Multi-modal multi-acupoint nerve function regulation system and collaborative stimulation method

Abstract

The invention provides a multi-mode multi-acupoint neural function regulation system and a collaborative stimulation method, and relates to the technical field of neural function regulation device.Neurophysiological data of central nerves and peripheral nerves of a user are collected and comprise electroencephalogram signals, electromyographic signals and / or multi-acupoint bio-electricity signals; inputting the neurophysiological data into a pre-trained machine learning model, and outputting a corresponding neurophysiological state type; obtaining multi-modal energy stimulation parameters of a plurality of preset acupuncture points from a preset mapping table according to the neurophysiological state type, and carrying out collaborative stimulation on the preset acupuncture points through the stimulation parameters; and carrying out real-time monitoring on the neurophysiological data after collaborative stimulation until a preset neurophysiological target threshold value is reached. The problem that stimulation parameters in the prior art are mostly fixed preset values and cannot be adjusted according to different neurophysiological states of a user is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of nerve function regulation devices, in particular to a multi-modal multi-acupoint nerve function regulation system and a collaborative stimulation method. BACKGROUND

[0002] In the field of nerve function regulation, the demand for non-drug intervention for chronic insomnia, anxiety, chronic neuralgia and other symptoms continues to grow, but the existing regulation devices have outstanding technical problems in terms of dynamic parameter adaptation and individual difference matching. The stimulation parameters of the current mainstream devices are mostly fixed preset mode values before leaving the factory, and only uniform standard output energy can be output, which cannot respond to the dynamic changes of the user's neurophysiological state in real time - for example, when the α wave proportion in the user's brain electrical signal increases from 15% to 22% (i.e. insomnia improvement trend), or the amplitude of the acupoint bioelectricity signal fluctuates, the device cannot adjust the stimulation intensity, frequency and other key parameters according to these real-time data.

[0003] This one-size-fits-all parameter setting mode results in a serious mismatch between the stimulation scheme and the different user's neural characteristics, such as central nervous excitability difference and different peripheral nerve sensitivity: some users may be uncomfortable due to excessive parameters, and some users may not achieve the regulation effect due to insufficient parameters, ultimately making it difficult to achieve precise and effective nerve function regulation for individual differences, and unable to meet the individualized intervention needs of users with different symptoms and different constitutions. SUMMARY

[0004] Embodiments of the present application provide a multi-modal multi-acupoint nerve function regulation system and a collaborative stimulation method, aiming to solve the problem that the stimulation parameters of the prior art are mostly fixed preset values and cannot be adjusted according to different user's neurophysiological states.

[0005] To achieve the above-mentioned purpose, in a first aspect, the present application provides a multi-modal multi-acupoint nerve function regulation system, comprising a data acquisition module, a machine learning analysis module, a multi-modal stimulation module and a result feedback module, the data acquisition module, the machine learning analysis module, the multi-modal stimulation module and the result feedback module are communicatively connected, wherein: The data acquisition module is configured to acquire neurophysiological data of the user's central nervous system and peripheral nervous system, wherein the neurophysiological data includes brain electrical signals, electromyographic signals and / or multi-acupoint bioelectricity signals; The machine learning analysis module is configured to input the neurophysiological data into a pre-trained machine learning model and output a corresponding neurophysiological state type; the machine learning model is configured as a SOM model; The multi-modal stimulation module is configured to obtain multi-modal energy stimulation parameters of a plurality of preset acupoints from a predetermined mapping table according to the neurophysiological state type, and perform collaborative stimulation on the preset acupoints through the stimulation parameters. The result feedback module is configured to monitor the neurophysiological data after the synergistic stimulation in real time until the neurophysiological data reaches a predetermined neurophysiological target threshold.

[0006] Further, the plurality of preset acupoints include Baihui acupoint, Sishencong acupoint, Fengchi acupoint, Fengfu acupoint, Yifeng acupoint, frontal lobe, perineum acupoint, Yongquan acupoint and / or foot bottom brain reflex area.

[0007] Further, the machine learning analysis module is further configured a feature optimization unit, which is configured to perform wavelet transform denoising on the collected neurophysiological data and extract core features of the neurophysiological data, the core features including α wave proportion and θ wave power spectral density of electroencephalogram signal, signal amplitude standard deviation and recruitment time of electromyogram signal, and peak amplitude and pulse frequency of multi-acupoint bioelectricity signal.

[0008] Further, the multi-modal stimulation module includes an energy sub-module group and an acupoint positioning unit, the energy sub-module group including a low-frequency magnetic stimulation sub-module, a near-infrared laser sub-module and / or an ultrasonic wave sub-module; the acupoint positioning unit is internally provided with a pressure sensor and an infrared locator, which are configured to detect the fitting degree of the plurality of preset acupoints in real time, and trigger an audible and light prompt when the fitting degree is lower than a preset fitting degree threshold.

[0009] Further, the result feedback module is further configured a dynamic threshold adjustment unit and an abnormality early warning unit; the dynamic threshold adjustment unit is configured to correct the predetermined neurophysiological target threshold according to user age and basic disease type; the abnormality early warning unit is configured to trigger the multi-modal stimulation module to pause output and send early warning information to the system when the neurophysiological data after the synergistic stimulation is monitored to be abnormal.

[0010] Further, the predetermined mapping table is a dynamic updating mapping table, and the machine learning analysis module is further configured a data iteration unit, which is configured to store the neurophysiological data, neurophysiological state type, stimulation parameter and adjustment result in each synergistic stimulation process into a database, for retraining and optimization of the SOM model, and for synchronously updating the mapping relationship between the neurophysiological state type and the multi-modal energy stimulation parameter in the dynamic updating mapping table.

[0011] Further, the multi-modal stimulation module is further configured a timing control unit, which is configured to set the output time sequence of the multi-modal energy according to the physiological correlation logic of the preset acupoints.

[0012] Further, the machine learning analysis module is further configured a weight adjustment unit, which is configured to adjust the weight of different neurophysiological data according to the neurosymptom input by the user in advance.

[0013] In a second aspect, the application provides a multi-modal multi-acupoint collaborative stimulation method, applied to the multi-modal multi-acupoint nerve function regulation system as described above, comprising the following steps: Collecting the neurophysiological data of the central and peripheral nerves of the user, the neurophysiological data including electroencephalogram signals, electromyogram signals and / or multi-acupoint bioelectricity signals; Inputting the neurophysiological data into a pre-trained machine learning model to output corresponding neurophysiological state types; the machine learning model is configured as a SOM model; Obtaining multi-modal energy stimulation parameters of a plurality of preset acupoints from a predetermined mapping table according to the neurophysiological state types, and collaboratively stimulating the preset acupoints through the stimulation parameters; Real-time monitoring the neurophysiological data after the collaborative stimulation until the neurophysiological data reaches a predetermined neurophysiological target threshold.

[0014] The above technical solution has the following technical effects: By collecting the neurophysiological data of the central and peripheral nerves of the user, including electroencephalogram signals, electromyogram signals and / or multi-acupoint bioelectricity signals; inputting the neurophysiological data into a pre-trained machine learning model to output corresponding neurophysiological state types; obtaining multi-modal energy stimulation parameters of a plurality of preset acupoints from a predetermined mapping table according to the neurophysiological state types, and collaboratively stimulating the preset acupoints through the stimulation parameters; real-time monitoring the neurophysiological data after the collaborative stimulation until the neurophysiological data reaches a predetermined neurophysiological target threshold. The application solves the problem that the stimulation parameters in the prior art are mostly fixed and preset, and cannot be adjusted according to different neurophysiological states of the user. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 FIG. 1 is a structural schematic diagram of a multi-modal multi-acupoint nerve function regulation system according to an embodiment of the application.

[0016] Figure 2 FIG. 2 is a flowchart of a multi-modal multi-acupoint collaborative stimulation method according to an embodiment of the application. DETAILED DESCRIPTION

[0017] To further illustrate the embodiments, the application provides accompanying drawings. These drawings are part of the disclosure of the application, mainly used to illustrate the embodiments, and can be used to explain the operating principle of the embodiments in conjunction with the related description of the specification. Those of ordinary skill in the art should be able to understand other possible implementations and advantages of the application by referring to these contents. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0018] The application will be further described in conjunction with the accompanying drawings and specific embodiments.

[0019] Figure 1 This is a schematic diagram of the structure of a multimodal multi-acupoint nerve function regulation system according to an embodiment of the present invention, as shown below. Figure 1 As shown, the system in this embodiment includes a data acquisition module, a machine learning analysis module, a multimodal stimulation module, and a result feedback module, wherein the data acquisition module, machine learning analysis module, multimodal stimulation module, and result feedback module are communicatively connected, wherein: The data acquisition module is used to collect neurophysiological data of the user's central and peripheral nervous systems. In one specific implementation, the neurophysiological data includes electroencephalogram (EEG) signals, electromyogram (EMG) signals, and / or bioelectric signals from multiple acupoints. In this embodiment, the data acquisition module acquires neurophysiological data through multi-carrier integrated sensors, providing basic data support for the system: EEG signal acquisition relies on the head helmet carrier, with dry electrodes set on the inner side of the helmet corresponding to the scalp areas of the frontal and temporal lobes, and a sampling frequency of 250-500Hz; EMG signal acquisition is achieved through a flexible neck patch and a foot insole carrier, with the neck patch attached to the surface of the posterior neck muscles and the foot insole attached to the foot muscle area, both with built-in flexible electrodes and a sampling frequency of 100-200Hz; multi-acupoint bioelectrical signal acquisition is completed through dedicated acupoint patch carriers, such as the Baihui acupoint patch adapted to the reserved position on the head helmet, the Yongquan acupoint patch integrated into the foot insole, and the Huiyin acupoint using an independent flexible patch, each patch having built-in microelectrodes and a sampling frequency of 100-300Hz.

[0020] The module has a built-in filtering unit that removes environmental electromagnetic interference and motion artifacts through wavelet transform, ensuring that the signal-to-noise ratio of the original data is ≥30dB. It supports Bluetooth 5.0 or Wi-Fi 6 protocols to transmit the pre-processed data to the machine learning analysis module and the matching terminal in real time, which not only meets the needs of real-time determination of neurophysiological state, but also facilitates the viewing of data acquisition status.

[0021] The machine learning analysis module is used to input neurophysiological data into a pre-trained machine learning model and output the corresponding neurophysiological state type; in one specific implementation, the machine learning model is configured as a SOM model. In one specific implementation, the machine learning analysis module is also configured with a feature optimization unit, which is used to perform wavelet transform denoising on the collected neurophysiological data and extract the core features of the neurophysiological data, including the proportion of alpha waves and the power spectral density of the EEG signal, the standard deviation of the signal amplitude and the recruitment time of the electromyography signal, and the peak amplitude and pulse frequency of the bioelectrical signals of multiple acupoints.

[0022] In this embodiment, the characteristics of the machine learning analysis module SOM model (self-organizing mapping neural network) solve the problem that existing devices cannot dynamically adapt to individual neural differences, which is implemented as follows: The module first receives the electroencephalogram signals, electromyogram signals and multi-acupoint bioelectricity signals transmitted by the data acquisition module, and starts a three-stage preprocessing process: In the first stage, db4 wavelet basis is used for wavelet transform to filter out 50Hz power frequency interference such as environmental electromagnetic noise and motion artifacts such as user head rotation and foot displacement, accurately retain the effective signal frequency band, and the specific frequency band is 0.5-30Hz for electroencephalogram, 10-500Hz for electromyogram, and 0.1-100Hz for acupoint bioelectricity; In the second stage, the denoised data is quantized to extract 6 core features, including the proportion of alpha waves in electroencephalogram (8-13Hz frequency band power ratio, reflecting the central nervous inhibition degree), the power spectral density of theta waves (4-8Hz, related to sleep and emotional state), the amplitude standard deviation of electromyogram (reflecting the stability of peripheral nerve and muscle coordination), the recruitment time (the time from the start of the signal to the peak, reflecting the nerve conduction velocity), the peak amplitude of multi-acupoint bioelectricity (corresponding to acupoint nerve ending activity), and the pulse frequency (reflecting the nerve excitability at the acupoint); In the third stage, the 6 features are uniformly scaled to the [0, 1] interval by the Min-Max normalization algorithm to form a 1x6-dimensional standardized feature vector, eliminating the interference of different signal magnitudes such as electroencephalogram in μV and acupoint bioelectricity in mV on model determination.

[0023] The SOM model used is an unsupervised clustering neural network composed of an input layer (including 6 neurons corresponding to 6 types of standardized features) and a 4x4 structure competition layer (including 16 neurons, each representing a potential neurophysiological state). During the model training phase, 1000+ labeled sample data covering different populations such as insomnia, anxiety, chronic neuralgia patients and healthy people are input, and the training is completed through 5000+ iterations: In each iteration, after inputting the sample feature vector, calculate the Euclidean distance between it and all neurons in the competition layer, select the neuron with the smallest distance as the "winning neuron", and then adjust the connection weights of the winning neuron and its surrounding neurons through the Gaussian neighborhood function, so that the competition layer gradually forms a topological structure corresponding to the sample features. Finally, the neurophysiological state is clustered into 4 core types, including central excitation type (electroencephalogram beta wave proportion > 35%), central inhibition type (electroencephalogram alpha wave proportion < 15%), peripheral nerve sensitivity type (electromyogram amplitude standard deviation > 50μV), and peripheral nerve sluggishness type (electromyogram recruitment time > 100ms), and the feature threshold range of each type of state is stored. The model is deployed on the system embedded chip (STM32H7) and can complete the operation within ≤30s after inputting the feature vector.

[0024] The module fundamentally solves the core defects of existing devices through the unsupervised clustering characteristics of the SOM model: first, it avoids relying on fixed judgment standards and can automatically identify the differences in neural characteristics of different users, such as A user's alpha wave proportion of 18% being within the normal range, and B user's alpha wave proportion of 18% being central inhibition type; second, it supports real-time response to changes in neurophysiological data, and when a user's brain wave alpha wave proportion increases from 12% to 18% after receiving stimulation, the module can re-calculate and output a new state type to provide a basis for adjusting parameters for the multi-modal stimulation module; third, it continuously optimizes the judgment accuracy through model iteration to ensure that individual neural states can still be accurately matched in the long term, laying a foundation for subsequent dynamic parameter adjustment and breaking the limitations of existing devices' one-size-fits-all parameter settings.

[0025] In a specific implementation, the machine learning analysis module further includes a weight adjustment unit configured to adjust the weights of different neurophysiological data according to the user's pre-input neurological symptoms.

[0026] In this embodiment, the machine learning analysis module further includes a weight adjustment unit configured to differentially allocate feature weights of brain electrical signals, electromyographic signals, and multi-acupoint bioelectric signals according to the user's pre-input neurological symptoms such as insomnia, chronic neuralgia, and anxiety, to ensure that the state judgment is more in line with the core needs of the disease. Specifically, when the user inputs the insomnia symptom, the weight adjustment unit increases the weight of brain electrical signal features from the basic proportion of 30% to 45%, while reducing the weight of electromyographic signal features from 30% to 20% and the weight of multi-acupoint bioelectric signal features from 40% to 35%, to preferentially determine sleep-related neural states such as central inhibition type / excitement type through central nervous data; when the user inputs the chronic neuralgia symptom, the weight of electromyographic signal features is increased to 40%, the weight of brain electrical signal features is reduced to 25%, and the weight of multi-acupoint bioelectric signal features is reduced to 35%, to focus on capturing abnormal signals of peripheral nerves and muscles in coordination; when the user inputs the anxiety symptom, the feature weight of beta wave proportion in brain electrical signals is increased to 35%, and the weights of other features are reduced by 5%-10%, to focus on determining the central nervous excitement state. This design avoids the judgment deviation caused by general feature weights through disease-oriented weight allocation, further improves the accuracy of neurophysiological state judgment, provides a basis for the multi-modal stimulation module to match parameters more in line with the disease, and helps to solve the core problem of fixed parameters not being able to adapt to different individuals.

[0027] The multi-modal stimulation module is configured to obtain multi-modal energy stimulation parameters of a plurality of preset acupoints from a predetermined mapping table according to the neurophysiological state type, and to stimulate the preset acupoints through the stimulation parameters; In a specific implementation, the multi-modal stimulation module includes an energy sub-module group and an acupoint positioning unit. The energy sub-module group includes a low-frequency magnetic stimulation sub-module, a near-infrared laser sub-module, and / or an ultrasonic wave sub-module. The acupoint positioning unit is internally provided with a pressure sensor and an infrared locator, which are used to detect the fitting degree of a plurality of preset acupoints in real time. When the fitting degree is lower than a preset fitting threshold, an audible and light prompt is triggered. The plurality of preset acupoints include Baihui acupoint, Sishengcong acupoint, Fengchi acupoint, Fengfu acupoint, Yufeng acupoint, frontal lobe, Huiyin acupoint, Yungquan acupoint, and / or foot bottom brain reflex zone.

[0028] In the present embodiment, the multi-modal stimulation module realizes targeted stimulation of multiple acupoints and multiple energies through the process of "parameter matching-precise positioning-coordinated output". The specific implementation is as follows: the module first receives the neural physiological state type output by the machine learning analysis module, such as central excitation type + peripheral nerve sensitivity type, and retrieves the adaptive multi-modal energy stimulation parameters from the predetermined "state-parameter" mapping table. For example, for central excitation type, match Baihui acupoint low-frequency magnetic stimulation (frequency 0.8-1.5 Hz, intensity 200-280 mT), frontal lobe near-infrared laser (wavelength 808 nm, power 50-60 mW); for peripheral nerve sensitivity type, match Yungquan acupoint ultrasonic wave (frequency 1.5-2 MHz, acoustic intensity 0.2-0.3 W / cm 2 ), Huiyin acupoint low-frequency magnetic stimulation (frequency 2-3 Hz, intensity 150-200 mT).

[0029] In a specific implementation, the multi-modal energy stimulation of the multi-modal stimulation module covers multiple types of physical energy such as electromagnetic waves, light waves, ultrasonic waves, bioelectricity, hyperthermia, lasers, and magnetic stimulation. The characteristics of each type of energy, the adaptive acupoints, and the Chinese and Western medicine regulation logic are as follows, ensuring accurate matching of different neural physiological states and acupoint physiological characteristics: Electromagnetic wave stimulation: low-frequency pulse electromagnetic waves (frequency 0.5-5 Hz, electric field intensity 5-15 kV / m) are used, with a penetration depth of 3-5 cm, suitable for deep acupoints (such as Fengfu acupoint and Huiyin acupoint). In traditional Chinese medicine, it has the effect of "warming and unblocking meridians". In Western medicine, it can regulate the excitability of deep nerve plexus (such as pelvic autonomic nerve plexus). It is suitable for users with peripheral nerve sluggishness and can help improve the amplitude of electromyographic signals.

[0030] Light wave stimulation: visible light waveband (wavelength 620-660 nm) is selected, with a penetration depth of 0.5-1 cm, suitable for superficial acupoints (such as Yufeng acupoint and Neiguan acupoint). In traditional Chinese medicine, it belongs to the category of "dispersing wind and clearing heat". In Western medicine, it can improve the microcirculation of the acupoint area. It is used synchronously when collecting acupoint bioelectricity signals to improve the signal-to-noise ratio and avoid environmental interference.

[0031] Ultrasonic wave stimulation: focused ultrasonic waves (frequency 0.5-3 MHz, acoustic intensity 0.1-0.5 W / cm 2), mechanical vibration can activate nerve endings, adapt to foot acupoints (Yongquan acupoint, foot bottom brain reflex area), in line with the theory of "treating upper disease from lower" in traditional Chinese medicine, and western medicine can regulate the central nervous system through reflex, aiming at insomnia users, and auxiliary to improve the proportion of brain alpha wave.

[0032] Bioelectricity stimulation: output low-frequency electric pulse (frequency 20-100 Hz, intensity 0.1-1 mA), directly acting on nerve fibers, adapting to acupoints in limbs (such as Taixi acupoint, Neiguan acupoint), traditional Chinese medicine to achieve the effect of "unblocking meridians and activating collaterals", western medicine can improve the conduction velocity of peripheral nerves, aiming at chronic neuralgia users, and relieve pain signal transmission.

[0033] Thermal therapy stimulation: adopt far infrared heat therapy (temperature 38-42℃, power 10-30W), relax local muscles through thermal effect, adapt to acupoints behind the neck (Fengchi acupoint, Fengfu acupoint), traditional Chinese medicine corresponds to the effect of "dispelling cold and relieving pain", western medicine can relieve the compression of occipital nerve by muscle tension in the neck and occipital region, and auxiliary low-frequency magnetic stimulation to improve the regulation effect.

[0034] Laser stimulation: choose near-infrared laser (wavelength 808-980nm), strong directivity and energy concentration, penetration depth 2-5mm, adapt to acupoints in head (Baohui acupoint, Sishencong acupoint), traditional Chinese medicine belongs to "brain health and spirit calming" category, western medicine can improve the blood oxygen saturation of cerebral cortex, aiming at anxiety users, and regulate the overactive state of brain beta wave.

[0035] Magnetic stimulation: divided into low-frequency magnetic stimulation (frequency 0.5-5Hz, magnetic field intensity 100-500mT) and high-frequency magnetic stimulation (frequency 5-20Hz, magnetic field intensity 50-200mT), low frequency adapts to central related acupoints (Baohui acupoint, frontal lobe), regulates the excitability of cerebral cortex; high frequency adapts to peripheral nerve acupoints (Huiyin acupoint, Taixi acupoint), improves the sensitivity of peripheral nerves, and realizes "central-peripheral" coordinated regulation by combining traditional Chinese and western medicine.

[0036] These energy types can be flexibly combined according to the type of neurophysiological state, combined with acupoint positioning unit to ensure accurate delivery of energy, improve the pertinence and effectiveness of synergistic stimulation.

[0037] Preferably, the multi-modal stimulation module comprises an energy submodule group and an acupoint positioning unit: the energy submodule group integrates a low-frequency magnetic stimulation submodule (output frequency 0.5-5Hz, magnetic field intensity 100-500mT), a near-infrared laser submodule (wavelength 808-980nm, power 30-100mW), an ultrasonic wave submodule (frequency 0.5-3MHz, sound intensity 0.1-0.5W / cm 2), can be activated alone or in combination with the promoter module according to the parameter requirements, such as "low-frequency magnetic + near-infrared laser" dual-energy stimulation at Baihui acupoint, and ultrasonic stimulation alone at Yongquan acupoint; The acupoint positioning unit is located at the preset acupoint corresponding position of Baihui acupoint, Sishencong acupoint, Fengchi acupoint, Fengfu acupoint, Yifeng acupoint, frontal lobe, perineum acupoint, Yongquan acupoint, and foot bottom brain reflex area, etc. All are equipped with pressure sensors and infrared locators. The contact pressure (adhesion threshold is set to 90%) of the pressure sensor detection module and the acupoint skin is calibrated and positioned by the infrared locator through the skin temperature and the anatomical position of the acupoint. When the adhesion of any acupoint is less than 90%, the module will immediately trigger the audible and visual prompt (red light flashing + bee buzzing).

[0038] It should be noted that the preset acupoints covered by the multi-modal stimulation module not only match the overall regulation logic of nerve function of traditional Chinese medicine meridian theory, but also match the central-peripheral nerve target intervention requirements of Western medicine. The characteristics and function adaptation of traditional Chinese and Western medicine are as follows: Baihui acupoint: It is a main acupoint of the Governor Vessel in traditional Chinese medicine, located at the intersection of the midline of the top of the head and the line connecting the two ear tips, and is responsible for "spiritual tranquility, mental stability, and brain regulation". It corresponds to the central nervous system (cerebral cortex) in Western medicine. It is adapted to the synergistic effect of low-frequency magnetic stimulation and near-infrared laser. The former penetrates the skull to regulate the excitability of the cerebral cortex, and the latter improves local cerebral blood supply, especially for central excitatory users such as anxious users, to balance the proportion of brain electrical beta waves by stimulating this acupoint.

[0039] Sishencong acupoint: It is an extraordinary acupoint of meridian in traditional Chinese medicine, located 1 inch around Baihui acupoint, and belongs to "brain-related acupoint group". It is responsible for "brain health, intelligence, and mental tranquility". It corresponds to the parietal lobe and frontal lobe cortex area of Western medicine. It is adapted to near-infrared laser stimulation alone to enhance the central nervous regulation effect of Baihui acupoint and solve the problem of insufficient brain electrical alpha wave proportion of insomniac users.

[0040] Fengchi acupoint / Fengfu acupoint: Both belong to the intersection acupoints of the Governor Vessel and the Gallbladder Meridian of Foot-Shaoyang in traditional Chinese medicine. Fengchi acupoint is located in the concave part under the occipital bone at the back of the neck, and Fengfu acupoint is located 1 inch above the hairline on the median line at the back of the neck. Both are responsible for "dispersing wind, unblocking meridians, and opening orifices to awaken the spirit". They correspond to the vertebral artery, occipital nerve, and brainstem area in Western medicine. They are adapted to low-frequency magnetic stimulation, which can improve vertebral artery blood supply and relieve occipital nerve compression. It is suitable for chronic neuralgia users such as cervical headache to regulate peripheral nerve pain signal transmission by stimulation.

[0041] Yifeng acupoint: It belongs to the Sanjiaojing of Hand-Shaoyang in traditional Chinese medicine, located in the concave part between the mastoid process behind the ear and the mandibular angle, and is responsible for "ear opening, qi regulation, and pain relief". It corresponds to the facial nerve and vestibular nerve area in Western medicine. It is adapted to low-intensity near-infrared laser stimulation to assist in regulating facial nerve excitability, and at the same time, it collects bioelectric signals at this acupoint to reflect changes in peripheral nerve sensitivity.

[0042] The prefrontal cortex is a core target of the central nervous system in Western medicine (frontal cortex), mainly responsible for "emotional regulation and cognitive function". In traditional Chinese medicine, it is classified as "the area governed by the spirit of the brain". It is suitable for near-infrared laser stimulation, which can improve the blood oxygen saturation of the frontal cortex and regulate the overactive state of beta waves in the brain of anxious users, which is in line with the concept of "regulating the spirit and relieving depression" in traditional Chinese medicine.

[0043] Huiyin acupoint: In traditional Chinese medicine, it is an important acupoint on the Ren meridian, located between the anterior and posterior genitals. It is mainly used to "harmonize qi and blood, calm the mind and consolidate the body", corresponding to the pelvic autonomic nerve plexus in Western medicine. It is suitable for the synergistic use of low-frequency magnetic stimulation and bioelectric stimulation. The former regulates the function of the pelvic autonomic nerve, while the latter improves the conduction of peripheral nerves. For users with sluggish peripheral nerves, such as those with pelvic floor nerve dysfunction, it helps to increase the amplitude of electromyographic signals.

[0044] Yongquan acupoint / Brain reflex zone on the sole of the foot: Yongquan acupoint is the Jing-Well point of the Kidney Meridian of Foot-Shaoyin in Traditional Chinese Medicine. It is located in the depression of the anterior 1 / 3 of the sole of the foot and is mainly used to "nourish Yin and reduce fire, calm the mind and soothe the nerves". It corresponds to the nerve endings of the sole of the foot and the brain reflex zone in Western medicine. It is suitable for ultrasound stimulation, which activates the nerve endings of the sole of the foot through mechanical vibration, and at the same time reflexively regulates the state of the central nervous system. Combined with the Western medicine "foot reflex zone intervention" and the TCM theory of "treating the upper body by treating the lower body", it can improve the problem of insufficient central inhibition in insomnia users.

[0045] In one specific implementation, the predetermined mapping table is a dynamically updated mapping table. The machine learning analysis module is also configured with a data iteration unit. The data iteration unit stores the neurophysiological data, neurophysiological state type, stimulation parameters and regulation results in the database during each co-stimulation process. This data is used to retrain and optimize the SOM model and synchronously update the mapping relationship between the neurophysiological state type and the multimodal energy stimulation parameters in the dynamically updated mapping table.

[0046] In this embodiment, the data iteration unit first sets the data collection dimensions. After each co-stimulation process, it automatically captures key data from the entire chain, including the initial neurophysiological data before stimulation, the neurophysiological state type output by the machine learning analysis module, the specific parameters called by the multimodal stimulation module, and the adjustment results after stimulation (whether the real-time data reaches the target threshold, the time taken to reach the target, and the comfort score reported by the user). The data is stored in the local database according to the rule that one stimulation = one complete data record. At the same time, data validity judgment criteria are set, such as missing data items ≤10% and stimulation duration ≥15min, to filter invalid records to ensure sample quality.

[0047] When the database accumulates 100 sets of valid data, the data iteration unit automatically triggers the SOM model to retrain: First, the newly accumulated valid data is divided into a training set and a validation set in a 7:3 ratio. The training set is used to update the connection weights of neurons in the competitive layer of the SOM model—by recalculating the Euclidean distance between the feature vector and the neuron, the neighborhood range and weight update step size of the winning neuron are adjusted, and the clustering boundary of the neurophysiological state type is optimized. For example, the original threshold for the proportion of β waves in the central excitation type was >35%, which was corrected to >32% after iteration, making it more consistent with the actual distribution of user data. The validation set is used to test the judgment accuracy of the optimized model. If the accuracy is improved by ≥5% or stabilized at ≥92% compared with the previous model, the model update is confirmed to be effective; otherwise, the data samples are backtracked and retrained.

[0048] After model optimization, the data iteration unit synchronously updates the dynamic update mapping table: based on the neurophysiological state type output by the new model, the corresponding multimodal energy stimulation parameters are rematched. For example, the original central inhibition type was matched with a low-frequency magnetic intensity of 280-360mT at Baihui acupoint. After iteration, it was found that the average time to reach the target under this parameter was 22 minutes. After optimization, it was adjusted to 300-380mT, and the time to reach the target was shortened to 18 minutes. The data iteration unit then updates the correspondence between central inhibition type and low-frequency magnetic intensity of Baihui acupoint in the mapping table to 300-380mT. At the same time, the effect difference before and after parameter adjustment is recorded to form an association record of state type - parameter before optimization - parameter after optimization - effect comparison, which is convenient for subsequent tracking of the iteration logic.

[0049] Furthermore, the data iteration unit supports two update modes: automatic iteration, triggered by a fixed data threshold, ensuring the system continuously adapts to changes in the user's neural state over long-term use; and manual iteration, triggered by the user in the background, which allows for forced updates after supplementing labeled data for specific cases, improving the model and mapping table's adaptability to complex symptoms. This design achieves dynamic optimization of the SOM model and mapping table through data iteration, preventing the initially set model and parameters from becoming ineffective due to long-term changes in the user's neural state, such as increased neural sensitivity after prolonged use, and ensuring that the system's adjustment accuracy continuously improves with usage time.

[0050] In one specific implementation, the multimodal stimulation module is also configured with a timing control unit, which is used to set the output time sequence of multimodal energy according to the physiological correlation logic of preset acupoints.

[0051] In this embodiment, the timing control unit is based on the dual physiological correlation logic of central-peripheral nerve linkage and the flow of Qi and blood in traditional Chinese medicine meridians. It sets the precise output time sequence of multimodal energy to ensure the synergistic superposition of stimulation effects. The specific implementation is as follows: The timing control unit first pre-stores basic timing templates based on physiological correlations, and calls the corresponding templates for different neurophysiological states: for users with central excitation and normal peripheral nerve states, such as those with anxiety, the timing logic of "central priority regulation → peripheral auxiliary consolidation" is adopted. Specifically, in the 1st second, the prefrontal cortex near-infrared laser stimulation is activated (acting on the central nervous system first to quickly inhibit excessive β-wave activity), after a 2-second delay, the Baihui acupoint low-frequency magnetic stimulation is activated (to enhance the central inhibitory effect), after a 3-second delay, the Fengchi acupoint low-frequency magnetic stimulation is activated (to improve cerebral blood supply and assist in central regulation), and finally, after a 5-second delay, the Yongquan acupoint ultrasonic stimulation is activated (to indirectly stabilize the central state through plantar nerve reflexes). The energy activation interval of each acupoint is strictly controlled within 2-5 seconds to avoid local discomfort caused by energy superposition.

[0052] For users with chronic insomnia who exhibit "central inhibition + peripheral sluggishness," a temporal logic of "central-peripheral synchronous activation → core acupoint enhancement" is adopted. Specifically, in the first second, Baihui acupoint (low-frequency magnetic + near-infrared laser dual energy) and Yongquan acupoint ultrasonic stimulation are activated simultaneously (synchronous awakening of the central and peripheral nerves). After a 4-second delay, low-frequency magnetic stimulation of Huiyin acupoint is activated (through the Ren meridian connection, assisting the central nervous system in increasing the proportion of alpha waves). After another 3-second delay, the intensity of low-frequency magnetic stimulation of Baihui acupoint is increased by 10%-15% (core acupoint enhancement and regulation), forming a temporal rhythm of "synchronous activation - step-by-step enhancement," which is adapted to the need for enhanced activity of both the central and peripheral nerves.

[0053] For users with "peripheral nerve sensitivity + central nervous system normality," such as those with chronic neuralgia, a timing logic of "peripheral buffering start → central nervous system stabilization escort" is adopted. Specifically, in the first second, near-infrared laser stimulation of Yifeng and Taixi acupoints is initiated at 50% of the base intensity (slowly activating sensitive peripheral nerves and reducing tingling sensation). After a 3-second delay, the stimulation intensity of the surrounding acupoints is restored to the normal value, while low-frequency magnetic stimulation of Fengfu acupoint is initiated (to stabilize the central nervous system and avoid compensatory excitation of the central nervous system caused by peripheral nerve sensitivity). Finally, after a 2-second delay, bioelectric stimulation of Neiguan acupoint is initiated (to regulate the autonomic nervous system and assist in relieving pain signal transmission). Through the timing logic of "low intensity start - gradual increase in effectiveness," the stimulation effect and user tolerance are balanced.

[0054] Furthermore, the timing control unit supports customized adjustment of timing parameters. Medical staff can modify the start-up delay time and intensity enhancement nodes of energy for each acupoint through the system backend, such as delaying by 5s or 8s to strengthen core acupoints, to adapt to the physiological needs of special cases. It also has a built-in timing anomaly monitoring function. If the start-up delay of energy for a certain acupoint exceeds the set error (e.g., ±0.5s), the module immediately pauses output and displays a fault message, ensuring precise execution of the timing logic. This design, through timing control that aligns with physiological relationships, avoids the cancellation of effects or discomfort caused by disordered output of multiple energies, improving the accuracy and effectiveness of multimodal synergistic stimulation.

[0055] The results feedback module is used to monitor the neurophysiological data after synergistic stimulation in real time until the neurophysiological data reaches the predetermined neurophysiological target threshold.

[0056] In this embodiment, the result feedback module is linked with the data acquisition module. After the multimodal stimulation is initiated, the EEG signal, EMG signal and bioelectric signal of multiple acupoints are collected synchronously. After the collected data is denoised by wavelet transform, it is transmitted to the built-in threshold comparison unit.

[0057] The threshold comparison unit pre-stores neurophysiological target thresholds categorized by disease and population: For insomnia users, the central target threshold is set at ≥22% (30-50 years old) or ≥18% (60 years and older) of EEG alpha wave proportion and ≤30% of beta wave proportion; for users with chronic neuralgia, the peripheral nerve target threshold is set at a stable standard deviation of EMG signal amplitude of 25-45μV and a bioelectric pulse frequency of 8-12Hz at the Yongquan acupoint; for anxiety users, the central target threshold is set at ≤28% of EEG beta wave proportion and ≥85% stability of prefrontal cortex bioelectric signal (signal fluctuation amplitude <10%). After each data collection, the unit automatically compares the real-time data with the corresponding target threshold, generating three judgment results: qualified, unqualified, and abnormal.

[0058] If the target is met, the module continues to monitor for two more cycles. If the data remains stable within the threshold range, the stimulus effect is deemed satisfactory, triggering the multimodal stimulus module to stop outputting and generating a stimulus report. If the target is not met, the deviation data is immediately sent back to the machine learning analysis module, triggering a parameter rematching process to generate new, suitable stimulus parameters. If an anomaly is detected, the module immediately activates the anomaly response mechanism, suspends multimodal stimulus output, sends a warning message to the associated terminal, and retains the original data for 5 seconds before and after the anomaly for subsequent analysis of the cause of the fault.

[0059] Secondly, this application provides a multimodal, multi-acupoint synergistic stimulation method, applied to a multimodal, multi-acupoint neural function regulation system as described above, such as... Figure 2 As shown, the method includes the following steps: Collect neurophysiological data of the user's central and peripheral nervous systems, including electroencephalogram (EEG) signals, electromyogram (EMG) signals, and / or bioelectric signals from multiple acupoints; Neurophysiological data is input into a pre-trained machine learning model, which outputs the corresponding neurophysiological state type; the machine learning model is configured as a SOM model. Based on the neurophysiological state type, multimodal energy stimulation parameters of multiple preset acupoints are obtained from a predetermined mapping table, and the preset acupoints are synergistically stimulated using the stimulation parameters. The neurophysiological data after synergistic stimulation are monitored in real time until the neurophysiological data reaches the predetermined neurophysiological target threshold.

[0060] Although the invention has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to the invention without departing from the spirit and scope of the invention as defined in the appended claims, all of which shall be within the scope of protection of the invention.

Claims

1. A multimodal, multi-acupoint neural function regulation system, characterized in that, It includes a data acquisition module, a machine learning analysis module, a multimodal stimulation module, and a result feedback module, which are communicatively connected. The data acquisition module is used to collect neurophysiological data of the user's central and peripheral nervous systems, including electroencephalogram (EEG) signals, electromyogram (EMG) signals, and / or bioelectric signals from multiple acupoints. The machine learning analysis module is used to input the neurophysiological data into a pre-trained machine learning model and output the corresponding neurophysiological state type; the machine learning model is configured as a SOM model. The multimodal stimulation module is used to obtain multimodal energy stimulation parameters of multiple preset acupoints from a predetermined mapping table according to the neurophysiological state type, and to perform synergistic stimulation on the preset acupoints through the stimulation parameters. The result feedback module is used to monitor the neurophysiological data after the synergistic stimulation in real time until the neurophysiological data reaches the predetermined neurophysiological target threshold.

2. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The preset acupoints include Baihui (GV20), Sishencong (EX-HN1), Fengchi (GB20), Fengfu (GV16), Yifeng (TE17), frontal lobe, Huiyin (CV1), Yongquan (KI1), and / or the foot reflex zone.

3. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The machine learning analysis module is also equipped with a feature optimization unit, which is used to perform wavelet transform denoising on the collected neurophysiological data and extract the core features of the neurophysiological data. The core features include the alpha wave ratio and the theta wave power spectral density of the EEG signal, the standard deviation of the signal amplitude and recruitment time of the electromyography signal, and the peak amplitude and pulse frequency of the bioelectrical signals of multiple acupoints.

4. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The multimodal stimulation module includes an energy submodule group and an acupoint positioning unit. The energy submodule group includes a low-frequency magnetic stimulation submodule, a near-infrared laser submodule, and / or an ultrasound submodule. The acupoint positioning unit has a built-in pressure sensor and an infrared locator for real-time detection of the fit of multiple preset acupoints. When the fit is lower than a preset fit threshold, an audio-visual prompt is triggered.

5. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The result feedback module is also configured with a dynamic threshold adjustment unit and an anomaly warning unit; the dynamic threshold adjustment unit is used to adjust the predetermined neurophysiological target threshold according to the user's age and the type of underlying disease; the anomaly warning unit is used to trigger the multimodal stimulation module to pause output and send warning information to the system when abnormal neurophysiological data is detected after co-stimulation.

6. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The predetermined mapping table is a dynamically updated mapping table. The machine learning analysis module is also configured with a data iteration unit. The data iteration unit stores the neurophysiological data, neurophysiological state type, stimulation parameters and regulation results of each co-stimulation process into the database for retraining and optimizing the SOM model, and synchronously updates the mapping relationship between the neurophysiological state type and the multimodal energy stimulation parameters in the dynamically updated mapping table.

7. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The multimodal stimulation module is also equipped with a timing control unit, which is used to set the output time sequence of multimodal energy according to the physiological correlation logic of preset acupoints.

8. The multimodal multi-acupoint nerve function regulation system according to claim 1, characterized in that, The machine learning analysis module is also equipped with a weight adjustment unit, which is used to adjust the weights of different neurophysiological data according to the neurological symptoms pre-input by the user.

9. A multimodal, multi-acupoint synergistic stimulation method, characterized in that, The multimodal, multi-acupoint neural function modulation system described in any one of claims 1-8 comprises the following steps: Collect neurophysiological data of the user's central and peripheral nervous systems, including electroencephalogram (EEG) signals, electromyogram (EMG) signals, and / or bioelectric signals from multiple acupoints; The neurophysiological data is input into a pre-trained machine learning model, which outputs the corresponding neurophysiological state type; the machine learning model is configured as a SOM model. Based on the neurophysiological state type, multimodal energy stimulation parameters of multiple preset acupoints are obtained from a predetermined mapping table, and the preset acupoints are synergistically stimulated using the stimulation parameters. The neurophysiological data following the synergistic stimulation are monitored in real time until the neurophysiological data reaches a predetermined neurophysiological target threshold.

Images