. Next, on top of the gold layer, PDMS is deposited as an insulating layer to prevent the gold connected to the device from collecting any signals from the skin surface.

[0115] Electrode preparation method: ion conductive gel preparation

[0116] For all gels, we introduced 5% NaCl to improve ionic conductivity. PVCL, PVP, dopamine, and acrylic acid were dissolved in polyethylene glycol (PEG) 200 at 10% wt%. Alginate gels were prepared in one step by mixing water with sodium alginate, ionic crosslinkers, covalent crosslinkers, acrylamide, and thermal initiators, according to our previous work (Pan, L. et.al., Adv. Mater., 2020, 32, 2003723).

[0117] It is noteworthy that after the PVCL, PVP, dopamine, and acrylic acid polymer materials were dissolved into PEG 200, a gel was formed.

[0118] System impedance of electrodes and skin

[0119] The system impedance was measured by Zahner Scientific Instruments. During the test, two 1×1cm 2 The electrodes are attached to the skin at a distance of ~1 cm. The frequency range is 1 Hz to 10 5 Hz.

[0120] Results and discussion

[0121] The increased coupling area leads to a higher coupling probability between ion flux and electron flow, thus enhancing the ion-electron coupling efficiency. Our measurements on amputee skin ( Figure 3 ) showed that the impedance of the PVCL electrode was about 20 kΩ@1 Hz, which is almost 1 / 100 of the impedance of the commercial acrylic-based electrode ( Figure 4 ), and is 93.4% lower than that of alginate polyacrylamide (Alg-PAAm) electrodes (Pan, L. et al., Adv. Mater., 2020, 32, 2003723), which shows the lowest bioelectronic impedance on normal skin ( Figure 1c ). This suggests that PVCL gel electrodes can be used to monitor ultra-low (down to 1.5% MVC) myoelectric signals in amputees under dexterous prosthetic control.

[0122] Example 2. Molecular dynamics (MD) simulation of the molecular anchoring process of PVCL, polyvinylpyrrolidone (PVP), acrylic acid, dopamine (DOPA) and alginate

[0123] Inspired by these measurements, we performed molecular dynamics (MD) simulations to understand how PVCL molecules interact with and anchor to SC. We also studied 4 other adhesive molecules: polyvinylpyrrolidone (PVP), acrylic acid, dopamine (DOPA), and alginate to explain the molecular anchoring strategy.

[0124] Modeling of Monomer Structure

[0125] We studied the gel monomers of five types of molecules, namely PVCL, PVP, acrylic acid, dopamine (DOPA), and alginate, as shown in Figure 1b. The structures of the monomers were obtained using the CHARMM-GUI web service (https: / / www.charmm-gui.org / ).

[0126] Modeling the skin surface

[0127] According to previous studies, the surface layer of the skin, the stratum corneum (SC), is modeled by the lipid matrix of the SC (Gupta, R. et al., J. Phys. Chem. B, 2016, 120, 8987).

[0128] According to previous studies, the lipid matrix of SC was constructed by CHARMM-GUI web service using a heterogeneous mixture of long-chain ceramide (CER), cholesterol (CHOL), and free fatty acids (FFA) in a 1:1:1 ratio, as Figure 2 a and Figure 5 As shown (Harding, CR, Dermatol. Ther., 2004, 17, 6; Notman, R. & Anwar, J., Adv. Drug Deliv. Rev., 2013, 65, 237). The lipid matrix was constructed to be approximately 10×10 nm in size. 2 To determine the interaction between monomers and SC, Figure 5 As shown, 700 monomers were randomly added to the simulation box.

[0129] Molecular dynamics simulation method

[0130] Molecular dynamics (MD) simulations were performed using the CHARMM36 force field (Klauda, ​​JB et al., J. Phys. Chem. B, 2010, 114, 7830) using the GROMACS software package (Van Der Spoel, D. et al., J. Comput. Chem. 2005, 26). TIP3P (Jorgensen, W. Let al., The J. Chem. Phys., 1983, 79, 926) water molecules were used to solvate the system, and an appropriate amount of Na + ions to neutralize the system. The simulation was performed under periodic boundary conditions. All graphics and visualization analyses were performed by the Visual Molecular Dynamics VMD software package (Humphrey, W. et al., J. Mol. Graph., 1996, 14, 33).

[0131] Results and discussion

[0132] Consistent with the intrinsic hydrophobicity of the lipid matrix (composed of FFA, CER, and CHOL), the calculated potential of mean force (PMF) indicated that more hydrophobic monomers have a lower energy barrier (ΔG) to anchor to the lipid matrix and will be more easily embedded ( Figure 2 b).

[0133] Among the five monomers, PVCL showed the lowest ΔG (6.57 k BT). PVCL and PVP are widely used in biology and medicine and have good biocompatibility and chemical stability (Heinzmann, C. et al., Chem. Sov. Rev., 2016, 45, 342). Although PVP and PVCL have high hydrophobicity and homologous structures, the additional methylene groups in PVCL make it more hydrophobic than PVP. Therefore, the octanol-water partition coefficient (log P) of PVCL is 1.75, which is greater than the octanol-water partition coefficient of PVP (0.74) ( Figure 6 ). Acrylic acid, DOPA, and alginate are suitable adhesives and are widely used in electrodes for electromyographic recording. The Log P values ​​of acrylic acid (0.44), DOPA (-2.2), and alginate (-3.84) are much lower than those of PVP and PVCL. Alginate, which has the lowest hydrophobicity, shows the highest energy barrier (ΔG = 40.67 k B T), and are least able to anchor to the lipid matrix.

[0134] Example 3. Combining molecular dynamics (MD) simulations with the V-rescale algorithm at different temperatures to accelerate structural evolution

[0135] In addition, we investigated the interactions between the five monomers and the lipid matrix by MD simulations combined with the V-rescale algorithm at three temperatures of 310 K, 340 K, and 360 K to accelerate the structural evolution.

[0136] Combination of molecular dynamics (MD) simulation and V-rescale algorithm

[0137] MD simulations were performed according to the protocol in Example 2.

[0138] In order to accelerate the structural evolution, the simulation was carried out at three temperatures, namely 310K, 340K and 360K, respectively, in combination with the V-rescale algorithm (Bussi, G. et al., J. Chem. Phys., 2007, 126, 014101). These systems were first balanced in the NPT ensemble, where the pressure was coupled under semi-isotropic conditions at 1 bar using the Parrinello-Rahman method (Parrinello, M. & Rahman, A., J. Appl. Phys., 1981, 52, 7182). In the equilibrium simulation, position constraints were imposed on all heavy atoms of the monomer and lipid matrix. After about 1ns of equilibrium simulation, all production simulations were carried out in the NVT ensemble. The LINCS algorithm is used to constrain the covalent bonds of hydrogen atoms. The time step is set to 2fs. The cutoff value of non-bonded interactions is set to 1.2nm. The long-range electrostatic interactions were calculated by the Particle Mesh Ewald (PME) method.

[0139] Umbrella Sampling

[0140] To calculate the potential mean force (PMF) of different monomers on the lipid matrix of the skin, the umbrella sampling method was used. In the umbrella sampling method, the reaction coordinate is defined as the distance between the center of mass (COM) of the monomer and the lipid matrix in the normal direction, as shown in Figure S5. To construct the structure of the system with different reaction coordinates, the PLUMED package was applied to add the monomers to the corresponding reaction coordinates. In the umbrella sampling simulation, the size of the lipid matrix was about 5.4 × 5.4 nm 2 . The spacing of the sampling windows was 0.1 nm. 41 windows were constructed for each system. In each window, 20 ns of MD simulation were performed to ensure convergence, resulting in umbrella sampling using a total simulation time of 800 ns in each monomer. The spring constant of the positional constraint in the umbrella sampling method was 1660.5 pN / nm. The umbrella sampling results were analyzed using the weighted histogram analysis method (WHAM).

[0141] Calculation of order parameters of lipid matrices

[0142] To describe how monomers affect the lipid matrix structure, we calculated the lipid matrix order parameter S z , as follows (Equation (1)):

[0143]

[0144] where θ is a vector (average orientation of the lipid matrix) and (The alkyl chain starts from C n+1 to C n-1 The brackets <·> indicate the angles averaged over the carbon atoms in the alkyl chain of the lipid.

[0145] Generally speaking, S z = 1 corresponds to perfect alignment of lipids in the matrix with respect to the normal direction; S z = 0 indicates random arrangement of lipids in the matrix; S z =1 / 2 indicates that the lipids are arranged perpendicular to the normal direction of the matrix.

[0146] Results and discussion

[0147] Further MD simulations combined with the V-rescale algorithm at three temperatures (310 K, 340 K, and 360 K) to accelerate the interaction also showed that PVCL was most effectively and quickly embedded into the SC ( Figure 2 c. Figure 8 and Fig. 9 ). Figure 2 c shows the interactions between the five molecules and the lipid matrix from 0 ns to 100 ns at 310 K to 360 K. The strength and depth of the interactions increase with increasing hydrophobicity of the monomers.

[0148] Calculation of the order parameter S of the lipid matrix z , to describe how monomers affect the lipid matrix structure. At 310 K, the ordered lipid matrix without monomers (control) has a high S on the 100 ns simulation time scale. Z (≈0.65)( Fig. 9 a)(See Fig.10 Since the molecular anchoring process is a long-time process, the order parameter of the lipid matrix remains at about 0.65 in the 100 ns simulation at 310 K, which is similar to the case without the anchoring process. When interacting with the PVCL monomer at 340 K, S Z Reduced to 0.072( Fig. 9 b), indicating that the monomer significantly disrupts the order of the lipid matrix ( Figure 2 c) For all other monomers, S Z The remaining ≈ 0.55. When the system is heated up to 360K, these monomers have different effects on the lipid matrix structure ( Fig. 9 (c) and Fig.10 ).like Fig.10 As shown, the simulation concluded that PVCL has the highest ability to anchor to the skin among the five molecules.

[0149] Example 4. Penetration process of gel electrode on bionic skin

[0150] Penetration Testing

[0151] First, the bionic skin was immersed in rhodamine 6G (0.002 mg / mL, solvent was ethanol) for 72 hours. Then, the bionic skin was rinsed with ethanol three times before drying. Second, the ion-conductive gel was mixed with Alexa 488 (0.004 mg / mL, solvent was dimethyl sulfoxide (DMSO)) for 24 hours. Third, the stained gel was transferred to the bionic skin with rhodamine 6G. We used a Leica confocal microscope to observe the penetration process.

[0152] Results and discussion

[0153] Confocal microscopy images also showed that PVCL was anchored into the bionic skin made of silk protein at a depth of 3.07 μm, which was deeper than that of PVP (2.43 μm), acrylic acid (1.18 μm), DOPA (0.97 μm), and alginate (0 μm). Fig. 9 df and Fig.11 ). These results indicate that among the five monomers, the most hydrophobic PVCL monomer has the best anchoring effect on the skin.

[0154] Example 5. Preparation of gel electrode and evaluation of mechanical and electrical properties

[0155] Based on the simulation results, we prepared five gel electrodes using these monomers and evaluated their mechanical and electrical properties (Figure 12).

[0156] Preparation of gel electrodes

[0157] Gel electrodes were prepared according to the protocol in Example 1. We used PDMS as the substrate and insulator for all electrodes. To improve the ionic conductivity, we added 5% NaCl to all gels.

[0158] Signal-to-Noise Ratio (SNR) equation (Equation (2))

[0159] SNR = 10log 10 (P 信号 / P 噪声 ) (2)

[0160] P 信号 is the average pixel value

[0161] P 噪声 is the standard deviation or error value of the pixel values.

[0162] Results and discussion

[0163] Figure 12a shows our electrode structure, with PDMS used as the substrate and insulator for the electrode to conform well to the curved amputee skin.

[0164] The adhesion between the electrode and the skin is macroscopic evidence of molecular anchoring due to the strong interaction between the gel and the SC matrix ( Figure 2 (b)). Figure 6 f), the standard 90° peel test on the bionic skin made of silk protein showed that the adhesion of PVCL was the largest (~189 N / m; Figures 12b and Fig.13 ). Similarly, in the 180° peel test, the PVCL gel still outperformed the other gels (Figure 14). Although all five electrodes had an intrinsic impedance of 200-300Ω at 1Hz ( Fig.15a ), but the PVCL electrode showed the lowest impedance on normal (~14 kΩ) and amputee (~20 kΩ) skin at 1 Hz (Fig. 12c, Figure 1c and Fig.15b). These values ​​are closer to the intrinsic impedance of human skin (6-10 kΩ) than those achieved by previous electrodes, allowing detection of low-level (1.5% MVC) EMG signals close to the human limit (~1% MVC) (Figure 12d-e). The key reason for this is the high ion-electron coupling efficiency of the EDL formed by the molecular anchoring process. The impedance of the original gel without NaCl at 1 Hz was 10 7 Taken together, these results suggest that anchoring PVCL to SC improves ionic-electronic coupling and significantly reduces impedance on the skin.

[0165] When the other four types of electrodes monitored ultra-low levels of muscle contraction (below 10% MVC), severe signal loss due to high bioelectrical impedance always occurred. As a basic parameter describing the electrode fidelity, we measured the signal-to-noise ratio of the recorded EMG signals below 30% MVC (Figure 12(f)). Except for the PVCL electrode, whose SNR was 5.02 for 1.7% MVC signals and 3.725 for 1.5% MVC, all other electrodes had an SNR below 5 for <2% MVC signals. Traditional recognition algorithms based on raw EMG signals, such as linear discriminant analysis (LDA), cannot be used to ensure high recognition accuracy for weak activities of phantom limbs of amputees. Since the signal source for myoelectric control requires an SNR above 5, the PVCL electrode has the potential to detect ultra-low EMG signals for fine prosthetic control.

[0166] Example 6. Image recognition algorithm based on convolutional neural network (CNN)

[0167] Lecun first demonstrated the effectiveness of convolutional neural networks (CNN) in handwritten digit recognition on the MNIST dataset (http: / / yann.lecun.com / exdb / mnist / ). Convolutional networks are believed to have the ability to extract spatial features of images.

[0168] Convolutional Neural Networks (CNN)

[0169] To perform image recognition using a convolutional neural network (CNN), we converted the EMG signals into images by mapping them in 1 × n × m pixels, where n is the number of channels on the electrode array and m is the time required to record each muscle contraction of the intended finger and wrist movement (in our work, m is 200 ms). The size of the electrode array depends on the size of the monitored muscle (the maximum data processing capacity of our system is n 最大 As training signal sources, we defined the EMG signals from low-level (≤20% MVC), high-level (≥20% MVC), and maximum-level (100% MVC) muscle contractions.

[0170] By designing the Convnet model, the basic and complex features of EMG mapping can be detected and learned through convolution operations. The basic Convnet consists of several layers: conv2D layer, maxpooling layer, dense layer, dropout layer, batch normalization layer and softmax layer.

[0171] The ReLU function is used, which includes the nonlinear opponent of the network. Adaptive moment estimation (Adam) is selected as the network optimizer. The structure of Convnet is shown in Figure 16b. For the conv3d layer, the number of filters of the first conv2D layer is set to 16 and the kernel size is set to 4×2; the number of filters of the second conv2D layer is set to 32, the kernel size is 3×3, and the size of the third kernel is 2×2. The first densely connected layer is set to have 64 units, and the second is 10 units, which is equal to the number of actions classified.

[0172] Results and discussion

[0173] Our CNN-based EMG recognition includes the following layers: extracting features from the EMG (conv2D layer), reducing the parameters of the model (max pooling layer), accelerating network convergence (batch normalization layer), preventing overfitting (discarding layer), and outputting predicted tensors (dense connection and softmax layers) (Figure 16b). The conv2d layer is mainly used to perform convolution operations, which extract features from the EMG map. The max pooling layer performs a downsampling operation and reduces the parameters of the model. The batch normalization layer prevents gradient vanishing and makes the network converge faster. The discarding layer randomly inactivates neurons, which solves the overfitting problem. Through the dense connection layer and the softmax layer, Convnet finally outputs a predicted tensor, the number of which is equal to the number of categories.

[0174] Example 7. Introducing attention mechanism in electromyography mapping

[0175] To improve the processing efficiency and classification accuracy of ultra-low EMG signals with low (<5) SNR, we designed an attention block (MAV block) that contributes to the network in two ways: (1) it evaluates the importance of each channel and enhances channels with obvious features while suppressing channels without obvious features; and (2) it extracts MAV features from different channels of EMG signals and retains these features for transfer learning, so that new tasks can be accurately performed with shorter training time (Figure 16c). Since the same muscles are used in different finger and wrist actions, the attention block evaluates and weights each muscle channel by measuring the average value (MAV). This block increases the weight of important channels and reduces the weight of unimportant channels.

[0176] Attention Block

[0177] In the general task of attention block, people usually focus on the different and similar features between source and target domains. Similar features should be preserved, and differences should be learned and calibrated. Even a small change in muscle contraction strength in the task can have a huge impact on gesture classification. Due to the independence of EMG gesture samples with various contraction strengths, it is difficult for the CNN network to learn obvious different causes of the shift, so we need to let the network pay more attention to the similar features between samples with different motion forces and use fine-tuning to calibrate the network.

[0178] Assume that the EMG signal has n channels and each channel has m sampling points. It can be regarded as an input image (Equation (3)).

[0179] I∈R 1×m×n (3)

[0180] For a channel, the mean value (MAV) is first calculated (Equation (4)).

[0181]

[0182] Then, the original MAV value is obtained by equation (5).

[0183] F r =[mav p ,mav p ,..,mav p ] (5)

[0184] After passing through two MLP layers and activated by the sigmoid function, eight weight vectors are obtained through equation (6),

[0185] F r =σ(MLP(F))∈R 1×n (6)

[0186] Where σ represents the sigmoid function.

[0187] Finally, the output image is expressed by equation (7),

[0188] I aug =F scale (F r ,I) (7)

[0189] where F scale Represents column-form multiplication.

[0190] Results and Discussion

[0191] Attention mechanism has been widely used in deep learning to improve the performance of the network and achieve higher classification accuracy. We designed an attention block named MAV block. The structure of MAV block is shown in Figure 16c.

[0192] After extracting the same features from the 100% MVC EMG, we obtain a series of activated electrodes with associated weights for each finger and wrist action. The 100% MVC map is used to exclude the influence of signals with low SNR. Using the output of the MAV block extraction (called feature map) as the initial judgment of the real-time EMG mapping, the delay time in the recognition process can theoretically be reduced from 200ms to 20ms.

[0193] Example 8. PVCL-based gel electrode array for measuring low-level electromyographic signals

[0194] As a proof of concept, we built PVCL-based gel electrode arrays and used them to measure low-level myoelectric signals from the residual limbs of amputees. The electrodes were connected to a wireless data logging box via flexible cables, and the measured myoelectric signals were used to control a prosthetic arm, which in our work is represented by a robotic car (Figure 16). Here, we focus on using low-level myoelectric signals to achieve dexterous finger movements, such as grasping with a prosthetic limb.

[0195] Preparation of gel electrodes

[0196] Gel electrodes were prepared according to the protocols in Examples 1 and 5.

[0197] Image recognition via CNN

[0198] CNN-based myoelectric mapping recognition is performed according to the scheme in Example 6. By following the scheme in Example 7 to subject the myoelectric pixel image to the MAV attention block, the processing efficiency and classification accuracy of ultra-low myoelectric signals with low (<5) SNR are improved.

[0199] Myoelectric control of a robotic car arm driven by a below-elbow amputee

[0200] We applied this system to enable below-elbow amputees to grasp a candle and place it into a hole. For amputees, using prosthetic fingers to grasp small objects with reasonable force and approach the target with high accuracy is a common daily activity.

[0201] To demonstrate the accuracy of prosthetic control, the diameter of the hole (2 cm) was made only slightly larger than the diameter of a candle (1.5 cm). Six finger and wrist actions (G1-G6) (Figure 17) triggered by the severed limb were used to drive the robot car's arm to release (G1, Figure 18), grasp (G2, Figure 19), move left (G3, Figure 20), move right (G4, Figure 21), move up (G5, Figure 22), and move down (G6, Figure 23).

[0202] Results and discussion

[0203] Among all 5 electrodes, the PVCL electrode showed the highest CNN recognition accuracy (97.6%) for low-level signals (Figures 16d-e and Tables 1 and 2), enabling the robotic arm to successfully grasp the candle and place it into the hole (Figure 16f). Between 22.5s and 36.4s, the robotic car used ultra-low-level and high-quality EMG mapping to deftly approach the small hole. In order for the robotic arm to grasp the candle, the strength of the muscle contraction of the intended action generated by the severed limb must match the strength of the normal finger action. If it does not match, grasping does not occur. In summary, our results show that PVCL electrodes firmly anchored to the skin can detect low-level EMG signals that are suitable for CNN-based recognition and fine prosthetic control.

[0204] Table 1. Confusion matrix for high-level muscle contraction (≤20% MVC) using CNN

[0205] G1 G2 G3 G4 G5 G6 Reset G1 107 0 0 0 0 0 4 G2 0 96 0 0 0 0 0 G3 0 0 87 0 0 0 0 G4 0 0 0 88 0 0 3 G5 0 0 0 0 75 0 0 G6 0 0 0 0 3 73 0 Reset 0 2 2 2 4 1 95

[0206] Table 2. Confusion matrix for high-level muscle contraction (≥20% MVC) using CNN

[0207] G1 G2 G3 G4 G5 G6 Reset G1 107 0 0 0 0 0 4 G2 0 96 0 0 0 0 0 G3 0 0 87 0 0 0 0 G4 0 0 0 88 0 0 3 G5 0 0 0 0 65 0 0 G6 0 0 0 0 3 73 0 Reset 0 2 2 2 4 3 95

[0208] In summary, we proposed a strategy to improve the detection and fidelity of low-level EMG signals by developing PVCL gel electrodes anchored to the SC. This strong anchoring enhanced ionic-electronic coupling and minimized the impedance on the amputated skin (down to 20 kΩ). The overall mechanical and electrical performance of our electrodes exceeded that of other gel electrodes (Table 3), allowing the detection of ultra-low (1.5% MVC) EMG signals close to the human limit. We designed a CNN-based mapping recognition algorithm that can classify and predict the expected finger and wrist movements with 97.6% accuracy based on the EMG signals obtained using our PVCL electrodes. With such high-fidelity signals and classification accuracy, our system enables amputees to drive prostheses with finger dexterity. Such fine movements are expected to improve the quality of life of amputees who rely on prostheses to complete daily tasks.

[0209] Table 3. Mechanical and electrical properties of five gel electrodes.

[0210]

Claims

1. An ion conductive gel comprising: a polymeric binding material in an amount of 5 to 20 wt % of the gel; a metal salt in an amount of 2 to 10 wt % of the gel; as well as The rest is a hydrophilic polymer material, wherein a gel matrix is ​​formed by the polymer binding material and the hydrophilic polymer material, and the metal salt is distributed throughout the gel matrix.

2. The ion conductive gel according to claim 1, wherein The number average molecular weight of the hydrophilic polymer material is 200 to 15,000 Daltons.

3. The ion conductive gel according to claim 1 or claim 2, wherein: The hydrophilic polymer material is polyethylene glycol or polyol.

4. The ion conductive gel according to claim 3, wherein: The hydrophilic polymer material is polyethylene glycol, optionally wherein the polyethylene glycol has a number average molecular weight of 200.

5. An ion-conducting gel according to any one of the preceding claims, wherein The polymer binding material is selected from one or more of the group consisting of polysaccharides, cellulose hydrogels, protein hydrogels, nucleic acid gels, biohydrogels, and more particularly poly(N-vinyl caprolactam) (PVCL), polyvinylpyrrolidone (PVP), polydopamine and polymerized acrylics.

6. The ion conductive gel according to claim 5, wherein The polymer binding material is selected from one or more of the group consisting of poly(N-vinyl caprolactam) (PVCL), polyvinyl pyrrolidone (PVP), polydopamine and polymerized acrylic.

7. The ion conductive gel according to claim 6, wherein The polymer adhesive material is poly(N-vinyl caprolactam) (PVCL).

8. An ion-conducting gel according to any one of the preceding claims, wherein The polymeric binding material comprises about 10 wt % of the total weight of the gel.

9. An ionically conductive gel according to any one of the preceding claims, wherein The metal salt is selected from one or more of the group consisting of iron salts, sodium salts, potassium salts, calcium salts and magnesium salts.

10. The ion conductive gel according to claim 9, wherein The metal salt is sodium chloride.

11. An ionically conductive gel according to any one of the preceding claims, wherein The metal salt accounts for about 5 wt % of the total weight of the gel.

12. A flexible electrode device, comprising: a flexible substrate material having a first surface and a second surface; a metal electrode sheet on a first surface of the substrate; as well as An ion conductive gel layer covering the metal electrode sheet, wherein the ion conductive gel is as described in any one of claims 1 to 11.

13. The electrode device according to claim 12, wherein: The device further comprises two or more metal electrode sheets on the first surface of the substrate, wherein each of the two or more metal electrode sheets is electrically connected to at least one conductive path, wherein each of the two or more metal electrode sheets is covered by an ion conductive gel layer, and Therein, each conductive path is covered by an insulating material.

14. The electrode device according to claim 12 or claim 13, wherein: The impedance of the electrode device is between 1 and 10 6 About 10 at Hz -3 Up to 10 5 Ω.

15. The electrode device according to claim 13, wherein: The two or more metal electrode sheets are 2 to 512, such as 8 to 256, such as 15 to 128 metal electrode sheets.

16. A prosthetic control system comprising one or more electrode devices according to claims 12 to 15.

Images