Brain-controlled closed-loop rehabilitation training system for swallowing function training

Abstract

This disclosure provides a brain-controlled closed-loop rehabilitation training system for swallowing function training, which can be applied to the field of intelligent medical device technology. The brain-controlled closed-loop rehabilitation training system may include: a multimodal signal acquisition module; a control module configured to: determine the i-th contraction degree index of peripheral neuromuscular muscles based on the i-th electromyographic signal at time i, and generate a first control signal and a second control signal based on the target EEG signal, the i-th contraction degree index, the pre-stored EEG signal of the target subject under full swallowing, and a benchmark contraction index corresponding to the contraction degree of peripheral neuromuscular muscles under full swallowing, where i is an integer greater than 1; a first stimulation module configured to electrically stimulate the cerebral cortex region at time i+1 under the control of the first control signal, and a second stimulation module configured to electrically stimulate the peripheral neuromuscular muscles at time i+1 under the control of the second control signal, to assist the target subject in swallowing at time i+1.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent medical device technology, and more specifically, to a brain-controlled closed-loop rehabilitation training system for swallowing function training. Background Technology

[0002] Post-stroke dysphagia is one of the most common and serious functional impairments in stroke patients. Swallowing rehabilitation methods include behavioral training, compensatory strategies, and peripheral neuromuscular electrical stimulation (NMES). However, these methods often rely on the therapist's experience or pre-set stimulation protocols, resulting in a lack of passive training and limited efficiency in neural remodeling. Summary of the Invention

[0003] In view of this, the present disclosure provides a brain-controlled closed-loop rehabilitation training system for swallowing function training.

[0004] One aspect of this disclosure provides a brain-controlled closed-loop rehabilitation training system for swallowing function training, comprising: a multimodal signal acquisition module configured to acquire electromyographic signals of peripheral neuromuscular tissues used for swallowing by a target subject, and electroencephalographic signals of the cerebral cortex region corresponding to the peripheral neuromuscular tissues; and a control module connected to the multimodal signal acquisition module and configured to: when a swallowing intention is detected in the target subject, determine an i-th contraction degree index of the peripheral neuromuscular tissues based on the i-th electromyographic signal acquired by the multimodal signal acquisition module in the i-th time period, and determine the target electroencephalographic signal, the i-th contraction degree index, and the pre-stored electroencephalographic signal of the target subject under full swallowing based on the target electroencephalographic signal from the multimodal signal acquisition module and the target contraction degree index; and determine the i-th contraction degree index based on the target electroencephalographic signal from the multimodal signal acquisition module, the i-th contraction degree index, and the target subject's pre-stored electroencephalographic signal under full swallowing. A baseline contraction index corresponding to the degree of contraction of peripheral neuromuscular contraction during full-force swallowing generates a first control signal and a second control signal. The i-th time period is located after the intention generation period of swallowing intention. The target EEG signal is the EEG signal collected by the multimodal signal acquisition module during the intention generation period, where i is an integer greater than 1. The first stimulation module and the second stimulation module are each connected to the control module. The first stimulation module is configured to electrically stimulate the cerebral cortex region in the i+1 time period under the control of the first control signal. The second stimulation module is configured to electrically stimulate the peripheral neuromuscular contraction in the i+1 time period under the control of the second control signal to assist the target subject in swallowing in the i+1 time period.

[0005] According to embodiments of this disclosure, a brain-controlled closed-loop rehabilitation training system for swallowing function training is provided. In this system, a multimodal signal acquisition module can acquire electromyographic (EMG) signals of the peripheral neuromuscular region of the target subject, which is involved in swallowing, and EEG signals of the corresponding cerebral cortex region. Upon detecting a swallowing intention in the target subject, the control module determines the i-th contraction degree index of the peripheral neuromuscular region based on the i-th EMG signal at the i-th time period. It then generates a first control signal and a second control signal based on the target EEG signal during the time period in which the swallowing intention is generated, the pre-stored EEG signal of the target subject under full swallowing, the i-th contraction degree index, and a benchmark contraction index corresponding to the contraction degree of the target subject's peripheral neuromuscular region under full swallowing. Based on this, the control module can use the first and second control signals to drive the first stimulation module and the second stimulation module respectively, to synchronously electrically stimulate the peripheral neuromuscular region and the central nervous system.

[0006] Since the first and second control signals are generated by the control module based on the target EEG signal when the target object has a swallowing intention, the pre-stored EEG signal of the target object under full swallowing, the i-th contraction degree index, and the benchmark contraction index corresponding to the contraction degree of the target object's peripheral neuromuscular system under full swallowing, the first and second stimulation modules, under the control of the first and second control signals, can synchronously electrically stimulate the peripheral neuromuscular system and central nervous system of the target object for swallowing, with the swallowing intention corresponding to the pre-stored EEG signal of the target object under full swallowing as the intention benchmark and the muscle contraction degree corresponding to the electromyographic signal of the target object under full swallowing as the muscle contraction benchmark. This allows the contraction degree of the target object's peripheral neuromuscular system to be as close as possible to or higher than the contraction degree of the target object under full swallowing, thereby achieving precise and effective assistance to the target object's swallowing function.

[0007] Based on this, the brain-controlled closed-loop rehabilitation training system of this disclosure uses swallowing-related multimodal physiological signals (such as the aforementioned electroencephalogram (EEG) and electromyogram (EMG) signals as trigger sources. It drives the first and second stimulation modules based on the swallowing intention represented by the EEG signals and the muscle contraction status represented by the EMG signals. This enables synergistic effects between central nervous system regulation and peripheral neuromuscular stimulation. Furthermore, it can perform the next round of electrical stimulation based on the target subject's multimodal physiological signals after the previous round of electrical stimulation, achieving closed-loop feedback in the training process. This improves the initiative of the training process and enhances the coordination between stimulation and movement, making it suitable for efficient swallowing function training. Thus, it achieves precise rehabilitation intervention for the target subject's peripheral neuromuscular system used for swallowing when the swallowing intention and muscle drive are matched in time, constructing a central-peripheral-central closed-loop rehabilitation mechanism that meets the needs of swallowing nerve function remodeling. Attached Figure Description

[0008] The above and other objects, features and advantages of this disclosure will become clearer from the following description of embodiments of this disclosure with reference to the accompanying drawings, which will be explained in conjunction with the drawings.

[0009] Figure 1 A schematic diagram of a brain-controlled closed-loop rehabilitation training system according to an embodiment of the present disclosure is shown.

[0010] Figure 2 A schematic diagram of a brain-controlled closed-loop rehabilitation training system according to another embodiment of the present disclosure is shown. Detailed Implementation

[0011] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0012] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0013] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0014] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0015] In the embodiments disclosed herein, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to safeguard user personal information security, network security, and national security.

[0016] In the embodiments of this disclosure, user authorization or consent is obtained before acquiring or collecting user personal information.

[0017] Non-invasive central nervous system modulation techniques such as transcranial direct current stimulation can be used to promote the recovery of neurological function after stroke. However, central stimulation is usually implemented independently of the specific swallowing rhythm and is difficult to effectively coordinate with the patient's actual swallowing behavior.

[0018] In some approaches, brain-computer interface technology can be applied to limb motor rehabilitation. However, in swallowing rehabilitation, there is still a lack of systematic solutions that can use the swallowing intention recognition results to synchronously drive central and peripheral stimulation and form a closed-loop control mechanism of "central-peripheral-central".

[0019] Therefore, how to achieve the core driving signal of the patient's swallowing intention, organically coordinate the central nervous system regulation and peripheral muscle stimulation, and dynamically adjust the stimulation parameters through physiological feedback to construct a complete closed-loop rehabilitation pathway during the swallowing rehabilitation process has become an urgent technical problem to be solved.

[0020] Based on this, the swallowing rehabilitation training system in related technologies has problems such as strong passivity in the training process, disconnect between stimulation and movement, and lack of closed-loop regulation.

[0021] In view of this, the present disclosure provides a brain-controlled closed-loop rehabilitation training system based on brain-computer interface technology. This system is driven by the swallowing intention represented by electroencephalogram (EEG) signals and the actual contraction status represented by the peripheral neuromuscular signals used for swallowing. It achieves synergistic effects between central nervous system regulation and peripheral neuromuscular stimulation. Furthermore, it can perform the next round of electrical stimulation based on the target subject's EEG and EMG signals after the previous round, realizing closed-loop feedback in the training process. This improves the initiative of the training process and enhances the coordination between stimulation and movement, enabling efficient swallowing function training.

[0022] Figure 1 A schematic diagram of a brain-controlled closed-loop rehabilitation training system according to an embodiment of the present disclosure is shown.

[0023] like Figure 1As shown, the brain-controlled closed-loop rehabilitation training system of this embodiment may include a multimodal signal acquisition module, a control module, a first stimulation module, and a second stimulation module.

[0024] The multimodal signal acquisition module can acquire electromyographic signals of the peripheral neuromuscular region of the target object used for swallowing, as well as electroencephalographic signals of the corresponding cerebral cortex region.

[0025] In this embodiment, the multimodal signal acquisition module may include one or more EEG electrodes, one or more EMG electrodes, and a signal acquisition circuit electrically connected to the EEG electrodes and EMG electrodes. The EEG electrodes may be, for example, wireless EEG acquisition electrode caps. The EMG electrodes may be, for example, flexible surface EMG electrodes.

[0026] Electroencephalogram (EEG) electrodes can be attached to the scalp of the target subject. Specifically, EEG electrodes can be attached to areas of the cerebral cortex used to control swallowing movements, such as the central-lateral sensorimotor area. The EEG electrodes can make contact with the target subject's scalp through the dry electrode structure of the EEG electrodes or through conductive gel, with the contact impedance controlled to be below 20 kΩ.

[0027] Electromyographic electrodes can be placed on the peripheral neuromuscular tissues of the target subject's pharynx used for swallowing. For example, these peripheral neuromuscular tissues may include at least one of the suprahyoid muscles or pharyngeal constrictors. The suprahyoid muscles may include the mylohyoid muscle. In some embodiments, multiple electromyographic electrodes can be attached to the surface of the mylohyoid muscle of the target subject, with a center-to-center spacing of 2 cm between the electrodes. The attachment positions of the electrodes should avoid areas of skin damage and hair. The arrangement of the electrodes should align with the direction of the muscle fibers to improve the sensitivity and stability of electromyographic signal acquisition.

[0028] The signal acquisition circuit can acquire electromyographic (EMG) signals from the peripheral neuromuscular system of the target object used for swallowing via EMG electrodes, and electroencephalogram (EEG) signals from the corresponding cerebral cortex region via EEG electrodes. It is important to note that the acquisition frequencies of the EEG and EMG signals need to be matched to facilitate subsequent fusion analysis of the EEG and EMG signals, which can improve the accuracy of swallowing intention recognition. For example, the sampling frequency can be set to 1000Hz. However, it should be understood that the signals acquired in this embodiment are not limited to this. In other embodiments, pressure sensors can also be used to acquire pressure signals from swallowing-related muscle groups during swallowing, so as to fuse and analyze the pressure signals, EEG signals, and EMG signals; these will not be elaborated upon here.

[0029] For example, the signal acquisition circuit can acquire the i-th electromyographic signal in the i-th time period, where i is an integer greater than 1. This i-th electromyographic signal can characterize the potential of the peripheral neuromuscular system of the target object in the i-th time period. In some embodiments, the i-th electroencephalogram (EEG) signal can also be acquired in the i-th time period, which will not be elaborated here.

[0030] Furthermore, this disclosure is not limited to this. Prior to this (e.g., before training using the brain-controlled closed-loop rehabilitation training system of this disclosure), the target subject's electroencephalogram (EEG) and electromyographic (EMG) signals during forceful swallowing can be collected. The EEG signal can characterize the reference potential of the aforementioned cerebral cortex region during forceful swallowing. The EMG signal can characterize the reference potential of the aforementioned peripheral neuromuscular region during forceful swallowing. Furthermore, the EEG signal can be pre-stored before training to obtain a pre-stored EEG signal. Similarly, the EMG signal can be pre-stored before training to obtain a pre-stored EMG signal. It should be noted that the target subject's "forceful swallowing" does not necessarily mean completing a full swallowing action. In some embodiments, when the target subject's swallowing-related muscle groups are detected to contract due to the target subject's exertion, and the target subject indicates that they have exerted their full effort (e.g., through a pre-agreed gesture), even if the swallowing-related muscle groups have not contracted to the target state that satisfies the target subject's actual swallowing needs, the target subject's EEG signal at this time can still be identified as the target subject's EEG signal during full-force swallowing, and the target subject's EMG signal at this time can also be identified as the target subject's EMG signal during full-force swallowing. Therefore, the EEG signal and the EMG signal can be used as signal references for the training process. Furthermore, to ensure the target subject's safety, the swallowing process can be conducted under the supervision of medical personnel, which will not be elaborated upon here.

[0031] The control module may include, for example, a terminal device. The control module can be electrically connected to the multimodal signal acquisition module. Thus, the control module can receive pre-stored electroencephalogram (EEG) signals and pre-stored electromyographic (EMG) signals of the target subject during forceful swallowing from the multimodal signal acquisition module. Furthermore, the control module can determine a baseline index of peripheral neuromuscular contraction based on the pre-stored EMG signals. This baseline contraction index characterizes the degree of muscle contraction during forceful swallowing.

[0032] During the training process of the target subject using the brain-controlled closed-loop rehabilitation training system of this embodiment, the control module can also detect the swallowing intention of the target subject based on the electroencephalogram (EEG) and electromyogram (EMG) signals received from the multimodal signal acquisition module. If the swallowing intention detection result indicates that the target subject has a swallowing intention, the EEG signal acquired at this time can be identified as the target EEG signal. This target EEG signal can characterize the potential of the aforementioned cerebral cortex regions during the period when the target subject has a swallowing intention.

[0033] After determining that the target subject has the intention to swallow (e.g., time period i), the control module can also receive the target's electroencephalogram (EEG) signal and the i-th electromyographic (EMG) signal from the multimodal signal acquisition module during time period i, and determine the i-th contraction degree index of the peripheral neuromuscular system based on the i-th EMG signal. The i-th contraction degree index can characterize the degree of muscle contraction of the target subject during time period i.

[0034] The control module can generate a first control signal and a second control signal based on the target EEG signal, the pre-stored EEG signal of the target object under full swallowing, the i-th contraction degree index, and the benchmark contraction index corresponding to the contraction degree of peripheral neuromuscular contraction under full swallowing.

[0035] For example, the control module can determine the brainwave activation characteristics of the target object when it intends to swallow, based on the target's EEG signal and the pre-stored EEG signal of the target object during full swallowing. These brainwave activation characteristics can characterize the effective activation level of the aforementioned cerebral cortex regions during the intention to swallow, using the target object's EEG signal during full swallowing as a benchmark.

[0036] For example, the control module can determine the activation level value of the i-th muscle of the peripheral neuromuscular system based on the i-th contraction level index and the benchmark contraction index corresponding to the contraction level of the peripheral neuromuscular system under full-force swallowing. This activation level value of the i-th muscle can characterize the activation level of the peripheral neuromuscular system used for swallowing in the target subject under the aforementioned target EEG signal. Based on this, the control module can determine a first stimulation parameter and a second stimulation parameter based on the EEG activation characteristics and the activation level value of the i-th muscle. The control module can generate a first control signal based on the first stimulation parameter and a second control signal based on the second stimulation parameter.

[0037] The control module can be electrically connected to the first stimulation module and the second stimulation module. For example, the first stimulation module can be used to implement transcranial direct current stimulation. The first stimulation module may include a constant current source, etc. The second stimulation module can be used to implement peripheral neuromuscular electrical stimulation. The second stimulation module may include a bidirectional square wave stimulation circuit, etc. The control module can send a first control signal to the first stimulation module and simultaneously send a second control signal to the second stimulation module.

[0038] For example, under the control of a first control signal, the first stimulation module can electrically stimulate the cerebral cortex region of the target object (such as the central-lateral sensorimotor area mentioned above, which corresponds to the central nervous system that drives peripheral nerves and muscles) during the i+1 time period. Specifically, the first stimulation module may have a first output electrode, which may be attached to the cerebral cortex region of the target object. The first stimulation module can output a first stimulation signal through the first output electrode to electrically stimulate the cerebral cortex region.

[0039] For example, under the control of a second control signal, the second stimulation module can provide electrical stimulation to the peripheral nerve and muscle in time period i+1, synchronously with the first stimulation module. Specifically, the second stimulation module may have a second output electrode, which can be attached to the peripheral nerve and muscle of the target object. The second stimulation module can output a second stimulation signal through the second output electrode to provide electrical stimulation to the peripheral nerve and muscle. In this way, the first stimulation module and the second stimulation module can assist the target object in swallowing during time period i+1.

[0040] Based on this, in this embodiment, the multimodal signal acquisition module can acquire the electromyographic (EMG) signals of the peripheral neuromuscular region of the target object used for swallowing and the electroencephalogram (EEG) signals of the corresponding cerebral cortex region. When the control module detects that the target object has a swallowing intention, it determines the i-th contraction degree index of the peripheral neuromuscular region based on the i-th EMG signal at the i-th time period. Based on the target EEG signal during the time period in which the target object generates the swallowing intention, the pre-stored EEG signal of the target object under full swallowing, the i-th contraction degree index, and the benchmark contraction index corresponding to the contraction degree of the target object's peripheral neuromuscular region under full swallowing, the control module generates a first control signal and a second control signal. On this basis, the control module can use the first and second control signals to drive the first stimulation module and the second stimulation module respectively, to synchronously electrically stimulate the peripheral neuromuscular region and the central nervous system.

[0041] Since the first and second control signals are generated by the control module based on the target EEG signal, the target object's pre-stored EEG signal under full swallowing, the i-th contraction degree index, and the benchmark contraction index corresponding to the contraction degree of the target object's peripheral neuromuscular system under full swallowing, the first and second stimulation modules, under the control of the aforementioned first and second control signals, can synchronously electrically stimulate the target object's peripheral neuromuscular system and central nervous system used for swallowing, using the swallowing intention corresponding to the pre-stored EEG signal under full swallowing as the intention benchmark and the muscle contraction degree corresponding to the target object's electromyographic signal under full swallowing as the muscle contraction benchmark. This ensures that the contraction degree of the target object's peripheral neuromuscular system is as close as possible to or higher than the contraction degree of the target object under full swallowing, thereby achieving precise and effective assistance to the target object's swallowing function.

[0042] Based on this, the brain-controlled closed-loop rehabilitation training system of this disclosure uses swallowing-related multimodal physiological signals (such as the aforementioned electroencephalogram (EEG) and electromyogram (EMG) signals as trigger sources. It drives the first and second stimulation modules based on the swallowing intention represented by the EEG signals and the muscle contraction status represented by the EMG signals. This enables synergistic effects between central nervous system regulation and peripheral neuromuscular stimulation. Furthermore, it can perform the next round of electrical stimulation based on the target subject's multimodal physiological signals after the previous round of electrical stimulation, achieving closed-loop feedback in the training process. This improves the initiative of the training process and enhances the coordination between stimulation and movement, making it suitable for efficient swallowing function training. Thus, it achieves precise rehabilitation intervention for the target subject's peripheral neuromuscular system used for swallowing when the swallowing intention and muscle drive are matched in time, constructing a central-peripheral-central closed-loop rehabilitation mechanism that meets the needs of swallowing nerve function remodeling.

[0043] It should be understood that the brain-controlled closed-loop rehabilitation training system of this disclosure can perform the above operations repeatedly.

[0044] Furthermore, in some embodiments, during the period prior to the i-th time period, such as the (i-1)-th time period within the same training phase, the control module may first detect the swallowing intention of the target object. If the target object is detected to have a swallowing intention during the (i-1)-th time period, the control module may first generate a first preliminary control signal and a second preliminary control signal, respectively controlling the first stimulation module and the second stimulation module to perform preliminary electrical stimulation during the i-th time period. It should be further noted that the i-th time period can be divided into a first sub-time period and a second sub-time period following the first sub-time period. During the first sub-time period, the first stimulation module and the second stimulation module may respectively perform electrical stimulation on the cerebral cortex region and the peripheral neuromuscular region. During the second sub-time period, the control module may use the electromyographic signal received from the multimodal signal acquisition module as the i-th electromyographic signal.

[0045] Based on this, in this embodiment, since the electromyographic electrodes need to be in contact with the peripheral neuromuscular tissues of the target object used for swallowing, and the second output electrode of the second stimulation module also needs to be in contact with these peripheral neuromuscular tissues, when the second stimulation module outputs a second stimulation signal, if the multimodal signal acquisition module simultaneously acquires electromyographic signals via the electromyographic electrodes, the accuracy of the electromyographic signal will be affected. Therefore, in this embodiment, electrical stimulation of the peripheral neuromuscular tissues is performed in the first sub-period of the i-th time period, and the electromyographic signal of the second sub-period of the i-th time period is used as the i-th electromyographic signal to drive the second stimulation module. This allows for accurate acquisition of the electromyographic signal, thereby enabling accurate synchronous electrical stimulation of the central nervous system and peripheral neuromuscular tissues, achieving precise and effective assistance to the target object's swallowing function.

[0046] According to embodiments of this disclosure, the multimodal signal acquisition module can acquire electroencephalogram (EEG) signals and electromyogram (EMG) signals from the (i-1)th time period to the ith time period.

[0047] The control module can perform swallowing intention detection on the (i-1)th EEG signal and (i-1)th EMG signal during the (i-1)th time period to obtain the swallowing intention detection result. In some embodiments, the (i-1)th EEG signal of the target object may include swallowing-specific EEG signals of the target object when swallowing intention is generated. Therefore, the (i-1)th EEG signal can be combined with the (i-1)th EMG signal of the same time period to determine whether the target object has swallowing intention during the (i-1)th time period.

[0048] In some embodiments, the control module deploys a model library that stores various swallowing assessment information and model parameters associated with these assessments. Different swallowing assessments correspond to different degrees of swallowing disorders. The control module can also acquire object swallowing assessment information input by the target object and call the model library to determine the target swallowing assessment information that matches the object swallowing assessment information from the various assessments.

[0049] For example, during the model training and personalized adaptation phases, k-fold cross-validation can be used to optimize the model parameters of the initial neural network, while simultaneously building a model library associated with the rehabilitation records of multiple individuals. These rehabilitation records are uploaded to the model library with the authorization of each individual.

[0050] For example, swallowing assessment information may include at least one of the subject's status information or swallowing dysphagia assessment results. Subject status information may include at least one of the following: age, stroke type, or time of onset. Swallowing dysphagia assessment results may include at least one of the following: swallowing screening scale, functional oral feeding scale, leak-aspiration scale based on video-fluorescence swallowing imaging, or functional dysphagia scale score.

[0051] Swallowing assessment information and corresponding optimal model parameters for different subjects can be pre-entered into the model library. Thus, when the target subject uses the brain-controlled closed-loop rehabilitation training system of this disclosure, the target subject's swallowing assessment information can be input into the system. The brain-controlled closed-loop rehabilitation training system can match the target subject's swallowing assessment information with multiple swallowing assessment information stored in the model library to determine the target swallowing assessment information that matches the target subject's swallowing assessment information. The system then uses the model parameters associated with the target swallowing assessment information to configure the initial neural network, obtaining the target neural network. It should be understood that, in this disclosure embodiment, "matching" can mean that the object state information included in the object swallowing assessment information is the same as or similar to the object state information included in the target swallowing assessment information, or it can mean that the swallowing disorder assessment result included in the object swallowing assessment information is the same as or similar to the swallowing disorder assessment result included in the target swallowing assessment information; these details will not be elaborated upon here.

[0052] In some embodiments, the target neural network can preprocess the electroencephalogram (EEG) and electromyogram (EMG) signals of the target object during the training period prior to the (i-1)th time period and output the corresponding swallowing intention detection results. Thus, the model parameters of the target neural network can be adjusted based on the swallowing intention represented by the swallowing intention detection results and the actual swallowing intention of the target object, thereby improving the accuracy of the swallowing intention detection results output by the target neural network. For example, the accuracy of the swallowing intention detection results output by the target neural network can be greater than or equal to 70%. Furthermore, the control module can determine the accuracy of the swallowing intention detection results for each training stage. If the accuracy of the swallowing intention detection results in a later training stage is lower than the accuracy of the swallowing intention detection results in a previous training stage by more than a predetermined percentage (e.g., 5%), a parameter optimization algorithm can be triggered to adjust parameters such as the attention weights of the target neural network, thereby improving the accuracy of the swallowing intention detection results output by the target neural network. Each training stage can be divided based on a training period (e.g., each natural day), which is not limited here.

[0053] In some embodiments, the initial neural network is based on a convolutional neural network and a long short-term memory network, that is, the target neural network is based on a convolutional neural network and a long short-term memory network. The control module can splice the (i-1)th EEG signal and the (i-1)th EMG signal to obtain the (i-1)th spliced ​​signal, and use the target neural network to process the (i-1)th spliced ​​signal to obtain the swallowing intention detection result.

[0054] For example, during training, the target subject can be guided to generate the intention to swallow through videos, animations, or other means. The motor intention generated by the target subject can be characterized using electroencephalogram (EEG) and electromyographic (EMG) signals. For example, the preset maximum detection duration for inducing the intention could be T. maxOnline analysis of EEG and EMG signals can be performed using a sliding window mechanism. The sliding window length can be set to T. w The step size of the sliding window can be set to T. s At the maximum detection duration T max Within this timeframe, the effective time window sequence obtained sequentially is as follows:

[0055] (1).

[0056] Where W1, W2, ..., W Q These represent Q time windows. Q is an integer greater than 1.

[0057] If motor intention is detected based on multimodal physiological signals (such as EEG and EMG signals mentioned above) within any time window, electrical stimulation can be initiated immediately. If the maximum detection duration T is within this timeframe... max If no motion intent is detected, the maximum detection time T will be reached. max Electrical stimulation is forcibly initiated at specific times. Each training process adopts a cyclical pattern of "intent detection - stimulus execution - rest," that is, after the predetermined stimulation duration, there is a pause for a period of rest, and then the above process is repeated.

[0058] In some embodiments, the multimodal signal acquisition module can preprocess the EEG and EMG signals for each time window. The number of EEG channels can be C. EEG For EEG signals, a 50Hz notch filter can be used to remove power frequency interference, and a 0.5~100Hz bandpass filter can be used to retain the effective neural electrical signal frequency band, outputting a pre-processed EEG signal with a high signal-to-noise ratio. The number of electromyography (EMG) channels can be C. EMG For electromyography (EMG) signals, a 10-300Hz bandpass filter can be used to remove low-frequency baseline drift and high-frequency noise, outputting a pre-processed EMG signal.

[0059] In some embodiments, the Qth time window can be used as the aforementioned (i-1)th time period, and the preprocessed (i-1)th EEG signal and the preprocessed (i-1)th EMG signal can be spliced ​​together in the channel dimension to obtain the (i-1)th spliced ​​signal. Subsequently, the (i-1)th spliced ​​signal can be processed using a target neural network to obtain the swallowing intention detection result.

[0060] However, in some embodiments, the EEG and EMG signals of each time period from the first to the (i-1)th time period can be spliced ​​together to obtain spliced ​​signals for each of the first to the (i-1)th time periods. The target neural network is then used to process the spliced ​​signals for each of the first to the (i-1)th time periods to obtain the swallowing intention detection result.

[0061] Specifically, for ease of description, the following will use the t-th time period (also called the t-th time window) as an example, but it should be understood that the t-th time period can be the i-1 time period mentioned above. In some embodiments, a multimodal signal matrix can be constructed based on the EEG and EMG signals collected in the t-th time period, where t is a positive integer less than or equal to T, and T is a positive integer.

[0062] (2).

[0063] Where X(t) represents the data in the t-th time window, i.e., the EEG and EMG signals in the t-th time period. e1(t)…e CEEG (t) represents C EEG The brain signals collected by each brainwave channel. m1(t)...m CEMG (t) represents C EMG The electromyographic signals acquired by each of the individual electromyographic channels. t represents time. C represents the total number of channels, C = C0 EEG + C EMG . Let N represent the real number field. N represents the total number of sampling points for both EEG and EMG signals during time period t.

[0064] In some embodiments, during real-time operation, multimodal physiological signals are segmented using a sliding window. The data for the t-th time window is represented as follows: For each channel's data within each time window, perform minimum-maximum normalization:

[0065] (3).

[0066] in, This represents the nth sampling point of the c-th channel within the t-th time window. This represents the minimum value among all sampling points of the c-th channel within the t-th time window. This represents the maximum value among all sampling points of the c-th channel within the t-th time window. This represents the normalized value. Normalized window data. for:

[0067] (4).

[0068] Subsequently, the normalized windowed data is input into the target neural network for end-to-end recognition. For example, the target neural network can be a fusion model of convolutional neural network-long short-term memory network (CNN-LSTM).

[0069] For example, normalized window data from the first time window to the t-th time window can be input into the target neural network. In this way, the sequence input formed within the detection time period... for:

[0070] (5).

[0071] in, These represent the normalized window data for t time windows.

[0072] The target neural network can have optimized structural parameters for the characteristics of bimodal signals. For example, a convolutional neural network used to extract temporal local features of multi-channel signals contains two convolutional blocks, each consisting of a convolutional layer, a batch normalization layer, and a pooling layer. The first convolutional layer has 32 kernels, a size of 1 (channel dimension) × 5 (time dimension), a stride of 1, uses a modified linear unit as the activation function, and employs max pooling (kernel size 1 × 2). The second convolutional layer increases the number of kernels to 64, while the remaining parameters are the same as the first convolutional layer.

[0073] The first and second convolutional layers can employ a one-dimensional convolutional structure to perform convolution operations along the time dimension. Taking the first convolutional layer as an example, the convolution operation of the first convolutional layer can be represented as:

[0074] (6).

[0075] in, This can be represented as the output feature of the first convolutional layer. This represents the convolution operation. This represents the weights of the first convolutional layer. This represents the bias of the first convolutional layer. BN represents the batch normalized pooling layer. ReLU represents the activation function.

[0076] The output features of the convolutional neural network are flattened to obtain the depth features for the t-th time window. Represented as:

[0077] (7).

[0078] in, This represents the output features of the second convolutional layer. Flatten indicates flattening.

[0079] The output features of a convolutional neural network with continuous time windows are sequentially input into a long short-term memory (LSM) network to model the dynamic evolution of swallowing intent. The LSM network contains 128 hidden units, and its recursive process is as follows:

[0080] (8).

[0081] in, Let represent the hidden state vector for the t-th time window. Let represent the hidden state vector for the (t-1)th time window. This represents the depth features of the t-th time window. LSTM stands for Long Short-Term Memory network.

[0082] To prevent overfitting in the Long Short-Term Memory (LSTM) network, a random deactivation layer is added after the LSM network, with a random deactivation rate of 0.2.

[0083] (9).

[0084] in, Let represent the hidden state vector for the t-th time window. This represents the hidden state vector of the t-th time window output of the randomly deactivated layer. Dropout represents random deactivation.

[0085] To enhance the discriminative contribution of key time windows, an attention mechanism is introduced to weight the output of the Long Short-Term Memory network:

[0086] (10).

[0087] (11).

[0088] in, w represents the hidden state vector of the t-th time window output by the randomly deactivated layer. T This represents the transpose of the weight vector w. This represents the attention weight for the t-th time window. This represents the hidden state vector of the T-th time window output by the randomly deactivated layer. This represents the features after attention weighting.

[0089] Attention-weighted features By inputting a fully connected layer and a normalized exponential classifier, we obtain the motion intent probability vector p at the current time step.

[0090] (12).

[0091] in, This represents the weight matrix of the fully connected layer. This represents the bias vector of the fully connected layer. Softmax can correspond to the normalization exponent.

[0092] In some embodiments, the Long Short-Term Memory (LSTM) network can output a motion intent probability vector from multiple candidate swallowing intent detection results. These candidate swallowing intent detection results may include swallowing intent detection results representing that the target object has a swallowing intent and swallowing intent detection results representing that the target object does not have a swallowing intent. The swallowing intent detection result with the largest motion intent probability vector from the multiple candidate swallowing intent detection results can be selected as the swallowing intent detection result representing the target object's swallowing intent at time t.

[0093] In some embodiments, if the swallowing intention detection result indicates that the target object has a swallowing intention in the (i-1)th time period, it can be determined that the target object has a swallowing intention. Simultaneously, the (i-1)th time period can be defined as the intention generation time period, and the (i-1)th EEG signal can be defined as the target EEG signal. Furthermore, in the (i-1)th time period before the i-th time period, a first preliminary control signal is generated according to a pre-set first preliminary stimulation parameter, and a second preliminary control signal is simultaneously generated according to a pre-set second preliminary stimulation parameter. The control module can synchronously send the first preliminary control signal and the second preliminary control signal to the first stimulation module and the second stimulation module, respectively.

[0094] In some embodiments, a target multimodal physiological signal matching the target subject's (i)-1th multimodal physiological signal (i.e., including the aforementioned (i)-1th electromyographic signal and (i)-1th electroencephalogram) signal can be determined from multiple multimodal physiological signals stored in the rehabilitation record database. It should be understood that "matching" here can mean that the aforementioned (i)-1th electromyographic signal included in the (i)-1th multimodal physiological signal is the same as or similar to the electromyographic signal included in the target multimodal physiological signal, or it can mean that the aforementioned (i)-1th electroencephalogram signal included in the (i)-1th multimodal physiological signal is the same as or similar to the electroencephalogram signal included in the target multimodal physiological signal; further details are omitted here.

[0095] In the rehabilitation record database, multiple multimodal physiological signals can be stored and associated with different preliminary stimulation parameter sets. These preliminary stimulation parameter sets are distinct from each other, and each set includes corresponding transcranial direct current stimulation (DC) parameters and peripheral neuromuscular electrical stimulation (PMES) parameters. The preliminary stimulation parameter set associated with the target multimodal physiological signal can be designated as the target preliminary stimulation parameter set. The DC stimulation parameters within the target preliminary stimulation parameter set can be designated as the first preliminary stimulation parameters, and the PMES parameters within the target preliminary stimulation parameter set can be designated as the second preliminary stimulation parameters.

[0096] In this embodiment, by combining the swallowing intention represented by the target subject's electroencephalogram (EEG) signals and the muscle contraction represented by the electromyographic (EMG) signals, the collected EEG and EMG signals, along with the swallowing intention detection results, can be synchronized to the patient's rehabilitation record database. This provides a decision-making basis for the adaptation of the first and second preliminary stimulation parameters and the evaluation of rehabilitation effects. The muscle contraction can be determined based on at least one of the dynamic characteristics, such as amplitude changes or waveform patterns of the EMG signals.

[0097] The first stimulation module can electrically stimulate the central nervous system during the i-th time period under the control of the first preliminary control signal. The second stimulation module can electrically stimulate the peripheral nerves and muscles during the i-th time period under the control of the second preliminary control signal to assist the target subject in performing initial swallowing during the i-th time period, so as to collect the target's electroencephalogram (EEG) signal and the i-th electromyogram (EMG) signal in the second sub-time period of the i-th time period.

[0098] In some embodiments, the first stimulation module may, under the control of a first preliminary control signal, increase the intensity of electrical stimulation to the central nervous system in each time period based on the step size of the voltage increase during the i-th time period, and perform electrical stimulation on the central nervous system based on the adjusted intensity of electrical stimulation.

[0099] For example, the anode patch of the EEG electrode can be precisely attached to the cerebral cortex region corresponding to the mylohyoid muscle of the target subject's head (e.g., it can be located based on brain imaging results), and the cathode patch of the EEG electrode can be attached to the skin of the opposite shoulder (reference electrode).

[0100] When a swallowing-specific EEG signal is detected, confirming the target subject's swallowing intention, the first stimulation module initiates stimulation using a stepped voltage ramp mode. Based on a pre-set ramp step size, the voltage is gradually increased to smoothly raise the DC current from 0mA to the predetermined initial stimulation current within 1 second, reducing discomfort caused by sudden current changes. This is followed by continuous stimulation for 5 seconds. For example, the predetermined initial stimulation current can be calculated using the following formula:

[0101] I step =C tDCS ×(1-ECI) (13)。

[0102] Among them, I step This indicates the predetermined initial stimulation current. (C) tDCS This represents the first preliminary stimulation parameter. ECI can represent the EEG activation characteristics of the target object. These EEG activation characteristics can be determined based on the average power spectral density of the EEG signal. A larger EEG activation characteristic indicates stronger activation.

[0103] In some embodiments, the contact resistance of at least one of the anode electrode patch or cathode electrode patch can be monitored in real time during stimulation. If the contact resistance is greater than 10kΩ, an alarm is immediately issued and stimulation is suspended. Stimulation is resumed only after at least one of the anode electrode patch or cathode electrode patch has been properly reattached.

[0104] Furthermore, in some embodiments, a stepped voltage reduction mode can be used at the end of the stimulation phase. Based on a pre-set voltage reduction step size, the voltage is gradually reduced to lower the current to 0mA. For example, the current can be reduced to 0mA within 2 seconds, improving the safety and comfort of the stimulation. The stimulation is then stopped to allow for the next intent detection. In some embodiments, the target object can be trained for multiple rounds until a preset training duration is reached or the training is stopped manually.

[0105] In some embodiments, the second stimulation module may, under the control of a second preliminary control signal, acquire the muscle strength assessment results of the peripheral neuromuscular system during the i-th time period, and perform electrical stimulation on the peripheral neuromuscular system based on the muscle strength assessment results.

[0106] For example, electromyographic electrodes (e.g., two medical-grade adhesive electrodes) can be symmetrically attached to the surface of the mylohyoid muscle of the target subject, with the electrode arrangement aligned with the direction of the muscle fibers. When a swallowing-specific EEG signal is detected, the second stimulation module initiates stimulation. The stimulation waveform is preferably a bidirectional square wave, with a stimulation frequency set, for example, in the range of 50–100 Hz, and a pulse width, for example, in the range of 200–400 μs. The initial stimulation current value is set based on the target subject's muscle strength assessment results (e.g., the initial current value range is 1–5 mA), ideally sufficient to induce visible muscle contraction without causing pain. For example, an 80 Hz square wave stimulation frequency and a stimulation pulse width of 300 μs can be used. After 6 seconds of continuous stimulation, stimulation can be paused for 5 seconds to allow for the next intention detection. This cycle continues until the preset training duration is reached or the stimulation is stopped manually. It should be noted that this embodiment is merely an example. In practice, in some embodiments, the stimulation periods of the second stimulation module and the first stimulation module can be synchronized.

[0107] Based on this, in the embodiments of this disclosure, the first stimulation module and the second stimulation module can synchronously implement synergistic stimulation based on the needs of the target object during the swallowing stage.

[0108] In some embodiments, the control module can determine a baseline average power spectral density based on pre-stored EEG signals of the target object during full swallowing. For example, the average power spectral density of the alpha band (8-13 Hz) of the central-lateral sensorimotor EEG during full swallowing can be calculated based on pre-stored EEG signals as the baseline average power spectral density.

[0109] The control module can determine the target average power spectral density based on the target EEG signal. For example, the average power spectral density of the alpha band (8~13Hz) of the central-lateral sensorimotor EEG of the target object during the intention generation period can be calculated as the target average power spectral density based on the target EEG signal.

[0110] The control module can determine the EEG activation characteristics of the cerebral cortex region during the intention generation period based on the target average power spectral density and the baseline average power spectral density. For example, it can calculate the ratio of the baseline average power spectral density to the target average power spectral density and select the smaller value between this ratio and a predetermined value as the EEG activation characteristic. For example, the predetermined value could be 0.9. It should be noted that if the target average power spectral density is less than the baseline average power spectral density, the baseline average power spectral density can be updated according to the target average power spectral density so that the next round of training can be performed based on the updated baseline average power spectral density.

[0111] In some embodiments, EEG activation characteristics can be determined according to the following formula:

[0112] ECI = min(0.9, P) min / P current (14).

[0113] ECI stands for Electroencephalographic Activation Characteristics. min P represents the reference average power spectral density. current This represents the target average power spectral density corresponding to the aforementioned target EEG signal.

[0114] The control module can determine the activation level of the i-th muscle in the i-th time period based on the i-th contraction level index and the baseline contraction index.

[0115] For example, during stimulation, electromyographic (EMG) signals from the mylohyoid muscle during contraction are acquired in real time using EMG electrodes. The degree of contraction can be determined based on at least one of these EMG signals, such as the amplitude or velocity of contraction. It should be understood that the degree of contraction determined from the EMG signal acquired at time i is the i-th degree of contraction index, and the degree of contraction determined from the EMG signal acquired when the target subject is making full swallowing effort is the baseline degree of contraction index.

[0116] For example, with a window length of 1 second and a step length of 1 second, the contraction amplitude index (which may include the root mean square of electromyographic signals) can be calculated using the following formula:

[0117] (15).

[0118] Where A represents the contraction amplitude index. N represents the total number of electromyographic signal sampling points within the current time period (e.g., the time period during which the target subject can swallow with full force, or the i-th time period mentioned above). n This represents the nth sampling point of the electromyography signal of the target object in the current time period, where n is a positive integer less than or equal to N.

[0119] The contraction rate index is calculated using the following formula:

[0120] (16).

[0121] Where V represents the contraction velocity index. N represents the total number of electromyographic signal sampling points within the current time period (e.g., the time period during which the target subject can swallow with full force, or the i-th time period). n This represents the nth sampling point of the electromyography (EMG) signal of the target object in the current time period. n+1 This represents the (n+1)th sampling point of the electromyography (EMG) signal of the target object in the current time period. n is a positive integer less than or equal to N.

[0122] In some embodiments, the degree of shrinkage index can be determined based on the shrinkage amplitude index and the shrinkage speed index:

[0123] (17).

[0124] (18).

[0125] Where MCI represents the degree of contraction index. w1 represents the first predetermined weight. w2 represents the second predetermined weight. A represents the magnitude of contraction index. V represents the rate of contraction index.

[0126] Based on this, the contraction state of muscles during stimulation can be effectively quantified.

[0127] In some embodiments, the i-th muscle activation value can be obtained by dividing the real-time measured i-th contraction degree index by the baseline contraction degree index. The activation level of the i-th muscle can be used to reflect the activation level of the mylohyoid muscle in real time and serve as feedback for dynamic adjustment of stimulation parameters. Thus, the stimulation intensity can be corrected in real time through the aforementioned closed-loop control strategy.

[0128] In some embodiments, the control module can determine the activation level of peripheral neuromuscular tissue based on the activation level value of the i-th muscle, and in response to detecting that the activation intensity meets a predetermined condition, determine a first stimulation parameter based on EEG activation characteristics, an i-th contraction degree index, and a first preliminary stimulation parameter. For example, the first stimulation parameter can be determined based on EEG activation characteristics, an i-th contraction degree index, the first preliminary stimulation parameter, and a first adjustable coefficient. A second stimulation parameter is determined based on the i-th difference between the activation level value of the i-th muscle and a predetermined activation level threshold, the EEG activation characteristics, and the second preliminary stimulation parameter.

[0129] For example, a doctor or therapist can set a predetermined activation threshold (MCI) based on the target individual's muscle strength assessment results before training begins. target .

[0130] When the activation level of the i-th muscle is... Greater than the predetermined activation threshold When the activation intensity meets the predetermined conditions, the first stimulation parameter can be calculated using the following formula:

[0131] (19).

[0132] Among them, T out This represents the parameter of the first stimulus. C tDCS This indicates the initial stimulation parameters. ECI represents the EEG activation characteristics. k t This represents the first adjustable coefficient.

[0133] The second stimulus parameter can be calculated using the following formula:

[0134] (20).

[0135] Where, N out This represents the second stimulus parameter. C NMES This represents the second preliminary stimulus parameter. k f Indicates the second adjustable coefficient. ECI represents the brainwave activation signature. MCI target Indicates the predetermined activation threshold. MCI actual This represents the activation level of the i-th muscle.

[0136] In this embodiment, the electrical stimulation intensity of the first stimulation module is proportional to the degree of active effort exerted by the target subject for swallowing. When the target subject has both strong intention (high EEG activation characteristics) and good execution (high muscle activation value), a stronger central stimulation can be given, thereby strengthening this successfully operating neuromuscular pathway based on the cerebral cortex and peripheral neuromuscular regions. Simultaneously, the second stimulation module can reduce the electrical intensity to prevent muscle fatigue.

[0137] Thus, the transcranial direct current stimulation achieved by the first stimulation module and the peripheral neuromuscular electrical stimulation achieved by the second stimulation module are no longer two independent treatments, but rather form an organic whole through EEG and EMG signals, jointly serving the rehabilitation goal of "intention-driven swallowing action".

[0138] In other embodiments, the control module may determine a first stimulation parameter based on EEG activation characteristics and a first preliminary stimulation parameter in response to detecting that the activation intensity does not meet a predetermined condition. For example, the first stimulation parameter may be determined based on EEG activation characteristics, the first preliminary stimulation parameter, and a first adjustable coefficient. Furthermore, the control module may determine a second stimulation parameter based on the i-th difference between the activation level value of the i-th muscle and a predetermined activation level threshold, the EEG activation characteristics, and a second preliminary stimulation parameter. For example, the second stimulation parameter may be determined based on the i-th difference between the activation level value of the i-th muscle and the predetermined activation level threshold, the EEG activation characteristics, the second preliminary stimulation parameter, and a second adjustable coefficient.

[0139] When the activation level value of the i-th muscle is MCI actual Less than the predetermined activation threshold MCI target The first stimulus parameter can be calculated using the following formula:

[0140] (twenty one).

[0141] Among them, T out This represents the parameter of the first stimulus. C tDCS This indicates the first preliminary stimulation parameters. ECI represents the brain electrical activation characteristics.

[0142] The second stimulus parameter can be calculated using the following formula:

[0143] (twenty two).

[0144] Where, N out This represents the second stimulus parameter. C NMES This represents the second preliminary stimulus parameter. k f Indicates the second adjustable coefficient. ECI represents the brainwave activation signature. MCI target Indicates the predetermined activation threshold. MCI actual This represents the activation level of the i-th muscle.

[0145] In some embodiments, the first stimulation module may first electrically stimulate the cerebral cortex region based on a first preliminary stimulation parameter to polarize the cerebral cortex region related to swallowing movements. At this time, the electrical stimulation intensity of the second stimulation module is proportional to the "gap between intention and execution". For example, when the user strongly desires to swallow (EEG activation characteristic close to 1), but the muscles only contract weakly (the i-th contraction degree index is very small), the second stimulation parameter will be relatively large, and the brain-controlled closed-loop rehabilitation training system of this embodiment will provide strong assistance to help achieve the swallowing action.

[0146] Furthermore, in some embodiments of this disclosure, the above method describes a feedback path implemented based on electroencephalogram (EEG) and electromyogram (EMG) signals. However, in other embodiments of this disclosure, the intensity of electrical stimulation can also be adjusted based on the subjective feedback from the target subject.

[0147] In this embodiment, upon receiving an instruction from the target object, the control module adjusts a first stimulation parameter and a second stimulation parameter according to the electrical stimulation intensity indicated by the instruction, resulting in adjusted first and second stimulation parameters. Subsequently, the control module generates a first control signal based on the adjusted first stimulation parameter and a second control signal based on the adjusted second stimulation parameter.

[0148] For example, the target can input an instruction corresponding to their own feelings (such as "too strong stimulation", "too weak stimulation" or "no discomfort") through the buttons or voice interaction module equipped in the control module. The control module can adjust the first stimulation parameter and the second stimulation parameter according to the instruction. For example, when the patient inputs "too strong stimulation", the first adjustable coefficient and the second adjustable coefficient are immediately reduced.

[0149] The control module has built-in coefficient adjustment thresholds, and all coefficient adjustments are performed within safe ranges. For example, the stimulation current range of the first stimulation module is limited to 0~2mA, and the stimulation current range of the second stimulation module is limited to 1~20mA. The stimulation frequency range of these two modules can be limited to 50~100Hz to reduce the safety risks to the target subject caused by the adjusted first and second stimulation parameters. After each parameter adjustment, the control module can automatically record the adjustment time, parameter values ​​before and after the adjustment, and changes in feedback signals, providing data support for subsequent optimization of personalized rehabilitation plans.

[0150] Based on this, in this embodiment of the disclosure, the control module can realize real-time feedback regulation stimulation based on the objective feedback path realized by EEG signals and EMG signals and the subjective feedback path based on the actual instructions of the target object, thereby realizing intention-driven and active participation in the swallowing training process, and realizing dynamic adjustment of the first stimulation parameter and the second stimulation parameter, further improving the rehabilitation effect.

[0151] In addition, during stimulation, high-frequency sawtooth wave detection can be performed on electromyographic signals to identify possible spastic abnormal contractions in real time.

[0152] For example, the control module can perform high-frequency sawtooth wave detection on the (i+j)th electromyographic signal in the (i+j)th time period from (i+2)th to (i+J)th time period to determine the (i+j)th high-frequency energy ratio of the (i+j)th electromyographic signal. Here, the (i+j)th high-frequency energy ratio represents the proportion of the (i+j)th high-frequency sawtooth wave in the (i+j)th electromyographic signal to the total (i+j)th electromyographic signal, where J is an integer greater than or equal to 2, and j is an integer greater than or equal to 2 and less than or equal to J.

[0153] For example, the control module can perform high-frequency sawtooth wave detection on the electromyography (EMG) signal of the i+j time period from the i+2 to the i+J time period to determine the high-frequency energy ratio of the i+j EMG signal. This includes: performing a fast Fourier transform on the i+j EMG signal to obtain the corresponding spectrum, and performing high-frequency sawtooth wave detection on the i+j EMG signal based on the spectrum to determine the high-frequency energy ratio.

[0154] For example, the high-frequency energy ratio of an electromyographic signal can be determined using the following formula:

[0155] (twenty three).

[0156] Among them, R HF Y(f) represents the high-frequency energy ratio. Y(f) represents the frequency spectrum. f represents the sampling frequency of the electromyographic signal.

[0157] The control module can determine the contraction degree index of the peripheral neuromuscular region based on the electromyographic signal of the i+jth time period. For example, the contraction degree index of the i+jth time period can be determined by formulas (15) to (18) above, which will not be elaborated here.

[0158] The control module can determine the activation level of the (i+j)th muscle based on the (i+j)th contraction level index and the baseline contraction index. For example, the activation level of the (i+j)th muscle can be determined based on the ratio of the (i+j)th contraction level index to the baseline contraction index.

[0159] The control module can determine the abnormal spasm result of peripheral neuromuscular nerves based on the high-frequency energy ratio and muscle activation level value of time periods i+2 to i+J. For example, if the muscle activation level value of time periods i+2 to i+J is greater than a predetermined overactivation threshold, and the high-frequency energy ratio of time periods i+2 to i+J is greater than a predetermined spasm threshold, the control module can generate a spasm detection result characterizing abnormal peripheral neuromuscular nerve spasm, and immediately generate a first stop signal and a second stop signal. The first stimulation module can stop the electrical stimulation of the central nervous system under the control of the first stop signal, and the second stimulation module can stop the electrical stimulation of the peripheral neuromuscular nerves under the control of the second stop signal, and can resume stimulation after the muscles relax. If the muscle activation level value of time periods i+2 to i+J is less than or equal to the predetermined overactivation threshold, and / or the high-frequency energy ratio of time periods i+2 to i+J is less than or equal to the predetermined spasm threshold, the control module can generate a spasm detection result characterizing normal peripheral neuromuscular nerve spasm.

[0160] Based on this, the control module can realize the synchronous triggering of central and peripheral stimulation, coordinated adjustment of parameters, and dynamic optimization based on feedback. It can also record rehabilitation process data, evaluate rehabilitation effects, and ensure the safety of the stimulation process.

[0161] Figure 2 A schematic diagram of a brain-controlled closed-loop rehabilitation training system according to another embodiment of the present disclosure is shown.

[0162] like Figure 2 As shown, the brain-controlled closed-loop rehabilitation training system in this embodiment may include the aforementioned control module, the aforementioned multimodal signal acquisition module, the aforementioned control module, the aforementioned first stimulation module, and the aforementioned second stimulation module. Furthermore, the brain-controlled rehabilitation training system may also include a power management module and a wireless transmission module. The wireless transmission module can be used to realize data transmission, and the power management module can be used to power the brain-controlled rehabilitation training system. For example, the power management module may be electrically connected to the aforementioned control module, the aforementioned multimodal signal acquisition module, the aforementioned control module, the aforementioned first stimulation module, the aforementioned second stimulation module, and the aforementioned wireless transmission module. The control module may also be electrically connected to the aforementioned multimodal signal acquisition module, the aforementioned control module, the aforementioned first stimulation module, the aforementioned second stimulation module, and the aforementioned wireless transmission module.

[0163] The aforementioned control module, multimodal signal acquisition module, control module, first stimulation module, second stimulation module, power management module, and wireless transmission module can be packaged in a portable housing. The brain-controlled rehabilitation training system of this disclosure can adopt a portable, integrated design, supporting wireless communication and remote data monitoring and control.

[0164] Those skilled in the art will understand that the features described in the various embodiments of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments of this disclosure can be combined and / or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0165] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A brain-controlled closed-loop rehabilitation training system for swallowing function training, characterized in that, include: The multimodal signal acquisition module is configured to acquire electromyographic signals of the peripheral neuromuscular system of the target object used for swallowing, as well as electroencephalographic signals of the cerebral cortex region corresponding to the peripheral neuromuscular system. A control module, connected to the multimodal signal acquisition module, is configured to: when the target object is detected to have a swallowing intention, determine the i-th contraction degree index of the peripheral neuromuscular system based on the i-th electromyography signal acquired by the multimodal signal acquisition module in the i-th time period, and generate a first control signal and a second control signal based on the target EEG signal from the multimodal signal acquisition module, the i-th contraction degree index, the pre-stored EEG signal of the target object under full swallowing, and a benchmark contraction index corresponding to the contraction degree of the peripheral neuromuscular system under full swallowing, wherein the i-th time period is located after the intention generation period of the swallowing intention, the target EEG signal is the EEG signal acquired by the multimodal signal acquisition module during the intention generation period, and i is an integer greater than 1; A first stimulation module and a second stimulation module are each connected to the control module. The first stimulation module is configured to electrically stimulate the cerebral cortex region during the (i+1)th time period under the control of the first control signal. The second stimulation module is configured to electrically stimulate the peripheral neuromuscular region during the (i+1)th time period under the control of the second control signal, so as to assist the target object in swallowing during the (i+1)th time period.

2. The brain-controlled closed-loop rehabilitation training system according to claim 1, characterized in that, The control module is further configured to: perform swallowing intention detection on the (i-1)th EEG signal and (i-1)th EMG signal in the (i-1)th time period to obtain a swallowing intention detection result; when the swallowing intention detection result indicates that the target object has a swallowing intention in the (i-1)th time period, generate a first preliminary control signal according to a first preliminary stimulation parameter, generate a second preliminary control signal according to a second preliminary stimulation parameter, determine the (i-1)th time period as the intention generation time period, and determine the (i-1)th EEG signal as the target EEG signal, where i is an integer greater than 1; The first stimulation module is further configured to electrically stimulate the cerebral cortex region during the i-th time period under the control of the first preliminary control signal, and the second stimulation module is further configured to electrically stimulate the peripheral neuromuscular region during the i-th time period under the control of the second preliminary control signal, so as to assist the target object in swallowing during the i-th time period.

3. The brain-controlled closed-loop rehabilitation training system according to claim 2, characterized in that, The first stimulation module is further configured to: under the control of the first preliminary control signal, during the i-th time period, increase the electrical stimulation intensity of the cerebral cortex region based on the boost step size and perform electrical stimulation on the cerebral cortex region based on the adjusted electrical stimulation intensity; The second stimulation module is further configured to: acquire the muscle strength assessment results of the peripheral neuromuscular system, and, under the control of the second preliminary control signal, electrically stimulate the peripheral neuromuscular system according to the muscle strength assessment results during the i-th time period.

4. The brain-controlled closed-loop rehabilitation training system according to claim 2, characterized in that, The control module is also configured to: Based on the pre-stored EEG signals of the target object during the full swallow, the baseline average power spectral density is determined; Based on the target EEG signal, determine the target average power spectral density; The EEG activation characteristics of the cerebral cortex region are determined based on the target average power spectral density and the benchmark average power spectral density. Based on the i-th contraction degree index and the baseline contraction index, the activation degree value of the i-th muscle in the i-th time period of the peripheral neuromuscular system is determined. The activation intensity of the peripheral neuromuscular system is determined based on the activation level value of the i-th muscle. In response to detecting that the activation intensity meets a predetermined condition, a first stimulation parameter is determined based on the EEG activation characteristics and the first preliminary stimulation parameter, and the first control signal is generated based on the first stimulation parameter. Based on the i-th difference between the i-th muscle activation level value and the predetermined activation level threshold, the EEG activation characteristics, and the second preliminary stimulation parameters, the second stimulation parameters are determined, and the second control signal is generated based on the second stimulation parameters.

5. The brain-controlled closed-loop rehabilitation training system according to claim 4, characterized in that, The control module is also configured to: In response to the detection that the activation intensity does not meet the predetermined condition, the first stimulation parameter is determined based on the EEG activation characteristics, the i-th contraction degree index, and the first preliminary stimulation parameter; The second stimulation parameter is determined based on the i-th difference between the i-th muscle activation level value and the predetermined activation level threshold, the EEG activation characteristics, and the second preliminary stimulation parameter.

6. The brain-controlled closed-loop rehabilitation training system according to claim 4 or 5, characterized in that, The control module is also configured to: Upon receiving an instruction from the target object, the first stimulation parameter and the second stimulation parameter are adjusted according to the electrical stimulation intensity indicated by the instruction, to obtain the adjusted first stimulation parameter and the adjusted second stimulation parameter. The first control signal is generated based on the adjusted first stimulation parameter; The second control signal is generated based on the adjusted second stimulation parameters.

7. The brain-controlled closed-loop rehabilitation training system according to any one of claims 2 to 5, characterized in that, The control module is also configured to: The i-1th EEG signal and the i-1th EMG signal are spliced ​​together to obtain the i-1th spliced ​​signal; The swallowing intention detection result is obtained by processing the (i-1)th spliced ​​signal using a target neural network, wherein the target neural network is based on a convolutional neural network and a long short-term memory network.

8. The brain-controlled closed-loop rehabilitation training system according to claim 7, characterized in that, The control module is equipped with a model library, which stores various swallowing assessment information and model parameters associated with the various swallowing assessment information. Different swallowing assessment information corresponds to different degrees of swallowing disorders. The control module is also configured to: Obtain the swallowing assessment information input by the target object; The model library is invoked to determine the target swallowing assessment information that matches the subject's swallowing assessment information from the various swallowing assessment information. The initial neural network is configured using model parameters stored in association with the target swallowing assessment information to obtain the target neural network.

9. The brain-controlled closed-loop rehabilitation training system according to any one of claims 1 to 5, characterized in that, The control module is also configured to: High-frequency sawtooth wave detection is performed on the electromyographic signal of the i+jth time period in the i+jth time period from the i+2th time period to the i+Jth time period to determine the i+jth high-frequency energy ratio of the i+jth electromyographic signal. The i+jth high-frequency energy ratio represents the proportion of the i+jth high-frequency sawtooth wave in the i+jth electromyographic signal to the i+jth electromyographic signal. J is an integer greater than or equal to 2, and j is an integer greater than or equal to 2 and less than or equal to J. The contraction degree index of the peripheral neuromuscular muscle is determined based on the electromyographic signal of the (i+j)th time period. The activation level of the (i+j)th muscle is determined based on the (i+j)th contraction degree index and the baseline contraction index. The results of the peripheral neuromuscular spasm detection are determined based on the high-frequency energy ratio from time period i+2 to time period i+J and the muscle activation level values ​​from time period i+2 to time period i+J. When the spasm detection results characterize the abnormal spasm of the peripheral neuromuscular system, a first stop signal and a second stop signal are generated. The first stimulation module is further configured to stop electrical stimulation of the cerebral cortex region under the control of the first stop signal, and the second stimulation module is further configured to stop electrical stimulation of the peripheral neuromuscular region under the control of the second stop signal.

10. The brain-controlled closed-loop rehabilitation training system according to any one of claims 1 to 5, characterized in that, The cerebral cortex region includes the central-lateral sensorimotor area; The peripheral neuromuscular muscles include the mylohyoid muscle.

Images