Method of stimulating hypoglossal nerve and nerve stimulation device

Abstract

The application relates to the field of information processing and discloses a hypoglossal nerve stimulation method and a nerve stimulation device. The nerve stimulation device comprises a multi-axis sensing unit, an execution unit and a pulse emission unit. The multi-axis sensing unit periodically acquires motion parameters of the nerve stimulation device in multiple axial directions at a first period. The execution unit periodically acquires the motion parameters in the multiple axial directions at a second period, removes gravity noise, obtains denoised data, determines an inhalation period and an exhalation period of a human body based on the denoised data, controls the pulse emission unit to emit pulses in the inhalation period, and prohibits pulse emission in the exhalation period. The problems that the existing hypoglossal nerve stimulation device causes patient discomfort and high power consumption due to continuous stimulation, or increases surgical trauma due to dependence on an external pressure sensor, are solved.

Description

Technical Field

[0001] This application relates to the field of information processing, and in particular to methods and devices for stimulating the hypoglossal nerve. Background Technology

[0002] Obstructive sleep apnea (OSA) is a serious sleep disorder in which patients experience repeated collapse of the upper airway during sleep, leading to pauses in breathing, decreased blood oxygen levels, and sleep disruption. This condition increases the risk of health problems such as high blood pressure, heart disease, and diabetes, and can also cause daytime sleepiness, poor concentration, and even affect cognition and mood.

[0003] Treatment options for OSA are diverse, primarily including lifestyle modifications, the use of mechanical assistive devices, and surgical intervention. Common non-surgical treatments include continuous positive airway pressure (CPAP), which delivers a continuous flow of air through a mask to prevent upper airway collapse. CPAP is the most widely used treatment, but some patients find it difficult to tolerate. For these patients, hypoglossal nerve stimulation (HNS) devices offer a new treatment option. These devices stimulate the hypoglossal nerve with electrical impulses, prompting the tongue to move forward and reducing airway obstruction. The device is surgically implanted and monitors breathing patterns in real time, sending electrical impulses to the hypoglossal nerve during inspiration. Stimulation of the hypoglossal nerve causes the tongue to move forward, expanding the upper airway space, preventing airway collapse, thereby reducing apnea and hypoventilation events, and improving sleep quality. HNS devices provide OSA patients with an effective and sustainable treatment option, significantly improving their quality of life and overall health.

[0004] Currently, the mainstream hypoglossal nerve stimulation devices mainly employ two approaches: continuous stimulation and external pressure sensors. However, with continuous stimulation, once the user turns on the device, stimulation is continuously output, keeping the hypoglossal nerve constantly activated and the tongue lolling forward, causing significant discomfort. Furthermore, continuous stimulation greatly increases the device's power consumption, reducing its continuous operating time. While external pressure sensors can detect the patient's breathing, stimulating only during inhalation and de-stimulating during exhalation to improve comfort, the sensor must be implanted in the chest (in the right intercostal extrapleural region, away from the heart), increasing the difficulty of the implantation surgery and patient trauma. Summary of the Invention

[0005] The purpose of this application is to provide a method and device for stimulating the hypoglossal nerve, in order to solve the problems of existing hypoglossal nerve stimulation devices causing patient discomfort and excessive power consumption due to continuous stimulation, or increasing surgical trauma due to reliance on external pressure sensors. The goal is to enable the nerve stimulation device to achieve accurate perception and on-demand stimulation of the respiratory cycle without additional surgical trauma, thereby balancing patient comfort and device power consumption.

[0006] To address the aforementioned technical problems, embodiments of this application provide a method for stimulating the hypoglossal nerve, applied to a nerve stimulation device implanted in the human body. The nerve stimulation device includes: a multi-axis sensing unit, an execution unit communicatively connected to the multi-axis sensing unit, and a pulse delivery unit communicatively connected to the execution unit. The method includes: the multi-axis sensing unit periodically acquiring motion parameters of the nerve stimulation device along multiple axes in a first cycle; the execution unit periodically acquiring the motion parameters along the multiple axes in a second cycle, removing gravity noise from the motion parameters along the multiple axes to obtain noise-reduced data, and determining the respiratory cycle of the human body based on the noise-reduced data, the respiratory cycle including an inspiratory cycle and an expiratory cycle; wherein the second cycle is greater than or equal to the first cycle; the execution unit controls the pulse delivery unit to deliver pulses during the inspiratory cycle and prohibits the delivery of pulses during the expiratory cycle.

[0007] Embodiments of this application also provide a neural stimulation device, comprising: a multi-axis sensing unit, an execution unit communicatively connected to the multi-axis sensing unit, and a pulse delivery unit communicatively connected to the execution unit; the multi-axis sensing unit is configured to periodically acquire motion parameters of the neural stimulation device in multiple axes in a first cycle; the execution unit is configured to periodically acquire the motion parameters in the multiple axes in a second cycle, perform gravity noise removal on the motion parameters in the multiple axes to obtain noise-reduced data, and determine the respiratory cycle of the human body based on the noise-reduced data, wherein the respiratory cycle includes an inspiratory cycle and an expiratory cycle; wherein the second cycle is greater than or equal to the first cycle; the execution unit is configured to control the pulse delivery unit to deliver pulses during the inspiratory cycle and to prohibit the delivery of pulses during the expiratory cycle.

[0008] In this embodiment, motion parameters are acquired by a multi-axis sensing unit in a first cycle, providing basic data for respiratory monitoring. The execution unit acquires these motion parameters in a second cycle and performs gravity noise stripping to obtain denoised motion parameters. Based on these denoised parameters, the inspiratory and expiratory cycles of the human body are determined. This allows the neurostimulation device to automatically identify the respiratory cycle through its built-in multi-axis sensing unit, eliminating the need for an external pressure sensor implanted in the chest. This avoids the increased surgical difficulty and patient trauma caused by increased implantation surgery. Because the execution unit controls the pulse delivery unit to deliver pulses during the inspiratory cycle and prohibits pulse delivery during the expiratory cycle, precise on-demand stimulation is achieved. Neuromodulation is performed only when the airway needs to be opened (i.e., during the inspiratory cycle), avoiding the discomfort caused by continuous stimulation that keeps the patient's hypoglossal nerve constantly activated. Simultaneously, because the pulses are delivered only during part of the respiratory cycle rather than continuously, the overall power consumption of the device is effectively reduced, avoiding the problem of significantly increased power consumption and reduced device usage time caused by continuous stimulation. Moreover, since the second cycle is greater than or equal to the first cycle, meaning that the data processing and judgment cycle of the execution unit can be longer than the acquisition cycle of the multi-axis sensing unit, this non-real-time processing mechanism further helps to reduce the computational power consumption of the execution unit. Attached Figure Description

[0009] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations do not constitute a limitation on the embodiments, and unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0010] Figure 1 This is a schematic diagram of the structure of a nerve stimulation device according to an embodiment of this application; Figure 2 This is a schematic diagram illustrating the effect of gravity on the data collected by the multi-axis sensing unit in one embodiment of this application; Figure 3 This is a schematic diagram of the curve corresponding to the data collected by the multi-axis sensing unit in one embodiment of this application when the patient is in a lying position; Figure 4 This is a schematic diagram of the curve corresponding to the data collected by the multi-axis sensing unit in one embodiment of this application when the patient is in a prone position; Figure 5 This is a schematic diagram of a process for determining the human respiratory cycle based on noise-reduced velocity data in one embodiment of this application; Figure 6 This is a schematic diagram of the structure of a nerve stimulation device according to another embodiment of this application; Figure 7 This is a flowchart of a method for stimulating the hypoglossal nerve based on a preset time period in another embodiment of this application; Figure 8A schematic diagram illustrating a specific example of determining the human respiratory cycle based on noise-reduced data in one embodiment of this application. Detailed Implementation

[0011] Currently, mainstream hypoglossal nerve stimulation devices have problems such as high power consumption, significant discomfort to patients, or the need to surgically implant an external pressure sensor into the patient's chest, which increases the difficulty of the implantation surgery and increases patient trauma.

[0012] Therefore, this application provides a method for stimulating the hypoglossal nerve, which can be applied to a nerve stimulation device implanted in the human body. The nerve stimulation device includes: a multi-axis sensing unit, an execution unit communicatively connected to the multi-axis sensing unit, and a pulse delivery unit communicatively connected to the execution unit. The multi-axis sensing unit periodically acquires motion parameters of the nerve stimulation device along multiple axes in a first cycle; the execution unit periodically acquires motion parameters along multiple axes in a second cycle, removes gravity noise from the motion parameters along multiple axes to obtain denoised data, and determines the human respiratory cycle based on the denoised data. The respiratory cycle includes an inspiratory cycle and an expiratory cycle; wherein the second cycle is greater than or equal to the first cycle; the execution unit controls the pulse delivery unit to deliver pulses during the inspiratory cycle and prohibits pulse delivery during the expiratory cycle. This addresses the problems of existing hypoglossal nerve stimulation devices causing patient discomfort and excessive power consumption due to continuous stimulation, or increasing surgical trauma due to reliance on external pressure sensors. It enables the nerve stimulation device to achieve accurate perception and on-demand stimulation of the respiratory cycle without additional surgical trauma, thus balancing patient comfort and device power consumption.

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of the present invention to facilitate a better understanding of this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.

[0014] In this embodiment of the application, the method for stimulating the hypoglossal nerve is applied to, for example... Figure 1The illustrated neurostimulation device, implanted in the human body, includes a multi-axis sensing unit 101, an execution unit 102, and a pulse delivery unit 103. The multi-axis sensing unit acquires motion parameters, such as acceleration, of the neurostimulation device along multiple axes during human respiration. The multi-axis sensing unit 101 can be disposed externally to the neurostimulation device, for example, on the surface of the housing; or it can be disposed internally to the neurostimulation device, for example, within the internal space defined by the housing. The multi-axis sensing unit 101 is communicatively connected to the execution unit 102. For example, the multi-axis sensing unit 101 and the execution unit 102 can be disposed in the same circuit and connected via a bus. Alternatively, the multi-axis sensing unit 101 and the execution unit 102 can be wirelessly connected. Similarly, the execution unit 102 is communicatively connected to the pulse delivery unit 103, either by being disposed in the same circuit and connected via a bus, or by being wirelessly connected. This application does not limit the specific connection methods between the multi-axis sensing unit 101 and the execution unit 102, or between the execution unit 102 and the pulse delivery unit 103. The electrical signal emitted by the pulse delivery unit 103 is transmitted to the neck through the wire 201 implanted under the skin, thereby stimulating the hypoglossal nerve.

[0015] In some exemplary embodiments, the multi-axis sensing unit is a triaxial accelerometer, providing motion parameters along multiple axes, including accelerations along the x, y, and z axes that are mutually perpendicular. In other examples, the multi-axis sensing unit can be a sensor that includes triaxial acceleration sensing functionality. For example, the multi-axis sensing unit can be a six-axis sensor that includes triaxial acceleration sensing and rotational angular velocity sensing functionality, or a six-axis sensor that includes triaxial acceleration sensing and orientation sensing functionality. Yet another example is a nine-axis sensor that includes triaxial acceleration sensing, rotational angular velocity sensing, and orientation sensing functionality.

[0016] In this embodiment, each data set acquired by the triaxial accelerometer includes acceleration values ​​along three axes: x-axis, y-axis, and z-axis, which are mutually perpendicular. This embodiment does not impose any particular limitation on the sampling precision of the acceleration values ​​along each axis; for example, sampling precision can be 8-bit, 16-bit, or 24-bit. Similarly, this embodiment does not impose any particular limitation on the sampling frequency of the triaxial accelerometer, which can be set according to the overall system's accuracy requirements. For example, the sampling frequency of the triaxial accelerometer can be 100Hz. The frequency at which the execution unit acquires the collected data can be equal to or less than the sampling frequency of the triaxial accelerometer.

[0017] Because the data collected by the multi-axis sensing unit reflects the absolute value of the acceleration of the nerve stimulation device, it cannot reflect the posture of the human body. Therefore, it cannot determine whether the current acceleration is in the inspiratory or expiratory cycle, and thus cannot determine whether to deliver or stop the pulse based on a zero-point reference. Furthermore, the data collected by the multi-axis sensing unit is significantly affected by gravity, causing considerable interference to the data. Figure 2 As shown, this data is from a lying position, where the motion direction is consistent with the Z-axis. It can be seen that the Z-axis data collected by the multi-axis sensing unit is significantly affected by gravity, causing considerable interference. Therefore, in order to eliminate the influence of gravity on the sampled data, in this embodiment, the execution unit 102 needs to strip gravity noise from the motion parameters in multiple axes to obtain denoised data.

[0018] In one example, the acceleration along one of multiple axes is determined to be closest to the gravitational acceleration G. The sine values ​​of the accelerations along the other two axes are used as the denoised data. For example, for the obtained acceleration values ​​along three axes, the acceleration value of the axis closest to the G value is discarded, and the sine values ​​of the accelerations along the other two axes are used as the denoised data to obtain the zero point. Alternatively, the acceleration along the axis closest to the gravitational acceleration G can be filtered. The filtered acceleration is then moduloed with the accelerations along the other two axes to obtain denoised data. For example, the axis closest to G is the z-axis. For the z-axis acceleration value, methods such as median filtering, wavelet transform, Kalman filtering, Fourier transform, and frequency domain filtering can be used to remove the influence of gravity. The obtained value z' is then moduloed with the acceleration values ​​along the other two axes to obtain the denoised data for finding the zero point. .

[0019] In some examples, the acceleration of the axis closest to the gravitational acceleration G among multiple axes is determined. This includes: determining the acceleration of the axis closest to the gravitational acceleration G among multiple axes in real time; or, determining the acceleration of the axis closest to the gravitational acceleration G among multiple axes at regular intervals; the period of the regular determination is greater than a second period; or, determining the acceleration of the axis closest to the gravitational acceleration G among multiple axes when a change in human posture is detected. Specifically, since human posture is dynamic, the influence of gravity on the data collected by the three-axis accelerometer is constantly changing. In one example, the axis whose acceleration value is closest to the G value among the three axes is detected in real time and discarded. In another example, the axis whose acceleration value is closest to the G value among the three axes is detected at regular intervals, and the acceleration value of that axis is continuously discarded until the next detection reconfirms the axis that needs to be discarded. The frequency is less than the data acquisition frequency of the execution unit. In another example, when a change in human posture is detected, it is reconfirmed which axis's acceleration value needs to be discarded.

[0020] In one example of this application, the respiratory cycle of the human body can be determined based on the noise-reduced data in the following manner: First, the zero-point data is determined based on the average of the denoised data from multiple respiratory cycles. Specifically, as mentioned above, the data collected by the triaxial accelerometer is in absolute value, and a zero point (i.e., a baseline) needs to be obtained as the basis for subsequent processing. In this embodiment, the current zero point is obtained based on historical denoised data. In one example, the current zero point is obtained based on the average of historical denoised data from multiple respiratory cycles. The average here can be one of the arithmetic mean, geometric mean, harmonic mean, or weighted average. For the weighted average, preferably, the closer to the current respiratory cycle, the higher the weight. Preferably, the previous 20-30 respiratory cycles are taken. In one example, the current zero point is obtained by averaging the extreme values ​​of the historical denoised data for each respiratory cycle in multiple respiratory cycles. Preferably, the extreme values ​​of the historical denoised data from the previous 10 respiratory cycles are taken. Here, the extreme values ​​are the minimum value of the corresponding trough and the maximum value of the corresponding peak.

[0021] After determining the zero-point data, if the first time interval from the moment the data crosses the zero-point upwards to the moment it crosses the zero-point downwards in the waveform curve corresponding to the denoised data is greater than the second time interval from the moment it crosses the zero-point downwards to the moment it crosses the zero-point upwards, then the period from the trough to the peak is determined to be the intake phase; if the first time interval is less than the second time interval, then the period from the peak to the trough is determined to be the intake phase. Specifically, the execution unit determines when and how the data crosses the zero point based on the acquired denoised data and the zero point. For example... Figure 3As shown, the execution unit compares the acquired denoised data with the zero point: if the first n data points are all greater than the zero point, while the current denoised data is less than the zero point, it indicates that the denoised data has crossed the zero point downwards. The execution unit controls the timing unit to record the occurrence time as the first time T0. If the first m data points are all less than the zero point, while the current denoised data is greater than the zero point, it indicates that the denoised data has crossed the zero point upwards. The execution unit controls the timing unit to record the occurrence time as the second time T1. This process is repeated to obtain a second first time T0' and a second second time T1'. This embodiment does not impose any particular restrictions on the specific values ​​of n and m, which can be set according to the sampling rate and the patient's condition. To prevent noise interference, the current data can be smoothed, and the processed data is compared with the zero point. This embodiment does not impose any particular limitations on the specific methods used for data smoothing, such as Moving Average, Exponential Moving Average (EMA), Median Filtering, Savitzky-Golay Filtering, Wavelet Transform, or Gaussian Smoothing. The first and second times here can be relative times based on a certain starting point (e.g., the time when equipment debugging is completed, the time when equipment is initialized), or absolute real times. Due to different human postures, the curves corresponding to the obtained data will represent different meanings. Figure 3 The curve corresponds to the patient's lying position. Figure 4 The curves represent the patient's prone position. Since the curves are similar in shape, it's difficult to determine whether the time interval from the trough to the peak represents inhalation, or vice versa. The inventors discovered that there is a transition period between exhalation and inhalation; therefore, as... Figure 3 As shown, if the difference between the second first time T0' and the first second time T1 (i.e., the first time interval) It is greater than the difference between the first second time interval T1 and the first time interval T0 (i.e., the second time interval). This indicates that the period from the trough to the crest represents the patient's inhalation phase. For example... Figure 4 As shown, if the difference between the second first time T0' and the first second time T1 is... Less than the difference between the first second time T1 and the first time T0 This indicates that the period from the peak to the trough represents the patient's inhalation phase.

[0022] Then, if the period from the trough to the peak is the inhalation phase, the time of the trough is determined based on the moment of downward crossing of the zero-point data and the moment of the next upward crossing of the zero-point data; and the time of the peak is determined based on the moment of the next upward crossing of the zero-point data and the moment of the next downward crossing of the zero-point data. If the period from the peak to the trough is the inhalation phase, the time of the peak is determined based on the moment of downward crossing of the zero-point data and the moment of the previous upward crossing of the zero-point data; and the time of the trough is determined based on the moment of downward crossing of the zero-point data and the moment of the next upward crossing of the zero-point data. Within the determined time from the trough to the peak, the pulse delivery unit is controlled to deliver a pulse. Specifically, the execution unit obtains the times corresponding to the maximum and minimum values ​​within the expiratory cycle, i.e., the times corresponding to the peak and trough, based on the acquired first and second times. Figure 3 As shown, the minimum time corresponds to the minimum value. ,in The maximum value corresponds to the maximum time. ,in .like Figure 4 As shown, the maximum value corresponds to the maximum value time. ,in The minimum time corresponding to the minimum value ,in In this way, the time corresponding to the extreme value can be determined without obtaining the extreme value within the respiratory cycle, which is more convenient. When the patient is inspiratory during the period from the trough to the peak, the execution unit will... Pulses are delivered within a specific timeframe. When the patient inhales from the peak to the trough, the execution unit delivers the pulse during the next respiratory cycle. Pulses are emitted within a specified time period.

[0023] In another example of this application, the human respiratory cycle can also be determined based on the denoised data in the following manner: First, a target interval subset is extracted from the denoised data. This target interval subset includes data between a maximum value and its adjacent minimum value, or data between a minimum value and its adjacent maximum value. The similarity between each of several pre-labeled standard datasets and the target interval subset is obtained. These pre-labeled standard datasets include standard datasets of human inspiratory cycles in several preset postures and standard datasets of human expiratory cycles in several preset postures. Based on the similarity between the target interval subset and each standard dataset, the human respiratory cycle is determined.

[0024] In other words, when implanting the neurostimulation device, the multi-axis sensing unit is first calibrated with the human posture to acquire standard acceleration data for the inspiratory and expiratory cycles in each of multiple human postures. After acquiring the collected data, the execution unit determines whether the extreme value range falls within the inspiratory or expiratory cycle of a certain posture based on the collected data and the standard data.

[0025] In some examples, the step of obtaining the similarity between each pre-labeled standard dataset and the target interval subset can employ any one or any combination of Jaccard similarity coefficient, cosine similarity, Euclidean distance, and Manhattan distance to obtain the similarity between each of the pre-labeled standard datasets and the target interval subset. Exemplarily, multiple data sets are collected to form a dataset (e.g., data collected based on the duration of the previous respiratory cycle). An interval subset between the maximum and minimum values ​​of the denoised data is obtained from this dataset. Then, methods such as Jaccard similarity coefficient, cosine similarity, Euclidean distance, and Manhattan distance are used to analyze the similarity between the interval subset and the acceleration standard data to obtain the similarity between the two. This determines whether the interval subset corresponds to the patient's breathing in a certain posture, and further determines whether the interval subset corresponds to an inspiratory or expiratory cycle based on whether the maximum to minimum value in that posture corresponds to an inspiratory or expiratory cycle. In other embodiments, a subset of intervals between the minimum and maximum values ​​of the denoised data can be obtained from the collected dataset. Then, similar methods can be used to analyze similarity and determine whether the subset corresponds to the patient's inspiratory or expiratory cycle in a particular posture. To improve analysis accuracy, multiple subsets of intervals between the maximum and minimum values ​​of the denoised data can be used. Similar methods such as Jaccard Similarity, Cosine Similarity, Euclidean Distance, and Manhattan Distance can be employed to perform similarity analysis with a standard dataset formed by standard acceleration data for each posture's inspiratory and / or expiratory cycles, thus determining whether the subset corresponds to the patient's inspiratory or expiratory cycle in that posture.

[0026] In other examples, the step of obtaining the similarity between each pre-labeled standard dataset and the target interval subset can also involve determining the surface formed by the target interval subset and the standard surface formed by each standard dataset. The similarity between the surface formed by the target interval subset and the standard surface is then used as the similarity between each standard dataset and the target interval subset. In other words, the similarity can be analyzed from the perspectives of surface shape and geometric features by comparing the surface formed by the interval subset with the standard surface formed by the acceleration standard data of each inspiratory or expiratory cycle at each posture, to determine whether the interval subset corresponds to the patient's inspiratory or expiratory cycle at that posture. For example, similarity can be analyzed using shape context, surface normal differences, Hausdorff distance, point cloud comparison algorithms such as ICP and FPFH (Fast Point Feature Histograms), and surface feature descriptors. The process involves determining whether the interval subset corresponds to the patient's inspiratory or expiratory cycle in that posture, thereby obtaining the duration of the inspiratory cycle, which serves as the basis for the stimulation delivery in the next cycle.

[0027] In other examples, the neurostimulation device may also include a pose sensor for acquiring the human body's pose. The respiratory cycle can also be determined based on the denoised data in the following ways: Extract a target interval subset from the denoised data. The target interval subset includes motion parameters from the maximum value to the adjacent minimum value, or the target interval subset includes motion parameters from the minimum value to the adjacent maximum value.

[0028] Among multiple pre-calibrated standard datasets, a first standard dataset and a second standard dataset are determined. The multiple pre-calibrated standard datasets include: standard datasets of inspiratory cycles for the human body in multiple preset postures, and standard datasets of expiratory cycles for the human body in multiple preset postures. The first standard dataset is the standard dataset of inspiratory cycles in the same posture currently detected by the posture sensor; the second standard dataset is the standard dataset of expiratory cycles in the same posture currently detected by the posture sensor.

[0029] Obtain a subset of the target interval and calculate its similarity to the first and second standard datasets, respectively. Based on the similarity of the subset of the target interval to the first and second standard datasets, determine the human respiratory cycle.

[0030] By incorporating a pose sensor into the neurostimulation device, only the standard acceleration data corresponding to the pose needs to be extracted. This eliminates the need to obtain the patient's current posture through similarity methods. Instead, the device directly determines whether the interval subset (or the interval subset from the minimum to the maximum acceleration value) corresponds to the patient's inspiratory or expiratory cycle in that posture, based on its corresponding period. This allows for the determination of the inspiratory cycle duration, which serves as the basis for the next stimulation cycle. This simplifies the calculation process and improves response speed. In this embodiment, determining whether the patient is in an inspiratory or expiratory cycle during data acquisition can be done in real-time or periodically. In this embodiment, the determination frequency can be less than or equal to the data acquisition frequency of the execution unit.

[0031] In the example above, the denoised data is the denoised acceleration data, which is used to determine the inhalation cycle and obtain its duration. In other examples, the denoised data can also be the denoised velocity data.

[0032] Specifically, the inspiratory cycle is determined by the rate of breathing, and its duration is obtained. While acquiring data from a triaxial sensor, the velocity data of the object in the x, y, and z directions is obtained through integration. The sensor should have a high sampling rate (e.g., 20 Hz to 100 Hz) to ensure the capture of rapidly changing motion. For each time interval Δt, an acceleration is assumed... a ( t The value remains constant during this time period, and the update rate is v(t+Δt)=v(t)+a(t). Similarly, the data along the axis whose acceleration is closest to the G value is removed to obtain the denoised velocity data. In some cases, higher-order numerical integration methods, such as the Runge-Kutta method, can be considered to improve the accuracy of velocity estimation. Additionally, data from other sensors (such as gyroscopes) can be combined with a Kalman filter to correct the velocity estimate in real time. The Kalman filter can dynamically adjust the estimate based on the prediction and observation models, reducing accumulated errors. Periodic calibration using known conditions (such as the velocity being zero at rest) can correct for drift errors that may occur during long-term operation. Through the above methods, triaxial acceleration data can be converted into velocity data, such as... Figure 5 As shown.

[0033] Similarly, the velocity obtained based on acceleration integral also requires obtaining the velocity zero point. Clearly, when the respiratory velocity crosses zero, it can be considered the transition between the expiratory and inspiratory cycles. The specific method for obtaining the velocity zero point can refer to the method for obtaining the acceleration zero point described above, and will not be repeated here. In one example, after obtaining the moment the velocity crosses zero, the time interval between two velocity zero-crossings can be obtained. After obtaining two adjacent time intervals, based on the characteristic that the expiratory cycle duration is longer than the inspiratory cycle duration, the shorter time interval is determined as the inspiratory cycle and used as the basis for the next pulse delivery. In another example, similarly, due to different human postures, the curves corresponding to the obtained velocity data represent different meanings. It is necessary to perform a similarity judgment between the noise-reduced velocity data and the calibrated velocity data to determine the patient's current posture, and based on this posture, whether the velocity crossing zero upwards represents an inspiratory or expiratory cycle, and whether the noise-reduced velocity interval above zero corresponds to an inspiratory or expiratory cycle, thereby determining the duration of the inspiratory cycle as the basis for the next pulse delivery. The specific method for determining the upward zero point of velocity can be found in the above-mentioned method for determining the zero point of acceleration. The method for determining the human respiratory cycle based on the noise-reduced velocity data can also be found in the above-mentioned method for determining the human respiratory cycle based on the noise-reduced acceleration data, which will not be elaborated here.

[0034] In some specific implementations, the execution unit can be configured to start performing the above operations at night to further save energy. For example... Figure 6 As shown, the neurostimulation device also includes a clock unit 104 for acquiring real time. During the initialization of the neurostimulation device, the time at which the execution unit begins performing the aforementioned operations is set, such as... Figure 7 As shown, after power-on, the execution unit determines whether the current time falls within a preset specific time period representing evening, such as 9 PM to 6 AM. If it does, the above operation is executed; otherwise, it enters standby mode. The execution unit can obtain the real time through the clock unit and then determine whether to execute the above operation based on the real time. Alternatively, the clock unit can be replaced by a timer unit. When the timer unit overflows, it triggers the execution unit, which then begins executing the above operation, and the timer unit is reset.

[0035] In other examples, the multi-axis sensing unit also includes a storage module for storing motion parameters along multiple axes. The execution unit periodically retrieves these motion parameters in batches from the storage module at a second cycle, where the second cycle is longer than the first cycle used by the multi-axis sensing unit to acquire the motion parameters. For example, the multi-axis sensing unit acquires data at a frequency of 100Hz (i.e., the first cycle is 10 milliseconds), and the execution unit retrieves the motion parameters in batches from the storage module at a lower frequency, such as 10Hz (i.e., the second cycle is 100 milliseconds).

[0036] The above configuration decouples data acquisition and data processing in terms of timing, allowing the multi-axis sensing unit to maintain high-frequency sampling to ensure the integrity of the original data, while the execution unit can perform intermittent batch reading and processing at a lower frequency. This significantly reduces the real-time computing load of the execution unit and the overall system power consumption while ensuring the accuracy of respiratory cycle recognition, which is beneficial to extending the battery life of the implantable device. An effective balance is achieved between high-frequency sampling and low-power processing, which can significantly reduce the power consumption of the system on the one hand, and increase the stability and reliability of respiratory cycle data on the other hand.

[0037] The embodiments of this application do not impose any particular limitation on the type of storage unit for the multi-axis sensing unit. In one example, the storage unit can be a register, which uses a first-in-first-out (FIFO) method for easy reading. In another instance, the storage unit can be an external memory connected via a data bus, such as a NOR flash memory. This embodiment does not impose any particular limitation on the capacity of the storage unit; it can be determined comprehensively based on the sampling frequency of the multi-axis sensing unit, the reading frequency of the execution unit, the size of the acquired data, and the system accuracy requirements.

[0038] Figure 8 This illustrates a specific example of determining the human respiratory cycle based on denoised data. For example... Figure 8As shown, the multi-axis sensing unit is a triaxial accelerometer with a sampling frequency of 10Hz, meaning it samples once every 0.1s and stores the data in a register. The register can store 10 data points, forming a dataset. The execution unit receives data at a frequency of 1Hz, meaning it reads the register once every 1s to obtain 10 sampled data points. Gravity denoising is then applied to the sampled data to obtain a denoised dataset. In this example, the method for obtaining the zero point is similar to that in the above embodiment and will not be described again. In the figure, the triaxial accelerometer samples 80 times within 8 seconds, and the execution unit obtains a total of 8 datasets, each containing 10 sampled data points. From 0s to 1s, the execution unit determines that the denoised data corresponding to the data obtained by the triaxial accelerometer is all greater than zero. From 1s to 2s, the execution unit determines that the first part of the denoised data is above zero, and the second part is below zero, meaning that the denoised data begins to cross the zero point. Specifically, the first four denoised data points are above zero, while the next six are below zero. The execution unit confirms this at time T0, which is 1.4s. From the 2nd to the 3rd second, the execution unit determines that all denoised data points are below zero. This is also true from the 3rd to the 4th, 4th to the 5th, and 5th to the 6th second. From the 6th to the 7th second, the execution unit determines that the first portion of the denoised data points are below zero, while the latter portion are above zero, meaning the denoised data is beginning to cross zero. Specifically, the first eight denoised data points are below zero, while the next two are above zero. The execution unit confirms this at time T1, which is 6.8s. From the 7th to the 8th second, the execution unit determines that all denoised data points corresponding to the data sampled by the triaxial accelerometer are greater than zero. This process is repeated sequentially. Between 14 and 15 seconds, the execution unit determines that the first portion of the denoised data is above zero, while the second portion is below zero. This means the denoised data is starting to cross zero again. Specifically, the first four denoised data points are above zero, and the last six are below zero. The execution unit then confirms the second first time interval T0' as 14.4 seconds. The execution unit uses a similar method to obtain the difference between the second first time interval T0' and the first second time interval T1. The difference between the first second time T1 and the first time T0 . Therefore, the period from the trough to the peak represents the patient's inhalation phase. Simultaneously, the minimum time corresponds to the minimum value. The maximum value corresponds to the maximum time. Therefore, the execution unit controls the pulse delivery unit to deliver stimulation pulses within the time range of [14.4+4.1-1.4, 14.4+12.5-1.4] in the next respiratory cycle.

[0039] Motion parameters are acquired in the first cycle by a multi-axis sensing unit, providing basic data for respiratory monitoring. The execution unit acquires these motion parameters in the second cycle and performs gravity noise stripping to obtain denoised motion parameters. Based on these, the inspiratory and expiratory cycles are determined, allowing the neurostimulation device to automatically identify the respiratory cycle through its built-in multi-axis sensing unit, eliminating the need for an external pressure sensor implanted in the chest. This avoids the increased surgical difficulty and patient trauma associated with implantation. Because the execution unit controls the pulse delivery unit to deliver pulses during the inspiratory cycle and disables pulse delivery during the expiratory cycle, precise on-demand stimulation is achieved. Neuromodulation is performed only when airway opening is needed (i.e., during the inspiratory cycle), avoiding the discomfort caused by continuous stimulation that keeps the patient's hypoglossal nerve constantly activated. Simultaneously, because pulses are delivered only during certain respiratory cycles rather than continuously, the overall power consumption of the device is effectively reduced, avoiding the problem of significantly increased power consumption and reduced device usage time caused by continuous stimulation.

[0040] The examples mentioned in the above embodiments can be freely combined, and any combination can be understood as an embodiment. The terms "embodiment" or "example" appearing in various locations in the specification do not necessarily refer to the same embodiment, nor are they independent or alternative embodiments mutually exclusive with other embodiments. Those skilled in the art will understand that the embodiments described herein can be combined with other embodiments.

[0041] Another embodiment of this application relates to a neurostimulation device, comprising: a multi-axis sensing unit, an execution unit communicatively connected to the multi-axis sensing unit, and a pulse delivery unit communicatively connected to the execution unit. The multi-axis sensing unit is used to periodically acquire motion parameters of the neurostimulation device along multiple axes in a first cycle; the execution unit is used to periodically acquire motion parameters along multiple axes in a second cycle, remove gravity noise from the motion parameters along multiple axes to obtain denoised data, and determine the human respiratory cycle based on the denoised data, the respiratory cycle including an inspiratory cycle and an expiratory cycle; wherein the second cycle is greater than or equal to the first cycle; the execution unit is used to control the pulse delivery unit to deliver pulses during the inspiratory cycle and to prohibit pulse delivery during the expiratory cycle.

[0042] It is not difficult to see that this embodiment is a device embodiment corresponding to the method embodiment, and this embodiment can be implemented in conjunction with the method embodiment. The relevant technical details mentioned in the method embodiment are still valid in this embodiment, and will not be repeated here to reduce repetition. Correspondingly, the relevant technical details mentioned in this embodiment can also be applied to the method embodiment.

[0043] Furthermore, in order to highlight the innovative aspects of this application, no units that are not closely related to solving the technical problems proposed in this application are introduced in this embodiment, but this does not mean that there are no other units in this embodiment.

[0044] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A method for stimulating the hypoglossal nerve, characterized in that, A neurostimulation device for implantation in the human body, the neurostimulation device comprising: a multi-axis sensing unit, an execution unit communicatively connected to the multi-axis sensing unit, and a pulse delivery unit communicatively connected to the execution unit; the method comprising: The multi-axis sensing unit periodically acquires the motion parameters of the neural stimulation device in multiple axes in a first cycle. The execution unit periodically acquires motion parameters along the multiple axes in a second cycle, removes gravity noise from the motion parameters along the multiple axes to obtain noise-reduced data, and determines the human body's respiratory cycle based on the noise-reduced data. The respiratory cycle includes an inhalation cycle and an exhalation cycle; wherein the second cycle is greater than or equal to the first cycle. The execution unit controls the pulse delivery unit to deliver pulses during the inspiratory cycle and to prohibit the delivery of pulses during the expiratory cycle.

2. The method for stimulating the hypoglossal nerve according to claim 1, characterized in that, The motion parameters along the multiple axes include the accelerations of the x-axis, y-axis, and z-axis, which are perpendicular to each other. The process of removing gravity noise from the motion parameters along the multiple axes to obtain denoised data includes: Determine the acceleration of the axis among the plurality of axes that is closest to the value of gravitational acceleration G; The chord values ​​of the accelerations in the other two axes are used as the denoised data; or, the acceleration of the axis closest to the gravitational acceleration G is filtered, and the modulus of the filtered acceleration and the accelerations in the other two axes is taken to obtain the denoised data.

3. The method for stimulating the hypoglossal nerve according to claim 2, characterized in that, Determining the acceleration of the axis among the plurality of axes that is closest to the value of gravitational acceleration G includes: In real time, determine the acceleration of the axis among the plurality of axes that is closest to the value of gravitational acceleration G; or, The acceleration of the axis among the plurality of axes that is closest to the value of gravitational acceleration G is determined periodically; the period for determining the timing is greater than the second period; or, When a change in the posture of the human body is detected, the acceleration of the axis among the plurality of axes that is closest to the value of gravitational acceleration G is determined.

4. The method for stimulating the hypoglossal nerve according to claim 1, characterized in that, Determining the human respiratory cycle based on the denoised data includes: Zero-point data is determined based on the average of the denoised data from multiple respiratory cycles; If the first time interval from the moment when the noise-reduced data passes upward through the zero-point data to the moment when it passes downward through the zero-point data is greater than the second time interval from the moment when it passes downward through the zero-point data to the moment when it passes upward through the zero-point data, then the inhalation phase is determined to be from the trough to the peak. If the first time interval is less than the second time interval, the inhalation phase is determined to be from the peak to the trough.

5. The method for stimulating the hypoglossal nerve according to claim 4, characterized in that, The execution unit controls the pulse delivery unit to deliver pulses during the inhalation cycle, including: If the inhalation phase is from the trough to the peak, the time of the trough is determined based on the moment when the wave passes downward through the zero point data and the moment when the wave passes upward through the zero point data again; and the time of the peak is determined based on the moment when the wave passes upward through the zero point data again and the moment when the wave passes downward through the zero point data again. If the inhalation phase is from peak to trough, the time of the peak is determined based on the moment when the wave passes downward through the zero-point data and the moment when the wave passes upward through the zero-point data again; and the time of the trough is determined based on the moment when the wave passes downward through the zero-point data and the moment when the wave passes upward through the zero-point data again. Within the time interval from the determined trough to the determined peak, the pulse emission unit is controlled to emit pulses.

6. The method for stimulating the hypoglossal nerve according to claim 4, characterized in that, The average includes any of the following averages: Arithmetic mean, geometric mean, harmonic mean, weighted mean.

7. The method for stimulating the hypoglossal nerve according to claim 1, characterized in that, Determining the human respiratory cycle based on the denoised data includes: Extract a target interval subset from the denoised data. The target interval subset includes data between the maximum value and the adjacent minimum value, or the target interval subset includes data between the minimum value and the adjacent maximum value. Obtain the similarity between each of the multiple pre-labeled standard datasets and the target interval subset; wherein, the multiple pre-labeled standard datasets include: standard datasets of human breathing cycles in multiple preset postures, and standard datasets of human breathing cycles in multiple preset postures. The respiratory cycle of the human body is determined based on the similarity between the target interval subset and each of the standard datasets.

8. The method for stimulating the hypoglossal nerve according to claim 7, characterized in that, The process of obtaining the similarity between each of the multiple pre-labeled standard datasets and the target interval subset includes: Using any one or any combination of Jaccard similarity coefficient, cosine similarity, Euclidean distance, and Manhattan distance, the similarity between each of the pre-labeled standard datasets and the target interval subset is obtained. Alternatively, determine the surface formed by the target interval subset and the standard surface formed by each of the standard datasets, and use the similarity between the surface formed by the target interval subset and the standard surface as the similarity between each standard dataset and the target interval subset.

9. The method for stimulating the hypoglossal nerve according to claim 1, characterized in that, The neural stimulation device further includes: a posture sensor for acquiring the posture of the human body; Determining the human respiratory cycle based on the denoised data includes: Extract a target interval subset from the denoised data. The target interval subset includes motion parameters from the maximum value to the adjacent minimum value, or the target interval subset includes motion parameters from the minimum value to the adjacent maximum value. Among multiple pre-calibrated standard datasets, a first standard dataset and a second standard dataset are determined; wherein, the multiple pre-calibrated standard datasets include: standard datasets of the inhalation cycles of the human body in multiple preset postures, and standard datasets of the exhalation cycles of the human body in multiple preset postures; the first standard dataset is a standard dataset of the inhalation cycles of the same posture currently detected by the posture sensor; the second standard dataset is a standard dataset of the exhalation cycles of the same posture currently detected by the posture sensor. Obtain the similarity scores between the target interval subset and the first and second standard datasets, respectively; The respiratory cycle of the human body is determined based on the similarity between the target interval subset and the first and second standard datasets, respectively.

10. The method for stimulating the hypoglossal nerve according to any one of claims 4 to 9, characterized in that, The denoised data includes denoised acceleration data or denoised velocity data.

11. The method for stimulating the hypoglossal nerve according to claim 1, characterized in that, The second period is longer than the first period; The multi-axis sensing unit is also equipped with a storage module for storing motion parameters in multiple axes; The execution unit periodically acquires motion parameters along the plurality of axes in a second cycle, including: The execution unit periodically retrieves the motion parameters along the plurality of axes from the storage module in a second cycle.

12. A nerve stimulation device, characterized in that, include: A multi-axis sensing unit, an execution unit communicatively connected to the multi-axis sensing unit, and a pulse delivery unit communicatively connected to the execution unit; The multi-axis sensing unit is used to periodically acquire motion parameters of the neural stimulation device in multiple axes in a first cycle. The execution unit is used to periodically acquire motion parameters in the plurality of axes in a second cycle, remove gravity noise from the motion parameters in the plurality of axes to obtain noise-reduced data, and determine the human body's respiratory cycle based on the noise-reduced data. The respiratory cycle includes an inhalation cycle and an exhalation cycle. The second cycle is greater than or equal to the first cycle. The execution unit is used to control the pulse delivery unit to deliver pulses during the inspiratory cycle and to prohibit the delivery of pulses during the expiratory cycle.

Images