Deep brain stimulation system for treating postural hypotension

Abstract

The invention aims to provide a deep brain stimulation system for treating postural hypotension, which comprises a data acquisition module for collecting physiological status information of a patient, and a data fusion module implanted in a pectoralis major subcutaneous area and used for receiving multi-source physiological data output by the data acquisition module, integrating the multi-source physiological data, filtering noise and outputting the multi-source physiological data to the patient. Generating comprehensive physiological status information (St = {BPt, HRt, ECGt, Post}) of the patient; the deep brain stimulation module is used for accurately stimulating a target brain region to enhance sympathetic nerve output; the response module is integrated in the data fusion module and is used for analyzing the physiological state information St and a future physiological state predicted value St + 1, dynamically adjusting the voltage, frequency and pulse width of the deep brain stimulation module and optimizing an adjustment strategy in real time according to physiological feedback data of the patient; the invention provides a deep brain stimulation system with real-time blood pressure monitoring, posture sensing and an intelligent reaction mechanism.

Description

Technical Field

[0001] The present invention belongs to the field of postural hypotension treatment, and in particular relates to a deep brain stimulation system for treating postural hypotension. Background Art

[0002] Orthostatic hypotension (OH) is a common clinical disease, mostly caused by autonomic dysfunction, manifested as a significant drop in blood pressure when the patient changes position (especially from lying or sitting to standing), which can lead to a series of uncomfortable symptoms, such as dizziness, blacking out, weakness and even syncope, seriously affecting the quality of daily life and safety of patients. In particular, the incidence of OH has increased significantly in the elderly population, becoming a key issue of clinical concern. Traditional treatment measures, such as drug adjustment, increased salt and water intake, and the use of elastic stockings, can relieve symptoms to a certain extent, but for some patients, especially those with severe symptoms, their effects are limited and may have side effects.

[0003] In recent years, deep brain stimulation (DBS) technology has made remarkable achievements in the treatment of movement disorders such as Parkinson's disease. It regulates abnormal neural activity and improves patients' motor function by precisely stimulating specific brain areas. A pioneering work by Green et al. revealed the potential of DBS in regulating the cardiovascular system (Green AL, Wang S, Owen SL, Xie K, Liu X, Paterson DJ, Stein JF, Bain PG, Aziz TZ. Deep brain stimulation can regulate arterial blood pressure in awake humans. Neuroreport. 2005 Nov 7;16(16):1741-5.). They found that stimulating different parts of the periventricular gray matter (PVG) and periaqueductal gray matter (PAG) regions can effectively increase blood pressure. This effect may be achieved by enhancing sympathetic nerve output and cardiac contractility, as well as increasing vascular tone. In addition, further studies have shown that this stimulation can maintain blood pressure stability without increasing basal blood pressure in the supine position, demonstrating its ability for fine-tuning treatment.

[0004] Therefore, developing a DBS device that integrates real-time blood pressure monitoring, body position sensing, and intelligent response, which can automatically monitor blood pressure changes when the patient's body position changes, intelligently analyze OH risks, and start or adjust DBS stimulation parameters in a timely manner, is of great significance for improving OH treatment effects and ensuring patient safety. Such a device will fill the gap in existing technologies, represent a major innovation in the field of OH treatment, and is expected to become the preferred solution for treating OH in the future. Summary of the invention

[0005] The purpose of the present invention is to provide a deep brain stimulation system with real-time blood pressure monitoring, posture sensing and intelligent response functions.

[0006] To achieve the above object, the present invention provides the following technical solution: a deep brain stimulation system for treating postural hypotension, comprising: The data acquisition module is used to collect the patient's physiological status information, including: Implantable pressure sensor, implanted in the artery, is used to monitor blood pressure data (BP) in real time. t ); ECG sensor, integrated into electronic wristwatch, used to collect ECG signals (ECG t ) and heart rate data (HR t ); The posture sensor is implanted in the subcutaneous area of ​​the pectoralis major muscle to detect the patient's body position (Pos t ); The data fusion module is implanted in the subcutaneous area of ​​the pectoralis major muscle to receive the multi-source physiological data output by the data acquisition module, integrate it, filter the noise, and generate the patient's comprehensive physiological status information (St={BP t ,HR t ,ECG t ,Pos t}); Deep Brain Stimulation Module, including: Directional electrodes, implanted in the patient's periventricular gray (PVG) and periaqueductal gray (PAG) regions, are used to precisely stimulate target brain areas to enhance sympathetic nerve output; The stimulation power supply assembly is implanted in the subcutaneous area of ​​the pectoralis major muscle and is used to provide voltage (V t ), frequency (f t ) and pulse width (PW t ) Adjustable electrical signals support segmented and directional stimulation to avoid interference with non-target areas; The response module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle to analyze the physiological status information S t, dynamically adjust the voltage, frequency and pulse width of the deep brain stimulation module, and optimize the adjustment strategy in real time based on the patient's physiological feedback data; The risk prediction module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle to process the comprehensive physiological status historical data {S t−n ,...,S t}, predict the patient's future physiological state S t+1 , and pass the predicted value to the response module; The reinforcement learning module, integrated together with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle, is used to dynamically update the stimulation parameter adjustment strategy of the response module according to the patient's physiological feedback data through the reinforcement learning algorithm.

[0007] Furthermore, the data fusion module generates the patient's real-time comprehensive physiological status information (St={BP t ,HR t ,ECG t ,Pos t}).

[0008] Furthermore, the directional electrode is composed of multiple independent electrode pieces, each of which can be individually controlled in on and off state. At the same time, each electrode piece is precisely distributed in the periventricular gray matter (PVG) and periaqueductal gray matter (PAG) areas according to the anatomical structure and treatment goals. Among them, the stimulation power supply component independently controls the on and off of each electrode piece to avoid affecting non-target areas, thereby improving the targetedness and safety of stimulation and achieving precise stimulation of specific target brain areas (such as PVG or PAG).

[0009] Further, the stimulus power supply assembly consists of a single power supply and a parallel switch matrix; A single power supply is implanted in the subcutaneous area of ​​the pectoralis major muscle of the patient, which also contains a voltage regulator, a pulse generator and a power control unit; it is used to provide adjustable stimulation parameters for the directional electrodes, including voltage (V t ), frequency (f t ) and pulse width (PW t ), where the voltage regulator adjusts the output voltage (V t ) amplitude to meet the stimulation intensity requirements of different target brain areas; the pulse generator generates the required stimulation frequency (f t ), frequency (f t ) affects the rhythm of the stimulation signal and the discharge response of the target neuron; at the same time, the pulse generator controls the pulse width (PW) by adjusting the duration of a single pulse. t ), pulse width (PW t) size determines the coverage depth and duration of each electrical stimulation signal; The parallel switch matrix is ​​located between a single power supply and directional electrodes, and includes multiple independent switches and switch control logic circuits; each switch is used to control the on / off state of one or a group of electrode sheets; the switch control logic circuit receives switch instructions from the response module and determines which switches are turned on or off, thereby configuring a specific electrode sheet activation combination for the purpose of segmented stimulation (activating a specific area) or directional stimulation (concentrating current on a specific path).

[0010] Furthermore, the response module sends the generated stimulation parameter instructions to the voltage regulator and pulse generator of the single power supply, and at the same time, sends the electrode activation instruction to the switch control logic circuit to specify the electrode sheet that needs to be activated.

[0011] Furthermore, a rule table is provided in the response module, which defines activation strategies and stimulation parameters under different states based on the patient's physiological state and treatment goals, combined with clinical data or individualized diagnosis results; the control methods of the response module include: S1, receiving the current comprehensive physiological state data (St={BP t ,HR t ,ECG t ,Pos t}); S2: Determine whether the patient needs to adjust the stimulation based on the real-time data, as well as the location and number of electrodes that need to be activated; S3, match the current physiological state with the preset rule table. If there is no complete match, select the closest matching rule or use the default safety parameters to determine the required activation area (PVG area, PAG area, or both), activation number (number and distribution of activated electrodes), stimulation parameters (voltage (V t ), frequency (f t ) and pulse width (PW t )); S4, firstly sending stimulation parameter instructions to a single power supply, and generating a target electrical stimulation signal by its internal components (voltage regulator, pulse generator); then sending a switch instruction to the switch control logic circuit of the parallel switch matrix, activating the target switch, and forming a target current path matching the electrical stimulation signal; S5, selectively activate the directional electrodes to achieve segmented and directional stimulation; at the same time, the data acquisition module continuously monitors the patient's physiological feedback (such as blood pressure BP t , heart rate HR t , electrocardiogram (ECG) t ); S6. If the feedback parameters do not reach the expected target, re-match the rule table and give priority to adjusting the stimulation parameters (such as voltage). If it is still ineffective, gradually increase the number of activated electrodes or adjust the position of the electrodes until the patient's physiological state returns to a safe range.

[0012] Furthermore, the risk prediction module outputs the predicted physiological state S at the next moment through the time series model. t+1 ={BP t+1 ,H Rt+1 ,ECG t+1 ,Pos t+1} to the response module, the decision logic of the response module is as follows: Rule table matching: Current Status t Match the rule table to obtain the immediate stimulation plan (including electrode activation and stimulation parameters); If the predicted state S t+1 Out of the safety range, matching rule table S t+1 Corresponding rules; Processing when the rule table cannot match: If the current state S t or future state S t+1 Unable to match the rule table; The response module automatically generates stimulus parameters, and the newly generated parameters and corresponding states are recorded for dynamic updating of the rule table; Decision output: Comprehensive consideration t and S t+1 The matching results generate the final stimulation instructions, including: the activated electrode combination (PVG area, PAG area or both), the stimulation parameter voltage (V t ), frequency (f t ) and pulse width (PW t ).

[0013] Furthermore, the steps for the response module to automatically generate stimulus parameters are as follows: S1. Obtain the current state S through the data fusion module t , obtain the future state S through the risk prediction module t+1 At the same time, the blood pressure target value and blood pressure deviation (ΔBP) are obtained through system presets. The blood pressure target value (BP 目标 ) is defined as: systolic blood pressure target value = baseline systolic blood pressure - 20 mmHg, diastolic blood pressure target value = baseline diastolic blood pressure - 10 mmHg; ΔBP=BP 目标 −BPt BPt is the current blood pressure monitored in real time; Deviation classification: ΔBP>0: Blood pressure is too low and stimulation intensity needs to be increased.

[0014] ΔBP≤0: Blood pressure is normal or high, no additional stimulation is needed.

[0015] S2. Based on the size of ΔBP, the stimulation parameters are generated according to the following rules: The voltage controls the intensity of the stimulation signal and directly determines the amplitude of the stimulation current. A higher voltage can more strongly activate the target neural area, but may cause unnecessary side effects (such as overstimulation). Its adjustment logic is to calculate the current blood pressure deviation ΔBP. t =Vinitial+k v ×ΔBP adjusts the voltage to increase or decrease the stimulation intensity; where k v Indicates the voltage value that needs to be increased for each unit of blood pressure deviation; at the same time, it is also necessary to ensure that the voltage does not exceed the safety range [V min ,V max ]; The frequency determines the rhythm of the stimulation signal and affects the frequency and efficiency of nerve excitation. A higher frequency can quickly accumulate stimulation effects, but may cause nerve adaptation and reduce efficacy. The adjustment logic is to calculate the current blood pressure deviation ΔBP; through f t =f 初始 +kf f ×ΔBP adjusts the frequency and controls the stimulation rhythm; where k f Indicates the frequency that needs to be increased for each unit of blood pressure deviation; at the same time, it is necessary to ensure that the frequency does not exceed the range that the nerve can tolerate [f min ,f max ]; Pulse width defines the duration of each stimulation and affects the cumulative intensity of stimulation. Prolonging the pulse width can increase the stimulation effect of each pulse, but it may increase power consumption. Its adjustment logic is to calculate the current blood pressure deviation ΔBP; through PW t =PW 初始 +k p ×ΔBP adjusts the pulse width and prolongs the single stimulation effect; where k P Indicates the pulse width that needs to be increased for each unit of blood pressure deviation; at the same time, it is necessary to ensure that the frequency does not exceed the range that the nerve can tolerate [PW min ,PW max ]; Activating more electrodes can expand the stimulation range and improve the treatment effect. According to the anatomy and target area (such as PVG and PAG), the combination of activated electrodes is reasonably selected. The adjustment logic is to calculate the current blood pressure deviation ΔBP; through the linear relationship N=N 初始 +(k n ×ΔBP) to adjust the number of electrodes and increase the stimulation coverage; where k nIndicates the number of electrodes that need to be added for each unit of blood pressure deviation; at the same time, it is also necessary to ensure that the number of activated electrodes does not exceed the maximum supported number N of the device max ; S3, the response module generates stimulation parameters according to the above adjustment rules: activation electrode combination (number N and target area (PVG, PAG or both)), stimulation parameter voltage (V t ), frequency (f t ) and pulse width (PW t ).

[0016] Furthermore, when the rule table is not completely matched or the risk prediction module has a large prediction deviation, the reinforcement learning module optimizes the stimulus parameter adjustment strategy through a feedback mechanism; the working steps of the reinforcement learning module are as follows: S1, receive the current state S from the data fusion module t , obtain the predicted status S from the risk prediction module t+1 , initialize the current Q(S t ,A t )value; S2, select actions based on the current Q-value function, using the ϵ-greedy strategy, ϵ-probability to select random actions (exploration), 1−ϵ1-probability to select the current optimal action (utilization), actions include adjusting voltage, frequency, pulse width and activating electrodes; S3, according to the selected action A t , activate the single power supply and parallel switch matrix through the response module; monitor the feedback state S after stimulation in real time t ′ and reward R t : Reward function:

[0017] S4. Update the Q-value using the Q-Learning algorithm: Among them, α is the learning rate, which controls the step size of each update, and γ is the discount factor, which measures the importance of future rewards; S5. If Q(S t ,A t ) value is stable for a long time and performs well, update it into the rule table as a new rule.

[0018] The deep brain stimulation system for treating postural hypotension proposed by the present invention combines deep brain stimulation technology with modern intelligent algorithms and has the following significant beneficial effects: 1. Accurately stimulate the target area to improve the treatment effect Directional electrodes: Through precise control of independent electrodes, it can specifically stimulate the periventricular gray matter (PVG) and periaqueductal gray matter (PAG) areas, significantly enhance sympathetic nerve output, improve blood pressure regulation, avoid interference with non-target areas, and reduce side effects.

[0019] Segmented and directional stimulation: Dynamically control the activation state of the electrode patch through a parallel switch matrix, supporting segmented stimulation of specific areas and directional concentrated current paths, further improving the accuracy and efficacy of stimulation.

[0020] 2. Real-time monitoring and dynamic adjustment to ensure safety Data acquisition module: real-time monitoring of the patient's blood pressure, heart rate, ECG signal and body position to ensure a comprehensive grasp of the patient's physiological information.

[0021] Data fusion module: fuses and filters multi-source physiological data to generate real-time comprehensive physiological status information St to ensure the accuracy and reliability of the data.

[0022] Response module: Dynamically adjust stimulation parameters (such as voltage, frequency, and pulse width) according to the real-time physiological state St and the predicted state St+1 to ensure that the stimulation effect reaches the target range while keeping the patient's physiological state within a safe range.

[0023] 3. Intelligent control to improve adaptability and self-learning ability Rule table matching: The response module has a built-in rule table to quickly match appropriate stimulation parameters and electrode activation schemes according to the patient's individual physiological state, reducing the need for manual intervention.

[0024] Risk prediction module: by predicting future physiological state S t+1 , achieve forward-looking control in stimulation parameter adjustment and avoid possible blood pressure fluctuations in advance.

[0025] Reinforcement Learning Module: In the case of incomplete rule table matching or large deviations in time series prediction, the stimulation strategy is dynamically optimized through a feedback mechanism.

[0026] It is able to gradually update and supplement the rule table through reinforcement learning algorithms, realize continuous optimization and self-adaptation of the system, and show strong advantages especially in complex or sudden situations.

[0027] 4. Personalized treatment to suit different patients The system can adjust the rule table and the initial values ​​of stimulation parameters according to the individual characteristics of the patient (such as age, body shape, medical history, etc.) to achieve personalized treatment.

[0028] The risk prediction module can optimize the prediction model based on the patient's long-term monitoring data, and the reinforcement learning module can learn the optimal treatment plan through long-term feedback.

[0029] 5. The system design is highly efficient and energy-saving, improving operational stability Single power supply vs. parallel switch matrix: A single power supply provides adjustable voltage, frequency, and pulse width, reducing hardware complexity while supporting multi-channel output to meet the needs of electrode segmentation and directional stimulation.

[0030] The parallel switch matrix dynamically activates the target electrode sheets through efficient switching logic, reducing energy consumption and extending the service life of the equipment.

[0031] 6. Reduce side effects and improve patient comfort Precise control: By optimizing stimulation parameters and activating electrode combinations, overstimulation or unnecessary energy waste can be avoided, reducing possible side effects.

[0032] Dynamic monitoring and adjustment: The system can quickly respond to changes in the patient's physiological state, avoid dizziness, fatigue or fainting caused by hypotension, and improve the patient's daily quality of life.

[0033] 7. Data-driven optimization of long-term efficacy Physiological data recording and analysis: The system records the patient's real-time physiological data and feedback results, providing doctors with a basis for long-term treatment effect evaluation.

[0034] Dynamic update of rule table: The risk prediction module and reinforcement learning module work together to continuously optimize the rule table through long-term feedback data to ensure that the system can adapt to the long-term treatment needs of patients.

[0035] 8. Innovation and clinical application potential The system integrates modern biosensor technology, intelligent control algorithms (rule table matching, time series prediction and reinforcement learning) and efficient power management technology, providing an innovative and comprehensive solution for the treatment of postural hypotension. Through accurate, real-time and personalized treatment methods, the system is likely to become the standard equipment for the treatment of postural hypotension in the future, providing patients with a safe and effective long-term treatment option.

[0036] The present invention combines deep brain stimulation technology with intelligent algorithms through the synergistic effect of multiple modules, solving the problem of poor effect and large side effects in the treatment of postural hypotension in the prior art. The system has shown significant advantages in accuracy, real-time, safety and personalized adaptation, and has extremely high clinical application value and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1It is a schematic diagram of the overall structure of the present invention; Figure 2 It is a schematic diagram of module connection in the present invention; In the figure, 100 is a directional electrode; 200 is a posture sensor; 300 is a single power supply; 400 is a pressure sensor; and 500 is an electronic watch. DETAILED DESCRIPTION

[0038] The purpose of the present invention is to provide a deep brain stimulation system for treating postural hypotension. The present invention is a deep brain stimulation system with real-time blood pressure monitoring, posture sensing and intelligent response mechanism.

[0039] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a deep brain stimulation system for treating postural hypotension, comprising: a data acquisition module, a data fusion module, a deep brain stimulation module, a response module, a risk prediction module and a reinforcement learning module; the data acquisition module is used to collect the patient's physiological state information, including: an implantable micro pressure sensor implanted in the patient's upper limb artery, which senses the change of blood flow pressure through a capacitive diaphragm structure and outputs the pressure value in real time through an LC resonant circuit integrated in an electronic watch; it has a high-sensitivity working characteristic and monitors blood pressure data BP in real time t Integrated in electronic watches, it uses an electrode-type single-lead ECG sensor that supports QRS wave detection to collect ECG signals (ECG t ) and heart rate data (HR t Implanted in the subcutaneous area of ​​the pectoralis major muscle, it is equipped with a 6-axis IMU (inertial measurement unit) and an attitude sensor with an integrated accelerometer and gyroscope to detect the patient's body position (Pos t ).

[0040] In such a closed-loop control system, BP t It directly reflects the severity of OH and is the “primary target variable” of system control; HR t It is an auxiliary tool for assessing the balance of autonomic nerves and cardiac compensation; ECG t Provides more comprehensive cardiac status information for safety monitoring and stimulation strategy fine-tuning; Pos t It can accurately detect changes in body position and is a key signal for triggering OH events. It can also serve as a forward-looking reference for risk prediction.

[0041] Among them, in order to facilitate the understanding of BP t , HR t , ECG t The relationship between the three dependent variables, BP is explained in detail here t , HR t , ECG t The detailed role played by the whole system; BP t The core function of (blood pressure data) is to be the most direct indicator for determining whether a patient has orthostatic hypotension (OH), and it is also the main reference for adjusting stimulation strategies (especially voltage, frequency, and pulse width).

[0042] Usage scenarios in the system: Determine whether stimulation is needed: When BP t When BP is within the safe range or slightly lower, the system may not stimulate or only stimulate at the lowest intensity. t When it falls below a certain threshold, the rule table or risk prediction module triggers the action of "increasing stimulation" or "changing the electrode activation method".

[0043] Real-time closed-loop feedback: After stimulation parameters are adjusted, BP t The change in blood pressure is the key to judging whether the stimulation is effective; if the blood pressure still does not recover, it is necessary to further increase the stimulation or adjust the activation electrode.

[0044] HR t (Heart rate data) is often closely related to the regulation of blood pressure and the state of the autonomic nervous system. Especially when OH occurs, some patients will experience a compensatory increase in heart rate (sympathetic nerve excitement), and there may also be a situation where heart rate changes are out of sync with blood pressure fluctuations. These can indicate the patient's current autonomic nervous function state.

[0045] Usage scenarios in the system: Assist in judging the balance of sympathetic / parasympathetic nerves: If blood pressure drops but heart rate does not rise significantly, it may indicate insufficient autonomic nervous response, and stronger DBS stimulation is needed to increase sympathetic output; if heart rate is too high but blood pressure is still low, it indicates insufficient vasoconstriction or ineffective heart rhythm compensation, and targeted stimulation is also needed; Warning of abnormal situations: whether the heart rate is too fast, too slow or ectopic rhythm occurs, which also serves as a reminder of the DBS safety boundary.

[0046] ECG t (Electrocardiogram signal) can reflect more cardiac electrophysiological characteristics, such as whether abnormalities such as premature contractions, atrial fibrillation, and premature ventricular beats occur. In patients with orthostatic hypotension, arrhythmias may sometimes occur.

[0047] Usage scenarios in the system: Safety monitoring: If ECG t If severe arrhythmia is displayed, the system will be more cautious when adjusting DBS parameters to avoid excessive stimulation frequency or current that increases the burden on the heart; Optimizing stimulation strategies: In some cases, PVG / PAG stimulation may affect the vagus nerve or other neural circuits, potentially affecting heart rhythm, ECGt These changes can be captured in time and the system can be assisted to make corresponding corrections.

[0048] The data fusion module is implanted in the subcutaneous area of ​​the pectoralis major muscle to receive the multi-source physiological data output by the data acquisition module, integrate it, filter the noise, and generate the patient's comprehensive physiological status information (S t ={BP t ,HR t ,ECG t ,Pos t}); In the above two modules, the specific working steps are as follows S1. First, the 6-axis IMU sensor (inertial measurement unit) implanted in the system records the patient's acceleration and angular velocity data in real time.

[0049] The following analysis steps are used to determine changes in body position: Static posture: The accelerometer detects the patient's current acceleration direction and distinguishes between standing, lying, and sitting postures.

[0050] Dynamic posture: The gyroscope records the angular velocity and determines the posture change (such as the transition from lying flat to standing).

[0051] Attitude change rate analysis: Calculate the acceleration value and time interval Δt of the body position change and evaluate the rate of change: ; If the rate v>v 阈值 (such as standing up quickly), flagged as a possible hypotensive triggering event.

[0052] Combined with the posture status (standing) and speed (quick standing), the system preliminarily marks potential hypotension events and enters the further data collection stage.

[0053] S2, implantable pressure sensor, real-time monitoring of blood pressure data (BP t ); Optical heart rate sensor collects heart rate data (HR t ); ECG sensor collects ECG signals (ECG t ); the integrated data forms the current state: The data fusion module integrates posture status and real-time physiological data to generate the current comprehensive status: S t ={BP t ,HR t ,ECG t ,Pos t} S3, data fusion module receives {BP t ,HR t ,ECG t ,Post}: Synchronous sampling: Align the timestamps of each signal to ensure data consistency.

[0054] Noise filtering: The Kalman filter is used to eliminate the fluctuation noise of the pressure sensor.

[0055] Feature extraction: Extract features such as blood pressure change rate ΔBP and heart rate change trend to enhance status assessment.

[0056] Weighted processing: weights are set according to the importance of the data. In this embodiment, the weight of blood pressure data is set higher than that of posture status:

[0057] S4. At the same time, the calculation module on the data fusion module also performs hypotension risk assessment; First, determine whether the current blood pressure BP is lower than the blood pressure target value (BP target), where the BP target is defined as: systolic blood pressure target value = baseline systolic blood pressure - 20 mmHg, diastolic blood pressure target value = baseline diastolic blood pressure - 10 mmHg; if the patient's real-time systolic blood pressure is less than baseline systolic blood pressure - 20 mmHg, the system marks it as a risk of hypotension, and if the heart rate HRt>100bpm, it is auxiliaryly determined as compensated hypotension (compensation is available but insufficient); posture state Pos t = Standing is an important condition for triggering risk, and low-intensity stimulation is required at this time; if the patient's real-time systolic blood pressure is less than the baseline systolic blood pressure -20 mmHg, the system marks it as a risk of hypotension. t <100bpm, auxiliary judgment is uncompensated hypotension (heart uncompensated); posture state Pos t =Standing is an important condition for triggering risks, and high-intensity stimulation is required at this time.

[0058] Deep Brain Stimulation Module, including: Directional electrodes, implanted in the patient's periventricular gray (PVG) and periaqueductal gray (PAG) regions, are used to precisely stimulate target brain areas to enhance sympathetic nerve output; The stimulation power supply assembly is implanted in the subcutaneous area of ​​the pectoralis major muscle and is used to provide an adjustable voltage (V t ), frequency (f t ) and pulse width (PW t ), supports segmented and directional stimulation, avoiding interference with non-target areas; The response module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle to analyze the physiological state information S t , dynamically adjust the voltage, frequency and pulse width of the deep brain stimulation module, and optimize the adjustment strategy in real time based on the patient's physiological feedback data; The risk prediction module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle to process the comprehensive physiological status historical data {S t−n ,...,S t}, predict the patient's future physiological state S t+1 , and pass the predicted values ​​to the response module for prospective adjustment of deep brain stimulation parameters; The reinforcement learning module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle, and is used to dynamically update the stimulation parameter adjustment strategy of the response module according to the patient's physiological feedback data through the reinforcement learning algorithm.

[0059] In one embodiment, the directional electrode is composed of a plurality of independent electrode sheets, each of which can control the on and off state individually. At the same time, each electrode sheet is precisely distributed in the periventricular gray matter (PVG) and periaqueductal gray matter (PAG) areas according to the anatomical structure and treatment goals. Among them, the stimulation power supply component independently controls the on and off state of each electrode sheet to avoid affecting non-target areas, thereby improving the targetedness and safety of the stimulation and achieving precise stimulation of specific target brain areas (such as PVG or PAG).

[0060] The stimulation power supply assembly consists of a single power supply and a parallel switch matrix; the single power supply is implanted in the subcutaneous area of ​​the patient's pectoralis major muscle, and is also equipped with a voltage regulator, a pulse generator and a power control unit; it is used to provide adjustable stimulation parameters for the directional electrodes, including voltage (V t ), frequency (f t ) and pulse width (PW t ), where the voltage regulator adjusts the output voltage (V t ) amplitude to meet the stimulation intensity requirements of different target brain areas; the pulse generator generates the required stimulation frequency (f t ), frequency (f t ) affects the rhythm of the stimulation signal and the discharge response of the target neuron; at the same time, the pulse generator controls the pulse width (PW) by adjusting the duration of a single pulse. t ), pulse width (PW t ) size determines the coverage depth and continuous effect of each electrical stimulation signal; at the same time, a power control unit is also provided in a single power supply: the power control unit realizes the power distribution and protection functions, is responsible for monitoring the output power, prevents overload or short circuit, and ensures stable operation of the system.

[0061] The parallel switch matrix is ​​located between a single power supply and directional electrodes, and includes multiple independent switches and switch control logic circuits; each switch is used to control the on / off state of one or a group of electrode sheets; the switch control logic circuit receives switch instructions from the response module and determines which switches are turned on or off, thereby configuring a specific electrode sheet activation combination for the purpose of segmented stimulation (activating a specific area) or directional stimulation (concentrating current on a specific path).

[0062] In one embodiment, the response module sends the generated stimulation parameter instructions to the voltage regulator and pulse generator of the single power supply, and at the same time, sends the electrode activation instructions to the switch control logic circuit to specify the electrode patch that needs to be activated.

[0063] The response module has a rule table, which defines the activation strategy and stimulation parameters under different states based on the patient's physiological state and treatment goals, combined with clinical data or individualized diagnosis results; here, the process of establishing the rule table is shown in detail: The first is the initial establishment process of the rule table; before the system is officially used or implanted, an "initialization" or "calibration" process is usually carried out to establish the initial rule table. The typical process is as follows: S1. Data collection and baseline measurement Collect the patient's resting blood pressure (BP) t , heart rate HR t , ECG t And position t Basic values ​​in different daily situations (lying, sitting, standing, etc.); The blood pressure and heart rate fluctuations of patients when they switched from lying to standing position without DBS stimulation were collected to evaluate the severity, occurrence rate and tolerance of OH in patients.

[0064] S2. Body position change induction test In a controlled environment (such as hospital monitoring conditions), let the patient change body position (lying → sitting or lying → standing), and measure the fluctuations of blood pressure BPt, heart rate HRt, and electrocardiogram ECGt in real time; According to the magnitude of blood pressure drop and the threshold for symptoms such as dizziness or fainting, the required stimulation voltage, frequency, and pulse width range are preliminarily determined; Determine the minimum effective stimulation value (stimulation parameters that can initially counteract the drop in blood pressure) and the maximum safe threshold (avoid overstimulation).

[0065] Segmented / directional electrode activation test Within the adjustable range, activate the electrode patches of PVG (periventricular gray matter) and PAG (periaqueductal gray matter) sequentially / combinatorially, and observe which activation combination is most effective in raising blood pressure with the least side effects; According to the test results, record the "most effective brain region stimulation combination" as the basic rule.

[0066] S3. Form the initial rule table Summarize the above data, combine the experience of clinicians and safety margins, and define a number of rules (Rule), each rule corresponding to the mapping relationship of "current state / predicted state → target stimulation plan"; For example, if the blood pressure is too low (BP t <X1), and the heart rate is high (HR t >Y1), then activate the electrodes in the low voltage + high frequency + PVG area; if the blood pressure is very low (BP t <X2), then the pulse width can be superimposed and more electrode patches can be activated, etc. These rules can support most of the system's decisions at the startup stage.

[0067] Among them, the control methods of the response module include: S1. Receive the current comprehensive physiological state data (S t ={BP t , HR t , ECG t , Pos t}) from the data fusion module; S2. Judge whether the patient needs to adjust the stimulation according to the real-time data, as well as the position and quantity of the electrode patches to be activated; S3. Match the current physiological state with the preset rule table. If there is no complete match, select the nearest matching rule or use the default safety parameters to determine the activation area required (PVG area, PAG area, or both simultaneously), the activation quantity (the quantity and distribution of the activated electrode patches), and the stimulation parameters (voltage (Vt), frequency (ft), and pulse width (PWt)); S4. First, send the stimulation parameter instruction to a single power supply, and its internal components (voltage regulator, pulse generator) generate the target electrical stimulation signal; then send the switch instruction to the switch control logic circuit of the parallel switch matrix to activate the target switch and form a target current path matching the electrical stimulation signal; S5. Selectively activate the directional electrode patches to achieve segmented and directional stimulation; at the same time, continuously monitor the patient's physiological feedback (such as blood pressure BP t , heart rate HR t , etc.) through the data acquisition module; S6. If the feedback parameters do not reach the expected target, re-match the rule table and give priority to adjusting the stimulation parameters (such as voltage). If it is still ineffective, gradually increase the number of activated electrodes or adjust the position of the electrodes until the patient's physiological state returns to a safe range.

[0068] In one embodiment, the risk prediction module outputs the predicted physiological state S at the next moment through the time series model. t+1 ={BP t+1 ,HR t+1 ,ECG t+1 ,Pos t+1} to the response module, the decision logic of the response module is as follows: Rule table matching: Current Status t , matching rule table, obtaining immediate stimulation scheme (including electrode activation and stimulation parameters); If the predicted state S t+1 , beyond the safety range, prioritize matching S t+1 , corresponding rules; Processing when the rule table cannot match: If the current state S t or future state S t+1 Unable to match the rule table; The risk prediction module automatically generates stimulation parameters, and the newly generated parameters and corresponding states are recorded for dynamic updating of the rule table; Decision output: Comprehensive consideration t and S t+1 The matching results generate the final stimulation instructions, including: the activated electrode combination (PVG area, PAG area or both), the stimulation parameter voltage (V t ), frequency (f t ) and pulse width (PW t ).

[0069] Taking into account S t and S t+1 When different weights are set, the final parameter = α × current state parameter + β × predicted state parameter; α, β are weight coefficients and can be adjusted according to actual conditions.

[0070] Combined with the patient's physiological feedback data (such as real-time changes in blood pressure and heart rate), the output parameters are further adjusted to form a closed-loop optimization; each automatically generated new parameter and its corresponding status are recorded in the rule table for subsequent rapid matching to improve the system's response speed and efficiency.

[0071] Furthermore, the steps for the response module to automatically generate stimulus parameters are as follows: S1. Get the current state S through the data acquisition module t , obtain the future state S through the risk prediction module t+1 At the same time, the blood pressure target value and blood pressure deviation (ΔBP) are obtained through system preset. The blood pressure target value (BP 目标 ) is defined as: systolic blood pressure target value = baseline systolic blood pressure - 20 mmHg, diastolic blood pressure target value = baseline diastolic blood pressure - 10 mmHg; ΔBP=BP 目标 −BP t BPt is the current blood pressure monitored in real time.

[0072] Deviation classification: ΔBP>0: Blood pressure is too low and stimulation intensity needs to be increased.

[0073] ΔBP≤0: Blood pressure is normal or high, no additional stimulation is needed.

[0074] S2. Based on the size of ΔBP, the stimulation parameters are generated according to the following rules: The voltage controls the intensity of the stimulation signal and directly determines the amplitude of the stimulation current. A higher voltage can more strongly activate the target neural area, but may cause unnecessary side effects (such as overstimulation). Its adjustment logic is to calculate the current blood pressure deviation ΔBP. t =Vinitial+k v ×ΔBP adjusts the voltage to increase or decrease the stimulation intensity; where k v Indicates the voltage value that needs to be increased for each unit of blood pressure deviation; at the same time, it is also necessary to ensure that the voltage does not exceed the safety range [V min ,V max ]; The frequency determines the rhythm of the stimulation signal and affects the frequency and efficiency of nerve excitation. A higher frequency can quickly accumulate stimulation effects, but may cause nerve adaptation and reduce efficacy. The adjustment logic is to calculate the current blood pressure deviation ΔBP; through f t =f 初始 +kf f ×ΔBP adjusts the frequency and controls the stimulation rhythm; where k f Indicates the frequency that needs to be increased for each unit of blood pressure deviation; at the same time, it is necessary to ensure that the frequency does not exceed the range that the nerve can tolerate [f min ,f max ]; Pulse width defines the duration of each stimulation and affects the cumulative intensity of stimulation. Prolonging the pulse width can increase the stimulation effect of each pulse, but it may increase power consumption. Its adjustment logic is to calculate the current blood pressure deviation ΔBP; through PW t =PW 初始 +k p×ΔBP adjusts the pulse width and prolongs the single stimulation effect; where k P Indicates the pulse width that needs to be increased for each unit of blood pressure deviation; at the same time, it is necessary to ensure that the frequency does not exceed the range that the nerve can tolerate [PW min ,PW max ]; Activating more electrodes can expand the stimulation range and improve the treatment effect. According to the anatomy and target area (such as PVG and PAG), the combination of activated electrodes is reasonably selected. The adjustment logic is to calculate the current blood pressure deviation ΔBP; through the linear relationship N=N 初始 +(k n ×ΔBP) to adjust the number of electrodes and increase the stimulation coverage; where k n Indicates the number of electrodes that need to be added for each unit of blood pressure deviation; at the same time, it is also necessary to ensure that the number of activated electrodes does not exceed the maximum supported number N of the device max ; S3, the response module generates stimulation parameters according to the above adjustment rules: activation electrode combination (number N and target area (PVG, PAG or both)), stimulation parameter voltage (V t ), frequency (f t ) and pulse width (PW t ).

[0075] In one embodiment, when the rule table is not completely matched or the risk prediction module has a large prediction deviation, the reinforcement learning module optimizes the stimulation parameter adjustment strategy through a feedback mechanism; the working steps of the reinforcement learning module are as follows: S1, receive the current state S from the data fusion module t , obtain the predicted status S from the risk prediction module t+1 , initialize the current Q(S t ,A t )value; S2, select actions based on the current Q-value function, using the ϵ-greedy strategy, ϵ-probability to select random actions (exploration), 1−ϵ1-probability to select the current optimal action (utilization), actions include adjusting voltage, frequency, pulse width and activating electrodes; S3, according to the selected action A t , activate the single power supply and parallel switch matrix through the response module; monitor the feedback state S after stimulation in real time t ′ and reward R t : Reward function:

[0076] S4. Update the Q-value using the Q-Learning algorithm: Among them, α is the learning rate, which controls the step size of each update, and γ is the discount factor, which measures the importance of future rewards; S5. If Q(S t ,A t ) value is stable for a long time and performs well, update it into the rule table as a new rule.

[0077] Here is a detailed explanation of the dynamic update of the rule table; during the long-term operation of the system, the rule table is not static, but will be dynamically updated with the continuous feedback from the reinforcement learning module and the risk prediction module: S1. Reinforcement Learning Module Feedback When the system finds that the existing rule table cannot effectively handle certain extreme or unexpected situations, it will try a new stimulation strategy; if the new strategy achieves good results (for example, blood pressure recovers quickly without side effects), the new strategy will be written into the rule table. In the long run, the rule table will be richer and more personalized.

[0078] S2. Risk prediction bias correction If the risk prediction module (based on the time series model) predicts that the future blood pressure deviates too much from the actual value, the system will record the rules triggered at that time and their execution effects, so as to correct or refine the corresponding rules while updating the prediction model.

[0079] 4. Example structure of rule table For a more intuitive understanding, a simplified example rule table is given below (for reference only, the actual system may be more complex):

[0080] "Recommended activation electrodes" may correspond to a variety of electrode combinations, which are only roughly described in the table.

[0081] The circuits and controls involved in the present invention are all prior art and will not be described in detail here.

[0082] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A deep brain stimulation system for treating postural hypotension, characterized in that: include: The data acquisition module is used to collect the patient's physiological status information, including: Implantable pressure sensor, implanted in the artery, is used to monitor blood pressure data (BP) in real time. t ); ECG sensor, integrated into electronic wristwatch, used to collect ECG signals (ECG t ) and heart rate data (HR t ); The posture sensor is implanted in the subcutaneous area of ​​the pectoralis major muscle to detect the patient's posture (Pos t ); The data fusion module is implanted in the subcutaneous area of ​​the pectoralis major muscle and is used to receive the multi-source physiological data output by the data acquisition module, integrate and filter the noise, and generate the patient's comprehensive physiological state information (St={BP t ,HR t ,ECG t ,Pos t }); Deep Brain Stimulation Module, including: Directional electrodes, implanted in the patient's periventricular gray (PVG) and periaqueductal gray (PAG) regions, are used to precisely stimulate target brain areas to enhance sympathetic nerve output; The stimulation power supply assembly is implanted in the subcutaneous area of ​​the pectoralis major muscle and is used to provide voltage (V t ), frequency (f t ) and pulse width (PW t ) Adjustable electrical signals support segmented and directional stimulation to avoid interference with non-target areas; The response module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle to analyze the physiological state information S t , dynamically adjust the voltage, frequency and pulse width of the deep brain stimulation module, and optimize the adjustment strategy in real time according to the patient's physiological feedback data; The risk prediction module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle to process the comprehensive physiological status historical data {S t−n ,...,S t }, predict the patient's future physiological state S t+1 , and pass the predicted value to the response module; The reinforcement learning module is integrated with the data fusion module in the subcutaneous area of ​​the pectoralis major muscle, and is used to dynamically update the stimulation parameter adjustment strategy of the response module according to the patient's physiological feedback data through the reinforcement learning algorithm.

2. A deep brain stimulation system for treating orthostatic hypotension according to claim 1, characterized in that: The data fusion module generates the patient's real-time comprehensive physiological state information (St={BP=BP) by synchronous processing, noise filtering, feature extraction and weighted fusion of blood pressure data, heart rate data, ECG signal and body position status. t ,HR t ,ECG t ,Pos t }).

3. A deep brain stimulation system for treating orthostatic hypotension according to claim 1, characterized in that: The directional electrode is composed of a plurality of independent electrode sheets, each of which can control the on / off state individually, and each electrode sheet is precisely distributed in the periventricular gray matter (PVG) and periaqueductal gray matter (PAG) areas according to the anatomical structure and treatment goals; wherein, the stimulation power supply component independently controls the on / off state of each electrode sheet to avoid affecting non-target areas, thereby improving the pertinence and safety of the stimulation and achieving precise stimulation of specific target brain areas (such as PVG or PAG).

4. A deep brain stimulation system for treating orthostatic hypotension according to claim 3, characterized in that: The stimulation power supply assembly is composed of a single power supply and a parallel switch matrix; A single power supply is implanted in the subcutaneous area of ​​the pectoralis major muscle of the patient, which also contains a voltage regulator, a pulse generator and a power control unit; it is used to provide adjustable stimulation parameters for the directional electrodes, including voltage (V t ), frequency (f t ) and pulse width (PW t ); wherein the voltage regulator adjusts the output voltage (V t ) amplitude to meet the stimulation intensity requirements of different target brain areas; the pulse generator generates the required stimulation frequency (f t ), frequency (f t ) affects the rhythm of the stimulation signal and the discharge response of the target neuron; at the same time, the pulse generator controls the pulse width (PW) by adjusting the duration of a single pulse t ), pulse width (PW t ) size determines the coverage depth and duration of each electrical stimulation signal; A parallel switch matrix is ​​located between the single power supply and the directional electrode, and includes a plurality of independent switches and a switch control logic circuit; wherein each switch is used to control the on / off state of one or a group of electrode sheets; the switch control logic circuit receives a switch instruction from the response module to determine which switches are turned on or off, thereby configuring a specific electrode sheet activation combination for the purpose of segmented stimulation (activating a specific area) or directional stimulation (concentrating current on a specific path).

5. A deep brain stimulation system for treating orthostatic hypotension according to claim 4, characterized in that: The response module sends the generated stimulation parameter instruction to the voltage regulator and pulse generator of the single power supply, and at the same time, sends the electrode activation instruction to the switch control logic circuit to specify the electrode sheet that needs to be activated.

6. A deep brain stimulation system for treating orthostatic hypotension according to claim 1, characterized in that: The response module is provided with a rule table, which defines activation strategies and stimulation parameters under different states based on the patient's physiological state and treatment goals in combination with clinical data or individualized diagnosis results; The control method of the response module includes: S1, receiving the current comprehensive physiological state data (St={BP t ,HR t ,ECG t ,Pos t }); S2. judging whether the patient needs to adjust the stimulation according to the real-time data, and the position and number of the electrode sheets that need to be activated; S3, matching the current physiological state with the preset rule table, if not completely matched, selecting the most recently matched rule or using the default safety parameters, determining the required activation area (PVG area, PAG area, or both), activation number (number and distribution of activated electrodes), stimulation parameters (voltage (V t ), frequency (f t ) and pulse width (PW t )); S4, first sending a stimulation parameter instruction to the single power supply, and generating a target electrical stimulation signal by its internal components (voltage regulator, pulse generator); then sending a switch instruction to the switch control logic circuit of the parallel switch matrix, activating the target switch, and forming a target current path matching the electrical stimulation signal; S5, selectively activate the directional electrode sheet to achieve segmented and directional stimulation; at the same time, continuously monitor the patient's physiological feedback (such as blood pressure BP) through the data acquisition module t , heart rate HR t , electrocardiogram (ECG) t ); S6. If the feedback parameters do not reach the expected target, re-match the rule table and give priority to adjusting the stimulation parameters (such as voltage). If it is still ineffective, gradually increase the number of activated electrodes or adjust the position of the electrodes until the patient's physiological state returns to a safe range.

7. A deep brain stimulation system for treating orthostatic hypotension according to claim 1, characterized in that: The risk prediction module outputs the predicted physiological state S at the next moment through the time series model t+1 ={BP t+1 ,H Rt+1 ,ECG t+1 ,Pos t+1 } to the response module, the decision logic of the response module is as follows: Rule table matching: Current Status t Match the rule table to obtain the immediate stimulation plan (including electrode activation and stimulation parameters); If the predicted state S t+1 Out of the safety range, matching rule table S t+1 Corresponding rules; Processing when the rule table cannot be matched: If the current state S t or future state S t+1 Unable to match the rule table; The response module automatically generates stimulus parameters, and the newly generated parameters and corresponding states are recorded for dynamic updating of the rule table; Decision output: Comprehensive consideration t and S t+1 The matching results generate the final stimulation instructions, including: the activated electrode combination (PVG area, PAG area or both), the stimulation parameter voltage (V t ), frequency (f t ) and pulse width (PW t ).

8. A deep brain stimulation system for treating orthostatic hypotension according to claim 7, characterized in that: The steps of automatically generating stimulation parameters by the response module are as follows: S1. Obtain the current state S through the data fusion module t , obtain the future state S through the risk prediction module t+1 At the same time, the blood pressure target value and blood pressure deviation (ΔBP) are obtained through system preset. The blood pressure target value (BP 目标 ) is defined as: systolic blood pressure target value = baseline systolic blood pressure - 20 mmHg, diastolic blood pressure target value = baseline diastolic blood pressure - 10 mmHg; ΔBP=BP 目标 −BPt BPt is the current blood pressure monitored in real time; Deviation classification: ΔBP>0: blood pressure is too low and stimulation intensity needs to be increased; ΔBP≤0: blood pressure is normal or high, no need to increase stimulation; S2. Based on the size of ΔBP, the stimulation parameters are generated according to the following rules: The voltage controls the intensity of the stimulation signal and directly determines the amplitude of the stimulation current. A higher voltage can more strongly activate the target neural area, but may cause unnecessary side effects (such as overstimulation). Its adjustment logic is to calculate the current blood pressure deviation ΔBP. t =Vinitial+k v ×ΔBP adjusts the voltage to increase or decrease the stimulation intensity; where k v Indicates the voltage value that needs to be increased for each unit of blood pressure deviation; at the same time, it is also necessary to ensure that the voltage does not exceed the safety range [V min ,V max ]; The frequency determines the rhythm of the stimulation signal and affects the frequency and efficiency of nerve excitation. A higher frequency can quickly accumulate stimulation effects, but may cause nerve adaptation and reduce efficacy. The adjustment logic is to calculate the current blood pressure deviation ΔBP; through f t =f 初始 +kf f ×ΔBP adjusts the frequency and controls the stimulation rhythm; where k f Indicates the frequency that needs to be increased for each unit of blood pressure deviation; at the same time, it is necessary to ensure that the frequency does not exceed the range that the nerve can tolerate [f min ,f max ]; Pulse width defines the duration of each stimulation and affects the cumulative intensity of stimulation. Prolonging the pulse width can increase the stimulation effect of each pulse, but it may increase power consumption. Its adjustment logic is to calculate the current blood pressure deviation ΔBP; through PW t =PW 初始 +k p ×ΔBP adjusts the pulse width and prolongs the single stimulation effect; where k P Indicates the pulse width that needs to be increased for each unit of blood pressure deviation; at the same time, it is necessary to ensure that the frequency does not exceed the range that the nerve can tolerate [PW min ,PW max ]; Activating more electrodes can expand the stimulation range and improve the treatment effect. According to the anatomy and target area (such as PVG and PAG), the combination of activated electrodes is reasonably selected. The adjustment logic is to calculate the current blood pressure deviation ΔBP; through the linear relationship N=N 初始 +(k n ×ΔBP) to adjust the number of electrodes and increase the stimulation coverage; where k n Indicates the number of electrodes that need to be added for each unit of blood pressure deviation; at the same time, it is also necessary to ensure that the number of activated electrodes does not exceed the maximum supported number N of the device max ; S3, the response module generates stimulation parameters according to the above adjustment rules: activation electrode combination (number N and target area (PVG, PAG or both)), stimulation parameter voltage (V t ), frequency (f t ) and pulse width (PW t ).

9. A deep brain stimulation system for treating orthostatic hypotension according to claim 1, characterized in that: When the rule table is not completely matched or the risk prediction module has a large prediction deviation, the reinforcement learning module optimizes the stimulation parameter adjustment strategy through a feedback mechanism; the working steps of the reinforcement learning module are as follows: S1, receive the current state S from the data fusion module t , obtain the predicted status S from the risk prediction module t+1 , initialize the current Q(S t ,A t )value; S2, select actions based on the current Q-value function, using the ϵ-greedy strategy, ϵ-probability to select random actions (exploration), 1−ϵ1-probability to select the current optimal action (utilization), actions include adjusting voltage, frequency, pulse width and activating electrodes; S3, according to the selected action A t , activate the single power supply and parallel switch matrix through the response module; monitor the feedback state S after stimulation in real time t ′ and reward R t : Reward function: ; S4. Update the Q-value using the Q-Learning algorithm: Among them, α is the learning rate, which controls the step size of each update, and γ is the discount factor, which measures the importance of future rewards; S5. If Q(S t ,A t ) value is stable for a long time and performs well, update it into the rule table as a new rule.

Images