Wearable electrical stimulation system based on multi-modal functional information closed-loop regulation

Abstract

The invention discloses a wearable electrical stimulation system based on multi-modal functional information closed-loop regulation and control, and the system comprises a signal collection and preprocessing module which is used for collecting and preprocessing a near-infrared signal and a surface electromyogram signal, and obtaining a multi-modal data set; the data analysis and decision module is used for establishing a personalized electrical stimulation prescription matched with the multi-modal data, forming an expert knowledge base based on clinical experience, training a multi-modal data set by using a DNN-sEMG-NIRS model, outputting the personalized electrical stimulation prescription, and generating or updating an electrical stimulation driving instruction according to a closed-loop algorithm; and the electrical stimulation execution module is used for receiving the electrical stimulation driving instruction and driving the high-precision constant-current electrical stimulator to apply precise electrical stimulation to the specific muscle or nerve part of the patient. According to the invention, the near-infrared signal, the surface electromyogram signal change and the electrical stimulation parameter adjustment process are presented in real time in an intuitive and understandable form, so that medical personnel can master the treatment progress at any time, make professional judgment quickly and optimize the treatment strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical stimulation, and in particular, to a wearable electrical stimulation system based on closed-loop regulation of multimodal functional information. Background Art

[0002] Functional electrical stimulation (FES) is a treatment method that improves or restores body functions through electrical stimulation and is currently widely used in fields such as stroke, spinal cord injury, and rehabilitation of chronic diseases in the elderly. Frequency, intensity, pulse width, waveform, and duty cycle are key parameters for electrical stimulation treatment. Clinically, the muscle is usually stimulated at a low frequency of 2 - 10 Hz to prevent muscle atrophy; the intensity is selected according to the patient's tolerance, treatment site, and treatment purpose; the pulse width refers to the duration of a single electrical pulse, and in low-frequency electrical stimulation, the pulse width for muscle contraction is generally between 0.1 - 1 ms. The square wave is the most common stimulation waveform with steep rising and falling edges, which can quickly cause neuromuscular excitation. The duty cycle refers to the interval time between two electrical stimulations. An appropriate intermittent time allows the muscle to have enough time to recover, and the duty cycle is usually 1:2 or 1:3.

[0003] To address the inherent drawbacks of traditional electrical stimulation technology, such as bulky equipment and a large subjective component in the treatment process, the current update of functional electrical stimulation devices shows two major trends: the transformation from traditional large-scale wired devices to small-scale wireless connections; and the replacement of open-loop passive regulation with closed-loop active regulation that dynamically feedbacks the muscle state based on real-time physiological signals of the human body.

[0004] In traditional large-scale wired electrical stimulation devices, the electrical stimulation generator uses a power source to generate electrical pulse signals with specific waveforms, frequencies, and intensities. The signals are transmitted through wires to electrodes that come into contact with the human body. The electrodes are mostly made of conductive metals or rubbers. After closely fitting the skin, the current flows into the subcutaneous tissue, stimulates the excited tissue, causes changes in the cell membrane potential, prompts the nerve to generate action potentials and conduct them to the effector organ, or directly causes the muscle to contract, thereby achieving the effects of rehabilitation and treating diseases. During the process, the patient's activities are restricted due to the wired connection. Currently, the traditional open-loop electrical stimulation used clinically mainly manually adjusts the electrical stimulation parameters by the physician based on the physician's experience and the patient's subjective feedback of the muscle state. This may lead to a lack of standardization in the treatment plan, and the patient's subjective feedback is interfered by various factors, which easily causes inaccurate parameter adjustment, resulting in muscle over-fatigue and affecting the treatment effect. Summary of the Invention

[0005] Based on this, to solve the problem of inaccurate parameter adjustment in traditional open-loop electrical stimulation, the present invention provides a wearable electrical stimulation system based on closed-loop regulation of multimodal functional information.

[0006] The present invention provides a wearable electrical stimulation system based on closed-loop regulation of multimodal functional information, including:

[0007] A signal acquisition and preprocessing module is used to acquire near-infrared signals and surface electromyography signals, and preprocess the near-infrared signals and surface electromyography signals to obtain a multimodal data set;

[0008] The data analysis and decision-making module is used to establish a personalized electrical stimulation prescription that matches the multimodal data through clinical expert consultation for each sample of the multimodal data set, form an expert knowledge base based on clinical experience, train the multimodal data set using the DNN-sEMG-NIRS model, output a personalized electrical stimulation prescription that complies with the principles of evidence-based medicine, and generate or update the electrical stimulation drive instructions according to the closed-loop algorithm;

[0009] The electrical stimulation execution module is used to receive the electrical stimulation driving instruction and drive the high-precision constant current electrical stimulator to apply precise electrical stimulation to specific muscles or nerve parts of the patient.

[0010] A high-precision electromyography sensor is used to collect surface electromyography signals of the target muscle group, and a near-infrared sensor is used to collect near-infrared signals. The collected near-infrared signals and surface electromyography signals are preprocessed to obtain digital signals that can be analyzed and processed by the data analysis and decision-making module. The digital signals are sent to the data analysis and decision-making module in a wireless transmission manner via a Bluetooth module.

[0011] The acquisition of the surface electromyography signal includes:

[0012] Surface electromyographic signals are collected using differential electrode pairs, vias with differential signal networks are reserved on the PCB board, and electrode female buckles are welded on the vias to connect the male buckles on the electrode sheet, and the two electrode female buckles are symmetrically distributed about the midline;

[0013] The surface electromyography signal enters the sEMG signal preprocessing circuit from the differential electrode pair via the in-phase input terminal +IN pin and the inverting input terminal -IN pin of the sEMG signal preprocessing circuit. The voltage difference between the two electrodes is first amplified by an instrument amplifier to obtain a differential signal, the differential signal is amplified by a gain resistor, and then AC coupling is performed through a capacitor to generate the surface electromyography signal.

[0014] The acquisition of the near infrared signal includes:

[0015] The three channels of the near-infrared light source generator are controlled by the I / O ports of the single-chip microcomputer to be alternately selected. The LED interfaces LED1, LED2, and LED3 corresponding to three different wavelengths are respectively connected to three independent I / O ports of the single-chip microcomputer. The single-chip microcomputer outputs a high-level signal to the corresponding I / O port to drive the corresponding triode to conduct, so that a working potential difference is formed between the positive and negative ends of the near-infrared light source corresponding to the wavelength, realizing the selection control of the light source; the ambient light sensor is used to capture the near-infrared signal and convert the near-infrared signal into an electrical signal, and the electrical signal is preprocessed by a second-order voltage-controlled voltage source low-pass filter, and finally the single-chip microcomputer reads the AD value of the near-infrared signal.

[0016] Based on the multi-modal dataset, the DNN-sEMG-NIRS model is introduced, and the network weight parameters are optimized through an end-to-end training paradigm, so that the DNN-sEMG-NIRS model has the ability to automatically extract physiological signal features from the original multi-modal dataset, and a non-linear mapping relationship between the multi-modal data and the electrical stimulation prescription is established.

[0017] The training of the DNN-sEMG-NIRS model includes,

[0018] Data cleaning is performed on the multi-modal dataset to ensure that each multi-modal data includes the original surface electromyogram signal and near-infrared signal, and each surface electromyogram signal and near-infrared signal corresponds to the electrical stimulation prescription given by the doctor.

[0019] The operations for data cleaning of the multi-modal dataset include,

[0020] Band-pass filtering with a frequency range of 20 - 450HZ is added to the surface electromyogram signal to remove noise, and low-pass filtering is added to the near-infrared signal to remove high-frequency noise;

[0021] The surface electromyogram signal and near-infrared signal after noise processing are aligned in the time domain, and the signal amplitudes of the surface electromyogram signal and near-infrared signal are normalized to the interval [0, 1];

[0022] The surface electromyogram signal and near-infrared signal are respectively input into two different branches of the DNN-sEMG-NIRS model, the outputs of the two different branches are spliced at the fusion layer, and the output layer outputs the electrical stimulation prescription corresponding to the input multi-modal signal.

[0023] After the collected surface electromyography (sEMG) signals and near-infrared (NIRS) signals are preprocessed, they are transmitted to the data analysis and decision-making module through a Bluetooth module. The data analysis and decision-making module extracts the time-domain eigenvalue iEMG and frequency-domain eigenvalue MF of the sEMG signals through a feature extraction algorithm, and the key features of the light absorption change rate, oxyhemoglobin concentration, and blood volume change of the NIRS signals at different wavelengths. After obtaining the time-domain eigenvalue iEMG, frequency-domain eigenvalue MF, the light absorption change rate of the NIRS signals at different wavelengths, oxyhemoglobin concentration, and blood volume change, an intelligent prescription generation DNN-sEMG-NIRS model driven by a trained deep neural network is introduced. The DNN-sEMG-NIRS model automatically maps out personalized electrical stimulation prescriptions, and the parameter values of the electrical stimulation prescriptions include stimulation intensity, stimulation frequency, and duty cycle. According to the electrical stimulation prescriptions, corresponding instructions for controlling electrical stimulation are generated.

[0024] The data analysis and decision-making module also performs real-time closed-loop evaluation on the multi-modal data. An evaluation is carried out every 5 s, and the evaluation duration is 1 s. When it is detected that the muscle state deviates from the sEMG signal threshold corresponding to the current prescription, the data analysis and decision-making module automatically updates the electrical stimulation parameters according to the current muscle state and verifies the rationality of the adjustment of the electrical stimulation parameters using an expert knowledge base.

[0025] Beneficial effects: The present invention deeply integrates two modal signals of near-infrared and electromyography. The NIRS signals characterize the overall blood oxygen state of the muscle, and the sEMG signals characterize the detailed characteristics of the muscle. The multi-dimensional signals complement each other, providing rich physiological data for precise rehabilitation treatment. The changes in NIRS signals, sEMG signals, and the process of adjusting electrical stimulation parameters are presented in an intuitive and easy-to-understand form in real time, facilitating medical staff to grasp the treatment progress at any time, quickly make professional judgments based on the visualized data, and optimize the treatment strategy.

[0026] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings are used to better understand the solution and do not constitute a limitation to the present invention. Among them:

[0028] Figure 1 is a system operation architecture diagram provided by the present invention;

[0029] Figure 2 is a detection structure diagram of sEMG signals and NIRS signals provided by the present invention;

[0030] Figure 3 is a physical structure schematic diagram of an electrical stimulation execution module provided by the present invention;

[0031] Figure 4 is the structural diagram of the edge computing module for myoelectric near-infrared perception decoding provided by the present invention;

[0032] Figure 5 is the physical structure diagram of the electrical stimulation execution module provided by the present invention;

[0033] Figure 6 is the circuit diagram of sEMG signal preprocessing provided by the present invention;

[0034] Figure 7 is the circuit diagram of three-channel near-infrared gating provided by the present invention;

[0035] Figure 8 is the NIRS signal optoelectronic conversion and preprocessing circuit provided by the present invention;

[0036] Figure 9 is the human-computer interaction interface diagram provided by the present invention;

[0037] Figure 10 is the electrical stimulation circuit diagram provided by the present invention;

[0038] Figure 11 is the architecture diagram of the data analysis and decision-making module provided by the present invention;

[0039] Figure 12 is the physical diagram of the system provided by the present invention. Detailed implementation manners

[0040] The following describes exemplary embodiments of the present invention with reference to the accompanying drawings. Various details of the embodiments of the present invention are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted below.

[0041] The present invention provides a wearable electrical stimulation system based on closed-loop regulation of multimodal functional information. Its specific implementation is as Figure 12 shown, where 17 is the system protection cover, 18 is the display screen of the computing terminal, 19 is the electrical stimulation charging compartment, 20 is the power supply port, 21 is the data interface, and 22 is the charging compartment for the myoelectric and near-infrared acquisition device, including:

[0042] The system operation architecture diagram is as Figure 1 shown, mainly including a signal acquisition and preprocessing module, a data analysis and decision-making module, and an electrical stimulation execution module.

[0043] A signal acquisition and preprocessing module is used to acquire near-infrared signals and surface electromyography signals, and preprocess the near-infrared signals and surface electromyography signals to obtain a multimodal data set;

[0044] The present invention combines surface electromyography (sEMG) and near infrared spectroscopy (NIRS) to detect muscle activity status.

[0045] The surface electromyographic signals of the target muscle group are collected using a high-precision electromyographic sensor, and the near-infrared signals are collected using a near-infrared sensor to characterize key information such as muscle contraction strength and fatigue level. The collected near-infrared signals and surface electromyographic signals are preprocessed to obtain digital signals that can be analyzed and processed by the data analysis and decision module, and the digital signals are sent to the data analysis and decision module in a wireless transmission manner through a Bluetooth module.

[0046] The acquisition of the surface electromyography signal includes:

[0047] Surface electromyographic signals are collected using differential electrode pairs. Vias with differential signal networks are reserved on the PCB board, and electrode female buckles are welded on the vias to connect the male buckles on the electrode sheets. The two electrode female buckles are symmetrically distributed about the midline. The physical distance between the centers of the two electrode female buckles is 28 mm. The symmetrical distribution can effectively offset the interference in differential signal processing and improve the signal-to-noise ratio.

[0048] Surface electromyography signals are obtained from the differential electrode pair through a sEMG signal preprocessing circuit (such as Figure 6 The in-phase input terminal +IN pin and the inverting input terminal -IN pin of the sEMG signal preprocessing circuit (as shown) first amplify the differential signal between the two electrodes through an instrumentation amplifier, and then perform AC coupling through a capacitor to generate the surface electromyography signal.

[0049] In the sEMG signal preprocessing circuit, the amplification gain G is related to the gain resistor R8. The present invention uses a 2.2kΩ gain resistor to amplify the above differential signal by 200 times, and then uses a 3.3uf capacitor to play the role of AC coupling. The front-stage circuit and the rear-stage operational amplifier circuit are isolated on the DC to prevent the DC level of the front-stage circuit from affecting the working point of the rear-stage operational amplifier, ensuring that the operational amplifier can work normally under appropriate DC bias. In the rear-stage operational amplifier circuit, a reference voltage is introduced at the R pin in the figure to provide a stable reference benchmark for signal processing, and finally the AD value at the label "sEMG" is read by the single-chip microcomputer.

[0050] The acquisition of the near infrared signal includes:

[0051] The three channels of the near-infrared light source generator are controlled by the I / O ports of the single-chip microcomputer to be alternately selected. The LED interfaces LED1, LED2, and LED3 corresponding to three different wavelengths are respectively connected to three independent I / O ports of the single-chip microcomputer. The single-chip microcomputer outputs a high-level signal to the corresponding I / O port to drive the corresponding triode to conduct, so that a working potential difference is formed between the positive and negative ends of the near-infrared light source corresponding to the wavelength, realizing the selection control of the light source. The ambient light sensor is used to capture the near-infrared signal and convert the near-infrared signal into an electrical signal. The electrical signal is preprocessed by a second-order voltage-controlled voltage source low-pass filter, and finally the single-chip microcomputer reads the AD value of the near-infrared signal.

[0052] Figure 2 Shows the detection structure diagrams of surface electromyogram signals and near-infrared signals. Figure 2 In it, 1 and 2 represent two reserved ports for electrode female buttons, 3 represents the near-infrared light source port, and 4 represents the ambient light sensor interface. Figure 3 Shows the physical structure diagram of the electrical stimulation execution module. 5 and 6 represent the reserved ports for electrical stimulation output electrode buttons, 7 represents the reserved port for the emergency stop case, and 8 represents the charging port. Figure 4 Is the structure diagram of the edge computing module for myoelectric near-infrared perception decoding. 9 and 10 are a pair of differential electrodes for measuring surface electromyogram signals. A via with a network and an outer diameter of 5 mm is reserved on the PCB, and the electrode female button is welded on this via.

[0053] As Figure 4 Shown, the surface electromyogram acquisition electrodes and the near-infrared probe are distributed in a "cross" shape on the bottom layer of the PCB board. During operation, the muscle information in the cross-shaped area is collected. It is stipulated that the sampling frequency of the surface electromyogram signal is 1000 Hz, and the sampling frequency of the near-infrared signal is 150 Hz. The collected data is sent to the data analysis and decision-making module through the Bluetooth module MX-02S.

[0054] Figure 4 In it, 11 is the near-infrared light source, which is a three-channel common-cathode near-infrared light source generator with a size of 3.5 mm × 2.8 mm, and its wavelength parameters are 735 nm, 810 nm, and 850 nm respectively. Through the selection circuit as Figure 7 Shown, it is controlled by the I / O port of the single-chip microcomputer to be alternately selected. In the figure, LED1, LED2, and LED3 are respectively connected to three independent I / O ports of the single-chip microcomputer. When it is necessary to select a near-infrared light source with a specific wavelength, the single-chip microcomputer outputs a high-level signal to the corresponding I / O port to drive the corresponding triode to conduct, so that a working potential difference is formed between the positive and negative ends of the near-infrared light source of this wavelength, realizing the selection control of the light source. The selection time for each channel is 20 ms, and the interval time between adjacent channels is also 20 ms. Figure 4 In it, 12 is the ambient light sensor OPT101, which is used to capture the near-infrared signal and convert the optical signal into an electrical signal recognizable by the single-chip microcomputer. AsFigure 8 As shown, when near-infrared photons irradiate the photosensitive area of OPT101, an electrical signal that is positively correlated with the light intensity is formed. The electrical signal passes through a second-order voltage-controlled voltage source low-pass filter composed of R25, R26, R27, C29, and C30 to preprocess the electrical signal, and finally the AD value of the read output pin 1 of the single-chip microcomputer is obtained. Since the propagation path of near-infrared photons in human tissue is approximately crescent-shaped, after multiple scatterings, they finally return to the surface of human tissue. In order to monitor the blood oxygen information of deep muscles in the present invention, the distance between the ambient light sensor and the light source emitter is designed to be about 25 mm. Figure 4 In Figure 13, 13 is a near-infrared light filter covering the ambient light sensor to avoid interference from light sources of other wavelengths.

[0055] Figure 5 Figure 14 shows the physical structure diagram of the electrical stimulation execution module. Among them, 14 and 16 are electrical stimulation output ports, and 17 is an emergency stop switch. Its main function is to quickly cut off the output of the electrical stimulation signal in case of an emergency to ensure the safety of the user.

[0056] The data analysis and decision-making module is used to conduct a consultation on each sample of the multi-modal data set by clinical experts, establish a personalized electrical stimulation prescription that matches the multi-modal data, form an expert knowledge base based on clinical experience, use the DNN-sEMG-NIRS model to train the multi-modal data set, output an individualized electrical stimulation prescription that conforms to the principles of evidence-based medicine, and generate or update electrical stimulation drive instructions according to the closed-loop algorithm;

[0057] As Figure 11 shown, the data analysis and decision-making module is a system computing center with the NXP IMX6ULL processor as the core. The DDR3L SDRAM memory chip provides operating memory support, the SLC NAND FLASH storage chip is responsible for data storage, and the BMS power management system ensures stable power supply. The video interface chip Sii9022 is connected to the RGB touch screen for display and interaction. The CH342 serial port chip is paired with a USB Bluetooth receiver to achieve serial communication and Bluetooth connection functions. Overall, it constitutes a computing system with storage, display, interaction, and communication capabilities.

[0058] The data analysis and decision-making module is externally connected to a dual-serial-port Bluetooth receiver to identify the sEMG and NIRS signals transmitted by different information detection modules through the USB3.0 protocol. The data analysis and decision-making module also includes as Figure 9The shown human-machine interaction interface mainly functions to visually process physiological signal features and select an electrical stimulation mode as active control or passive control. Signal channels to be displayed can be selected on the right side, and a two-dimensional coordinate on the left side is used for display, facilitating the user to judge whether the electrode mounting is effective. The "Mode Selection" button can switch the electrical stimulation between active control and passive control. Active control of electrical stimulation means that the electrical stimulation automatically updates the electrical stimulation parameters according to the input physiological signal features, while passive control means that the user or professional physician manually sets the electrical stimulation parameters according to personal experience and usage habits.

[0059] The data analysis and decision-making module has collected a multi-modal muscle function multi-modal dataset of a large number of patients with motor dysfunction. The multi-modal dataset covers the following core bio-information features: surface electromyogram (sEMG), including integrated electromyogram value (iEMG) and median frequency (MF); near-infrared spectroscopy (NIRS), dynamic change parameters of oxygenated hemoglobin concentration (Δ[HbO2]) and blood volume (BV).

[0060] Based on the multi-modal dataset, a DNN-sEMG-NIRS model is introduced. Through an end-to-end training paradigm, the network weight parameters are optimized, enabling the DNN-sEMG-NIRS model to automatically extract physiological signal features from the original multi-modal dataset and establish a non-linear mapping relationship between the multi-modal data and the electrical stimulation prescription.

[0061] The training of the DNN-sEMG-NIRS model includes,

[0062] Clean the multi-modal dataset to ensure that each multi-modal data includes the original surface electromyogram signal and near-infrared signal, and each surface electromyogram signal and near-infrared signal corresponds to an electrical stimulation prescription given by a doctor.

[0063] Clean the dataset, including operations such as deleting duplicate information, filling in missing data, and discarding data. Ensure that each data includes the original surface electromyogram (sEMG) and near-infrared (NIRS) signals, and each signal corresponds to an electrical stimulation prescription given by a doctor. Add a band-pass filter of 20 - 450 Hz to the sEMG signal to remove noise, add a low-pass filter to the NIRS signal to remove high-frequency noise; and align the sEMG signal and the NIRS signal in the time domain; normalize the signal amplitude to the interval of [0, 1].

[0064] The operations for cleaning the multi-modal dataset include,

[0065] Add a band-pass filter of 20 - 450 HZ to the surface electromyogram signal to remove noise, and add a low-pass filter to the near-infrared signal to remove high-frequency noise;

[0066] Align the surface electromyogram (sEMG) signal and near-infrared (NIRS) signal after noise processing in the time domain, and normalize the signal amplitudes of the sEMG signal and NIRS signal to the interval [0, 1].

[0067] Input the sEMG signal and NIRS signal into two different branches of the DNN-sEMG-NIRS model respectively, splice the outputs of the two different branches in the fusion layer, and the output layer outputs an electrical stimulation prescription corresponding to the input multimodal signal.

[0068] After the collected sEMG signal and NIRS signal are preprocessed, they are transmitted to the data analysis and decision-making module through a Bluetooth module. The data analysis and decision-making module extracts the time-domain eigenvalue iEMG and frequency-domain eigenvalue MF of the sEMG signal through a feature extraction algorithm, and the key features of the light absorption change rate, oxyhemoglobin concentration, and blood volume change of the NIRS signal at different wavelengths. After obtaining the time-domain eigenvalue iEMG, frequency-domain eigenvalue MF, light absorption change rate of the NIRS signal at different wavelengths, oxyhemoglobin concentration, and blood volume change, introduce the trained intelligent prescription generation DNN-sEMG-NIRS model driven by a deep neural network. The DNN-sEMG-NIRS model automatically maps out a personalized electrical stimulation prescription. The parameter values of the electrical stimulation prescription include stimulation intensity, stimulation frequency, and on-off ratio, and generate a corresponding instruction to control the electrical stimulation according to the electrical stimulation prescription.

[0069] The data analysis and decision-making module also performs real-time closed-loop evaluation on the multimodal data, conducts an evaluation every 5 s, and the evaluation duration is 1 s. When it detects that the muscle state deviates from the sEMG signal threshold corresponding to the current prescription, the data analysis and decision-making module automatically updates the electrical stimulation parameters according to the current muscle state and verifies the rationality of the adjustment of the electrical stimulation parameters using an expert knowledge base.

[0070] When iEMG decreases by 15% and BV increases by 20% (a decrease in iEMG indicates a weakening of muscle electrical activity, and an increase in BV indicates that more blood flows to the muscle area), the system automatically updates the electrical stimulation parameters according to the current muscle state and verifies the rationality of the adjustment of the electrical stimulation parameters using an expert knowledge base. Compared with the prescription before adjustment, the adjustment amplitude each time does not exceed 10%, the amplitude does not exceed 40 mA, and the minimum update interval is set to 30 s. Ensure that the parameter adjustment is within a safe range. Achieve a "perception - decision - intervention" closed loop.

[0071] An electrical stimulation execution module, which is used to receive the electrical stimulation driving instruction and drive a high-precision constant current electrical stimulator to apply precise electrical stimulation to specific muscles or nerve sites of the patient.

[0072] The electrical stimulation execution module drives a high-precision constant-current electrical stimulator to apply precise electrical stimulation to specific muscles or nerve sites of the patient according to the serial port protocol. The electrical stimulator adopts advanced constant-current source technology to ensure the stability and precision of the stimulation current, and avoid discomfort or injury to the patient caused by current fluctuations.

[0073] During the electrical stimulation process, near-infrared and surface electromyography signals are continuously received through the feedback loop to monitor the electrical stimulation effect in real time, dynamically optimize the electrical stimulation parameters, and form a complete closed-loop control system.

[0074] The physical structure diagram and physical structure diagram of the electrical stimulation execution module are respectively as Figure 3 、 Figure 5 shown. The weight of a single device is 6.2 g. On the PCB board, two 5-mm vias are reserved in advance for soldering the 4.0-mm electrode female buckle. These two vias are electrically connected to the output ports 14 and 16 of the electrical stimulation respectively. Thus, it can ensure that the electrical signal is smoothly conducted from the inside of the circuit board to the electrode female buckle, and then to the physiotherapy electrode patch, and finally the current flows through the human body to form a complete loop. 17 is an emergency stop switch, whose main function is to quickly cut off the output of the electrical stimulation signal in case of emergency to ensure the safety of the user. The electrical stimulation circuit is as Figure 10 shown. The electrical stimulation circuit selects ENS001 as the main control. The peripheral circuit includes a crystal oscillator, a power inductor, a decoupling capacitor, a power isolation circuit, a Boot startup circuit. The chip built-in 32-bit ARM Cortex-M0 CPU core, a boost circuit, a constant-current circuit, and can realize the control of the electrical stimulation current, frequency, and pulse width through programming. The device realizes receiving the control instructions sent by the terminal wirelessly through the MX-02S Bluetooth module. After the main control receives the control instructions through the serial port, it adjusts the electrical stimulation parameters accordingly. The specific parameters are shown in Table 1.

[0075] Table 1 Electrical Stimulation Parameter Table

[0076] Parameter type Parameter range (unit) Output current 0~60 (mA) Pulse width 10~1000 (uS) Stimulation frequency 1~10 (KHz) Stimulation waveform Square wave, triangular wave, sine wave

[0077] The adjustment accuracy of the output current of the electrical stimulation circuit reaches 1 μA, the pulse width is adjusted according to the percentage of the stimulation unit, and the adjustment accuracy of the stimulation frequency reaches 1 Hz.

[0078] To more clearly illustrate the implementation process of the technical solution of the present invention, the following will be specifically described in combination with typical clinical application scenarios:

[0079] In the initial stage of rehabilitation of a stroke patient, the limb muscle strength is weak and the movement coordination is poor. When using the functional electrical stimulator with closed-loop control of wearable multi-modal functional information, first attach the lightweight physiological signal acquisition device to the limb muscle groups that need rehabilitation of the patient through Ag / AgCl electrode patches, such as the biceps brachii and quadriceps femoris. After the device is started, the surface electromyogram and near-infrared acquisition modules start to work: the high-precision electromyogram sensor real-time collects the weak electrical signals generated by muscle contraction, realizes 200-fold amplification through the instrumentation amplifier AD8236, and then passes through a 3.3uf capacitor for AC coupling to remove DC interference and generate a stable surface electromyogram signal; at the same time, the three-channel common-cathode near-infrared light source generator (wavelengths are 735nm, 810nm, 850nm respectively) is sequentially and alternately gated, each channel is gated for 20ms, and the adjacent channels are spaced 20ms apart. The ambient light sensor OPT101 captures the near-infrared signal scattered back by the human tissue at a distance of about 25mm from the light source emitter and converts it into an electrical signal to monitor the deep muscle blood oxygen information and comprehensively reflect the real-time state of the muscle.

[0080] After the collected electromyogram and near-infrared signals are preprocessed, they are transmitted to the data analysis and decision-making module through the Bluetooth module MX - 02S. The data analysis and decision-making module introduces an intelligent prescription generation DNN-sEMG-NIRS model driven by a trained deep neural network (a comprehensive model that combines deep neural network, surface electromyogram and near-infrared spectroscopy technology) to quickly generate personalized electrical stimulation prescriptions. For example, according to the current state of the patient's extremely weak muscles and insufficient blood oxygen supply, the model gives electrical stimulation parameters of low intensity (such as 10mA), low frequency (such as 2Hz), and relatively wide pulse width (such as 500uS), the stimulation waveform is a square wave, and the on-off ratio is 1:3 to gently stimulate muscle contraction while avoiding excessive fatigue.

[0081] After receiving the instruction, the electrical stimulation execution module accurately applies electrical stimulation to the patient's muscles through the electrical stimulation output ports 14 and 16 (connected to the 4.0mm electrode female buckle welded on the 5mm via of the PCB board, and then connected to the physiotherapy electrode patch). During the electrical stimulation process, the physiological signal acquisition module continuously and real-time monitors the electromyogram and near-infrared signals as a feedback loop. Once it is found that the muscle fatigue intensifies or the blood oxygen changes abnormally, the electrical stimulation parameters are immediately dynamically optimized. For example, after several minutes of stimulation, it is found that the muscle contraction force has increased to a certain extent but signs of fatigue begin to appear. The system automatically adjusts to a lower frequency (such as 1Hz) and appropriately reduces the pulse width (such as 400uS) to ensure the safe and efficient progress of rehabilitation training. The patient or medical staff can also visually view the changes in muscle physiological signals through the human-computer interaction interface and select the electrical stimulation mode according to actual needs. If the patient feels that the current active closed-loop regulation effect is good, maintain the active control mode; if the medical staff believes that fine-tuning is needed based on professional experience, it can be switched to passive control. Figure 9Manually input the electrical stimulation parameters on the right interface shown. The data analysis and decision-making module generates corresponding serial port instructions for the lower computer according to the serial port protocol to adjust the electrical stimulation parameters.

[0082] Muscle recovery of athletes after the game

[0083] After intense competitions, athletes' muscles are often in a state of fatigue or even minor injuries. When using this electrical stimulator for assistance in recovery, the device is worn on the fatigued muscle parts, such as the gastrocnemius muscle in the leg, the deltoid muscle in the upper limb, etc. The edge computing module quickly collects the electromyogram and near-infrared signals of the muscle. At this time, due to muscle fatigue, the frequency domain eigenvalue MF of the surface electromyogram signal decreases, and the near-infrared signal shows a decrease in muscle blood oxygen saturation, a decrease in the concentration of oxyhemoglobin, and an increase in the concentration of deoxyhemoglobin.

[0084] After the signals are transmitted to the data analysis and decision-making module, an electrical stimulation prescription for the rapid recovery of athletes is generated based on rich characteristic data and intelligent models. Usually, it is a medium-intensity (such as 30 mA), specific frequency (such as 5 Hz) stimulation, the pulse width is finely adjusted according to the degree of muscle fatigue (such as 300 μS), the stimulation waveform is selected as a sine wave, and the on-off ratio is set to 1:2 to promote muscle blood circulation, accelerate the excretion of metabolic waste, and relieve fatigue.

[0085] The electrical stimulation execution module accurately executes the instructions, ensures the stable output of the stimulation current through stable constant current source technology, and avoids discomfort caused by fluctuations to the sensitive muscles of athletes. During the electrical stimulation process, real-time feedback enables the system to flexibly adjust parameters according to changes in muscle status. For example, if an athlete reports a strong sense of muscle soreness in a certain part, the system analyzes the real-time collected signals, increases the electrical stimulation intensity of that part to 35 mA, and slightly adjusts the frequency to 4 Hz at the same time to optimize the recovery effect. The human-computer interaction interface enables athletes to easily understand the muscle recovery process and independently select the electrical stimulation mode. If they want the system to automatically optimize according to the body's real-time reaction, they can turn on the active control; if they are used to adjusting manually according to past experience, they can switch to the passive control for manual operation.

[0086] In several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0087] The unit(s) may or may not be physically separated. The component(s) presented as a unit may be one physical unit or multiple physical units, i.e., they may be located in one place or distributed to multiple different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0088] In addition, in each embodiment of the present invention, each functional unit may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated unit(s) can be implemented in the form of hardware or in the form of software functional unit(s).

[0089] If the above-mentioned integrated unit(s) is / are implemented in the form of software functional unit(s) and sold or used as an independent product, it can be stored in a readable storage medium. Based on such understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions for causing a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, optical disks, and other various media that can store program codes.

[0090] The above are only the specific implementation manners of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A wearable electrical stimulation system based on closed-loop control of multimodal functional information, characterized in that: include: A signal acquisition and preprocessing module is used to acquire near-infrared signals and surface electromyography signals, and preprocess the near-infrared signals and surface electromyography signals to obtain a multimodal data set; The data analysis and decision-making module is used to establish a personalized electrical stimulation prescription that matches the multimodal data through clinical expert consultation for each sample of the multimodal data set, form an expert knowledge base based on clinical experience, train the multimodal data set using the DNN-sEMG-NIRS model, output a personalized electrical stimulation prescription that complies with the principles of evidence-based medicine, and generate or update the electrical stimulation drive instructions according to the closed-loop algorithm; The electrical stimulation execution module is used to receive the electrical stimulation driving instruction and drive the high-precision constant current electrical stimulator to apply precise electrical stimulation to specific muscles or nerve parts of the patient.

2. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 1, characterized in that: A high-precision electromyography sensor is used to collect surface electromyography signals of the target muscle group, and a near-infrared sensor is used to collect near-infrared signals. The collected near-infrared signals and surface electromyography signals are preprocessed to obtain digital signals that can be analyzed and processed by the data analysis and decision-making module. The digital signals are sent to the data analysis and decision-making module in a wireless transmission manner via a Bluetooth module.

3. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 2, characterized in that: The acquisition of the surface electromyography signal includes: Surface electromyographic signals are collected using differential electrode pairs, vias with differential signal networks are reserved on the PCB board, and electrode female buckles are welded on the vias to connect the male buckles on the electrode sheet, and the two electrode female buckles are symmetrically distributed about the midline; The surface electromyography signal enters the sEMG signal preprocessing circuit from the differential electrode pair via the in-phase input terminal +IN pin and the inverting input terminal -IN pin of the sEMG signal preprocessing circuit. The voltage difference between the two electrodes is first amplified by an instrument amplifier to obtain a differential signal, the differential signal is amplified by a gain resistor, and then AC coupling is performed through a capacitor to generate the surface electromyography signal.

4. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 3, characterized in that: The acquisition of the near infrared signal includes: The three channels of the near-infrared light source generator are controlled by the I / O port of the single-chip microcomputer to be selected in turn, and the LED interfaces LED1, LED2, and LED3 corresponding to the three different wavelengths are respectively connected to the three independent I / O ports of the single-chip microcomputer. The single-chip microcomputer outputs a high-level signal to the corresponding I / O port to drive the corresponding transistor to be turned on, so that a working potential difference is formed at the positive and negative ends of the near-infrared light source of the corresponding wavelength, thereby realizing the selection control of the light source; the ambient light sensor is used to capture the near-infrared signal and convert the near-infrared signal into an electrical signal, which is pre-processed by a second-order voltage-controlled voltage source low-pass filter, and finally the AD value of the near-infrared signal is read by the single-chip microcomputer.

5. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to any one of claims 2 to 4, characterized in that: Based on the multimodal dataset, a DNN-sEMG-NIRS model is introduced, and the network weight parameters are optimized through an end-to-end training paradigm, so that the DNN-sEMG-NIRS model has the ability to automatically extract physiological signal features from the original multimodal dataset and establish a nonlinear mapping relationship between the multimodal data and the electrical stimulation prescription.

6. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 5, characterized in that: The training of the DNN-sEMG-NIRS model includes: The multimodal data set is cleaned to ensure that each multimodal data includes original surface electromyography signals and near-infrared signals, and each surface electromyography signal and near-infrared signal corresponds to an electrical stimulation prescription given by a doctor.

7. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 6, characterized in that: The operation of performing data cleaning on the multimodal data set includes: Add 20-450HZ bandpass filtering to the surface electromyography signal to remove noise, and add low-pass filtering to the near-infrared signal to remove high-frequency noise; The surface electromyography signal and the near infrared signal after the noise processing are aligned in the time domain, and the signal amplitudes of the surface electromyography signal and the near infrared signal are normalized to the interval of [0, 1]; The surface electromyography signal and the near infrared signal are respectively input into two different branches of the DNN-sEMG-NIRS model, the outputs of the two different branches are spliced ​​in the fusion layer, and the output layer outputs the electrical stimulation prescription corresponding to the input multimodal signal.

8. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 7, characterized in that: The collected surface electromyographic signals and near-infrared signals are preprocessed and transmitted to the data analysis and decision-making module through the Bluetooth module. The data analysis and decision-making module extracts the time domain eigenvalue iEMG and frequency domain eigenvalue MF of the surface electromyographic signals, the key features of the light absorption change rate of the near-infrared signals at different wavelengths, the oxygenated hemoglobin concentration and the blood volume change through a feature extraction algorithm; after obtaining the time domain eigenvalue iEMG, the frequency domain eigenvalue MF, the light absorption change rate of the near-infrared signals at different wavelengths, the oxygenated hemoglobin concentration and the blood volume change, a trained deep neural network-driven intelligent prescription generation DNN-sEMG-NIRS model is introduced. The DNN-sEMG-NIRS model automatically maps out a personalized electrical stimulation prescription. The parameter values ​​of the electrical stimulation prescription include stimulation intensity, stimulation frequency and on-off ratio, and the corresponding electrical stimulation control instructions are generated according to the electrical stimulation prescription.

9. A wearable electrical stimulation system based on closed-loop control of multimodal functional information according to claim 8, characterized in that: The data analysis and decision-making module also performs real-time closed-loop evaluation on the multimodal data, with an evaluation every 5 seconds and an evaluation duration of 1 second. When it is detected that the muscle state deviates from the multimodal physiological signal threshold corresponding to the current prescription, the data analysis and decision-making module automatically updates the electrical stimulation parameters according to the current muscle state and uses the expert knowledge base to verify the rationality of the adjustment of the electrical stimulation parameters.

Images