Electronic Acquisition, Storage and Management System for Pacemaker Program Control Data

Abstract

The present invention relates to the technical field of pacemaker control, and discloses an electronic acquisition, storage and management system for pacemaker programming data, including a data acquisition module, a data preprocessing module, a data analysis module and a feedback mechanism module: The data acquisition module is used to acquire pacemaker characteristic data; The data preprocessing module is used to preprocess the first characteristic sequence and perform feature extraction to obtain the second characteristic sequence; The data analysis module is used to perform real-time optimization of the pacemaker characteristic data at the local end and long-term optimization in the cloud; The feedback mechanism module is used to coordinate the local end optimization and the cloud optimization to form a closed-loop control system. Through the collaborative closed-loop mechanism of real-time optimization at the local end and long-term optimization in the cloud, the present invention combines fuzzy control, gated recurrent unit and reinforcement learning to realize personalized pacing parameter regulation and improve the stability, adaptability and intelligent level of the pacemaker.

Description

Technical Field

[0001] The present invention belongs to the technical field of pacemaker control, and particularly relates to an electronic acquisition, storage, and management system for pacemaker programming data. Background Art

[0002] A pacemaker is an important implantable medical device for the heart, widely used in the treatment of heart diseases such as arrhythmia. Traditional pacemaker control parameters (such as pacing threshold, pacing frequency, sensing sensitivity, etc.) are usually manually adjusted by doctors during follow-up, making it difficult to dynamically adapt to the patient's physiological state. In addition, existing pacemaker data acquisition and management systems mainly rely on manual reading and storage, lacking efficient data analysis and intelligent optimization means, and are difficult to meet the needs of personalized treatment. Currently, some studies have proposed pacemaker remote monitoring technology, enabling pacemaker data to be regularly uploaded to the cloud for analysis. However, the existing technology has the following deficiencies: 1. Poor real-time performance, with a delay in cloud data processing and inability to quickly respond to changes in the patient's physiological state; 2. Lack of adaptive optimization, and the adjustment of pacing parameters still relies on manual intervention by doctors, unable to automatically make intelligent adjustments according to the patient's physiological data; 3. Low data utilization rate. Although some systems can collect and store the patient's physiological data, they lack effective analysis methods and cannot fully exploit the data value. Summary of the Invention

[0003] The present invention provides an electronic acquisition, storage, and management system for pacemaker programming data, which solves the technical problems of deficiencies in real-time performance, adaptive optimization, and data utilization in related technologies, and is difficult to achieve rapid response, intelligent adjustment, and efficient data analysis.

[0004] The present invention provides an electronic acquisition, storage, and management system for pacemaker programming data, including a data acquisition module, a data preprocessing module, a data analysis module, and a feedback mechanism module:

[0005] The data acquisition module is used to acquire pacemaker characteristic data, where the pacemaker characteristic data includes: pacing threshold, pacing frequency, sensing sensitivity, pacing mode, output energy, heart rate, heart rate variability, QT interval, and activity level;

[0006] Among them, the pacemaker characteristic data is acquired at a first preset time interval within a first preset time period to obtain a first characteristic sequence;

[0007] The j-th sequence unit of the first characteristic sequence represents the pacemaker characteristic data acquired at the j-th time point;

[0008] The data preprocessing module is used to preprocess the first characteristic sequence and perform feature extraction to obtain a second characteristic sequence;

[0009] The data analysis module is used to perform real-time optimization of the pacemaker characteristic data at the local end and long-term optimization in the cloud. Here, the local end represents the pacemaker device, and the cloud represents the remote server. At the local end, the physiological signals of the pacemaker are analyzed using a fuzzy control algorithm, and the first control parameter is adjusted in real time based on the first preset rule; in the cloud, a physiological state prediction model is constructed based on a gated recurrent unit to analyze the second feature sequence, and the first control parameter of the pacemaker is optimized in combination with a reinforcement learning algorithm; among them, the physiological signals include: heart rate, heart rate variability, QT interval, and activity level, and the first control parameter includes: pacing threshold, pacing frequency, and sensing sensitivity;

[0010] The feedback mechanism module is used to coordinate the local end optimization and the cloud optimization to form a closed-loop control system.

[0011] Furthermore, the pacing modes include: single-chamber pacing, dual-chamber pacing, and adaptive pacing, which are represented by numerical coding. The heart rate variability is represented by the standard deviation of the heart rate, and the activity level is represented by the numerical coding of the acceleration.

[0012] Furthermore, the preprocessing of the first feature sequence is performed, and the specific steps include:

[0013] S201, filling the missing values using the mean value in the first feature sequence;

[0014] S202, denoising the first feature sequence using a band-pass filtering method;

[0015] S203, performing statistical anomaly detection on the first feature sequence based on the standard deviation and 3 rules, screening out the outliers, and performing replacement processing using the linear interpolation method.

[0016] Furthermore, the specific steps of the feature extraction include:

[0017] S301, extracting the pacing frequency-energy comprehensive index feature according to the first feature sequence. The calculation formula of the pacing frequency-energy comprehensive index feature is: , where, represents the pacing frequency-energy comprehensive index feature, represents the current pacing frequency, represents the current output energy of the pacemaker, represents the current pacing threshold, represents the current sensing sensitivity, represents a minimum value between 0 and 1, represents the number of acquisition time points within the first preset time period, i represents the index of the time point, represents the pacing frequency at the i-th time point, represents the average pacing frequency at T time points within the first preset time period;

[0018] S302. Extract the heart beat stability coefficient feature according to the first feature sequence. The calculation formula of the heart beat stability coefficient feature is: , where represents the heart beat stability coefficient feature, represents the first preset time interval, represents the heart rate at the i-th time point, represents the average heart rate of T time points within the first preset time period, represents the pacing mode, represents the reference sensing sensitivity, S represents the sensing sensitivity, represents the QT change weight coefficient, represents the QT interval, represents the standard QT value;

[0019] S303. Extract the physiological recovery index feature according to the first feature sequence. The calculation formula of the physiological recovery index feature is: , where represents the physiological recovery index feature, represents the heart rate at the (i - 1)-th time point, and respectively represent the QT intervals at the (i - 1)-th time point and the i-th time point, represents the activity level at the i-th time point, represents the average activity level of T time points within the first preset time period;

[0020] S304. Concatenate the pacing frequency - energy comprehensive index feature, the heart beat stability coefficient feature, and the physiological recovery index feature after the sequence unit of the first feature sequence, and perform normalization to obtain the second feature sequence.

[0021] Furthermore, the specific steps of the fuzzy control algorithm include:

[0022] S401. Input fuzzyfication step, perform multi-layer fuzzy mapping on the physiological signal;

[0023] S402. Fuzzy inference step, analyze the adjustment trend of the first control parameter under the fuzzy logic framework according to the first preset rule, where the first preset rule is used to define the fuzzy adjustment boundaries corresponding to different physiological states;

[0024] S403. Defuzzyfication step, after the adjustment trend is clear, convert the control quantity in the fuzzy space into an executable numerical value and apply it to the first control parameter of the pacemaker.

[0025] Furthermore, the specific steps of S402 include:

[0026] S501. Calculate the membership degree of the physiological signal based on the triangular membership function. The membership degree represents the belonging degree of the input variable at this division level.

[0027] S502. Calculate the activation degree of the rule triggered by the input variable according to the membership degree of the input variable.

[0028] S503. Determine the adjustment trend of the first control parameter according to the activation degree of the rule.

[0029] Furthermore, the physiological state prediction model is constructed with a gated recurrent unit as the core. The composition structure of the physiological state prediction model includes: an input layer, a feature extraction layer, and an output layer. Among them, the input layer is used to input the second feature sequence. The feature extraction layer uses the gated recurrent unit to process the second feature sequence and extract cross-time-step correlation features. The output layer is used to perform dimensionality reduction mapping on the extracted cross-time-step correlation features and finally generate a physiological health score.

[0030] Furthermore, a reinforcement learning model is constructed with the physiological health score as the core. The state variables of the reinforcement learning model include the parameters in the pacemaker feature data and the physiological health score. The action variables include the adjustment values of the pacing threshold, pacing frequency, and sensing sensitivity. The reward function is composed of the change value of the physiological health score and the adjustment cost of the first control parameter. During the training process, deep deterministic policy gradient is adopted to optimize the first control parameter under different physiological states. Among them, the calculation formula of the reward function is:

[0031] ;

[0032] Where represents the value of the reward function, represents the physiological health score at the current time step, represents the physiological health score at the previous time step, represents the first weight coefficient, , and respectively represent the second weight coefficient, the third weight coefficient, and the fourth weight coefficient, represents the adjustment value of the pacing threshold, represents the adjustment value of the pacing frequency, represents the adjustment value of the sensing sensitivity.

[0033] Furthermore, the feedback mechanism module includes: local optimization feedback, cloud optimization feedback, and a closed-loop feedback mechanism. The local optimization feedback is adjusted in real time through a fuzzy control algorithm and a first preset rule, and the adjustment results are regularly uploaded to the cloud for long-term learning by the cloud. The cloud optimization feedback is used to output a physiological health score through a physiological state prediction model, and uses reinforcement learning to optimize the first control parameter to obtain an optimization strategy, and pushes it to the local side. The closed-loop feedback mechanism specifically includes: local real-time optimization, local feedback upload, cloud long-term learning, cloud push of optimization strategy, and local update of optimization strategy.

[0034] The beneficial effects of the present invention are as follows: The present invention constructs a collaborative closed-loop mechanism for local optimization and cloud optimization; uses a fuzzy control algorithm at the local side to realize real-time optimization of pacing threshold, pacing frequency, and sensing sensitivity; constructs a physiological state prediction model based on a gated recurrent unit at the cloud, combines reinforcement learning for long-term optimization, and regularly pushes optimization strategies to the local side; this hierarchical optimization architecture can not only meet short-term response requirements but also optimize personalized regulation based on long-term trends, improving the stability and adaptability of the pacemaker. Brief Description of the Drawings

[0035] Figure 1 It is a schematic diagram of the modules of the pacemaker programming data electronic acquisition, storage, and management system of the present invention. Detailed Embodiments

[0036] Now, the subject matter described herein will be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described relative to some examples can also be combined in other examples.

[0037] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should have the ordinary meaning understood by those of ordinary skill in the field to which the present invention pertains. In one or more embodiments of the present invention, the terms "first", "second" and similar words do not denote any order, quantity or importance, but are only used to distinguish different components. Words such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Upper", "lower", "left", "right", etc. are only used to represent relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0038] As Figure 1 shown, the electronic acquisition, storage and management system for pacemaker programming data includes:

[0039] A data acquisition module 101, configured to acquire pacemaker characteristic data, where the pacemaker characteristic data includes: pacing threshold, pacing frequency, sensing sensitivity, pacing mode, output energy, heart rate, heart rate variability, QT interval, and activity level;

[0040] Specifically, the pacing threshold represents the minimum voltage required for the pacemaker to stimulate cardiac contraction and is an important parameter for judging the sensitivity of the heart to electrical stimulation; the pacing frequency represents the number of stimulation pulses emitted by the pacemaker per minute and directly reflects the intensity of the device's intervention in the cardiac rhythm; the sensing sensitivity represents the minimum threshold for the pacemaker to detect intrinsic cardiac electrical signals and determines whether the device can accurately capture the natural cardiac rhythm; the pacing mode represents the pacing and sensing methods of the pacemaker for the atrium and ventricle, such as single-chamber, dual-chamber, or rate-responsive modes, reflecting the device's regulation strategy; the output energy represents the energy transmitted by the pacemaker each time it emits a pulse, usually measured in millijoules (mJ), which affects battery life and pacing effect; the heart rate represents the number of heartbeats per minute and is a basic physiological index for evaluating the working state of the heart; the heart rate variability reflects the regulatory ability of the autonomic nervous system on the heart through the standard deviation of adjacent R-R intervals; the QT interval represents the time interval from the start of the Q wave to the end of the T wave in the electrocardiogram, reflecting the cardiac repolarization process, and abnormalities may indicate the risk of arrhythmia; the activity level is measured by an acceleration sensor and reflects the patient's exercise state and physical activity intensity, which is of great significance for rate-responsive pacing;

[0041] Among them, the pacemaker characteristic data is acquired by the pacemaker at a first preset time interval within a first preset time period to obtain a first characteristic sequence;

[0042] The j-th sequence element of the first feature sequence represents the pacemaker feature data collected at the j-th time point;

[0043] The data preprocessing module 102 is used to preprocess the first feature sequence and perform feature extraction to obtain a second feature sequence;

[0044] The data analysis module 103 is used to perform real-time optimization of the pacemaker feature data at the local end and long-term optimization in the cloud. Here, the local end represents the pacemaker device, and the cloud represents the remote server. At the local end, the physiological signals of the pacemaker are analyzed using a fuzzy control algorithm, and the first control parameter is adjusted in real time based on the first preset rule; in the cloud, a physiological state prediction model is constructed based on a gated recurrent unit to analyze the second feature sequence, and the first control parameter of the pacemaker is optimized in combination with a reinforcement learning algorithm; where the physiological signals include: heart rate, heart rate variability, QT interval, and activity level, and the first control parameters include: pacing threshold, pacing frequency, and sensing sensitivity;

[0045] The feedback mechanism module 104 is used to coordinate local end optimization and cloud optimization to form a closed-loop control system.

[0046] In an embodiment of the present invention, the pacing modes include: single-chamber pacing, dual-chamber pacing, and adaptive pacing, and are represented by numerical coding. For example, single-chamber pacing is coded as 1, dual-chamber pacing is coded as 2, and adaptive pacing is coded as 3. Heart rate variability is represented by the standard deviation of the heart rate, and the activity level is represented by the numerical coding of the acceleration;

[0047] In an embodiment of the present invention, to ensure the consistency of the pacemaker feature data and make it applicable to the physiological state prediction model, all data needs to be unified in units. Specifically, the unit of the pacing threshold is millivolt, the unit of the pacing frequency is beats per minute, the unit of the sensing sensitivity is millivolt, the unit of the output energy is millijoule, the unit of the heart rate is beats per minute, the unit of the heart rate variability is millisecond, the unit of the QT interval is millisecond, and the unit of the activity level is meters per second squared.

[0048] In an embodiment of the present invention, the preprocessing of the first feature sequence specifically includes the following steps:

[0049] S201, filling the missing values using the mean value in the first feature sequence. Specifically, for each feature, calculate the average value of all collected values within the first preset time period, and then replace all missing values with this average value to ensure sequence continuity and data integrity;

[0050] S202, denoising the first feature sequence using a band-pass filtering method, which can effectively remove noises caused by electromyographic noise, electromagnetic interference, etc., improving the clarity of the signal and the accuracy of subsequent feature extraction;

[0051] S203, perform statistical anomaly detection on the first feature sequence based on the standard deviation and 3 rules, screen out the outliers, and use linear interpolation method for replacement processing. Specifically, calculate the mean and standard deviation of the pacemaker feature data within the first preset time period to obtain 3 range, and consider the data outside the range as outliers. Use the linear interpolation method to replace the outliers to ensure the smoothness and continuity of the first feature sequence.

[0052] In an embodiment of the present invention, the specific steps of the feature extraction include:

[0053] S301, extract the pacing frequency-energy comprehensive index feature according to the first feature sequence. The calculation formula of the pacing frequency-energy comprehensive index feature is: , where represents the pacing frequency-energy comprehensive index feature. If the FESI is too high, it indicates that the pacing frequency is high and the energy consumption is large at the same time, which may lead to energy waste and the need to optimize the sensing sensitivity. If the FESI is too low, it means that the pacing frequency of the pacemaker is low or the energy output is insufficient, which may lead to insufficient regulation of the patient's cardiac rhythm; represents the current pacing frequency, represents the current output energy of the pacemaker, represents the current pacing threshold, represents the current sensing sensitivity, represents a very small value between 0 and 1. Preferably, is set to , represents the number of acquisition time points within the first preset time period, and i represents the index of the time point. represents the pacing frequency at the i-th time point, represents the average pacing frequency of T time points within the first preset time period;

[0054] S302, extract the heart beat stability coefficient feature according to the first feature sequence. The calculation formula of the heart beat stability coefficient feature is: , where represents the heart beat stability coefficient feature, which is used to measure the stability of the heart rate. If the HSF is too high, it indicates that the heart rate fluctuates greatly and the regulation effect of the pacemaker is not stable enough. If the HSF is too low, it means that the heart rate is relatively stable and the control effect of the pacemaker is better; represents the first preset time interval, represents the heart rate at the i-th time point, represents the average heart rate of T time points within the first preset time period, represents the pacing mode, represents the reference sensing sensitivity, which is the preset initial value of the pacemaker sensing sensitivity, S represents the sensing sensitivity, represents the QT change weight coefficient, represents the QT interval, represents the standard QT value, that is, the QT interval value under normal physiological conditions;

[0055] S303, extract the physiological recovery index feature according to the first feature sequence, and the calculation formula of the physiological recovery index feature is: , where, represents the physiological recovery index feature, which is used to measure the patient's physiological adaptability. If the PRI is too high, it means that the patient's heart has a strong adaptability to physiological activities. If the PRI is too low, it means that the patient's heart adaptability is poor; represents the heart rate at the (i - 1)-th time point, and represent the QT intervals at the (i - 1)-th time point and the i-th time point respectively, represents the activity level at the i-th time point, represents the average activity level of T time points within the first preset time period;

[0056] S304, splice the pacing frequency - energy comprehensive index feature, the heartbeat stability coefficient feature and the physiological recovery index feature after the sequence unit of the first feature sequence, and perform normalization to obtain the second feature sequence.

[0057] In an embodiment of the present invention, the local end uses a fuzzy control algorithm to analyze the pacing threshold, pacing frequency and sensing sensitivity of the pacemaker, and realizes real-time adjustment based on the first preset rule to optimize the physiological adaptability of the pacemaker and improve the cardiac function stability of the patient.

[0058] In an embodiment of the present invention, the specific steps of the fuzzy control algorithm include:

[0059] S401, the input fuzzyfication step, perform multi-layer fuzzy mapping on the physiological signals. Specifically, divide the heart rate into three levels: low, normal and high, divide the heart rate variability into three levels: low, normal and high, divide the QT interval into three levels: short, normal and long, and divide the activity level into four levels: resting, slightly active, exercising and vigorously exercising. Fuzzy it to the corresponding level according to the specific value and actual situation, which can enhance the robustness and reduce the computational complexity;

[0060] S402, the fuzzy inference step, analyze the adjustment trend of the first control parameter under the fuzzy logic framework according to the first preset rule, where the first preset rule is used to define the fuzzy adjustment boundaries corresponding to different physiological states;

[0061] Among them, the first preset rules include but are not limited to:

[0062] When it is detected that the heart rate is higher than the first preset threshold and the heart rate variability decreases, the pacing frequency is reduced. Preferably, the first preset threshold is 130 bpm;

[0063] When it is detected that the heart rate is lower than the second preset threshold and the QT interval prolongs, the pacing frequency is increased. Preferably, the second preset threshold is 50 bpm;

[0064] When it is detected that the QT interval prolongs beyond the third preset threshold, the pacing threshold is reduced to reduce overstimulation. Preferably, the third preset threshold is 480 ms;

[0065] When it is detected that the heart rate is higher than the fourth preset threshold and the activity level is higher than the fifth preset threshold, the pacing frequency is increased to adapt to the physiological needs. Preferably, the fourth preset threshold is 100 bpm;

[0066] S403, the defuzzification step. After the adjustment trend is clear, the control quantity in the fuzzy space is converted into an executable numerical value and acts on the first control parameter of the pacemaker to ensure the accuracy of execution and optimize the control output.

[0067] In an embodiment of the present invention, the specific steps of S402 include:

[0068] S501, calculating the membership degree of the physiological signal based on the triangular membership function. The membership degree represents the belonging degree of the input variable at this division level. For example, when the heart rate is 100 bpm, the classification standard of the heart rate is: low: , normal: , high: , where R represents the heart rate. For , in the "high" level, when the heart rate is 90, the membership degree is 0; when the heart rate is 120, the membership degree is 1; when the heart rate is 150 and above, the membership degree is 0; the membership degree of the heart rate of 100 is: ; This step can eliminate the uncertainty of the data and make the pacemaker applicable to the individual differences of different patients;

[0069] S502, calculating the activation degree of the rule triggered by the input variable according to the membership degree of the input variable to ensure decision-making flexibility and prevent misjudgment of the fixed threshold. Specifically, for the input variable involved in the triggered rule, calculate its membership degree and take the minimum value in the membership degrees as the activation degree of the triggered rule. For example, the current heart rate input variable is 100 bpm, and the current activity level input variable is 2 , the rules triggered by the input variables are as follows: when it is detected that the heart rate is higher than the fourth preset threshold and the activity level is higher than the fifth preset threshold, increase the pacing frequency; the membership degree of high heart rate is calculated to be 0.33, and the membership degree of strenuous exercise is 0.67. By taking the minimum value operation, the activation degree of the triggered rule is obtained as 0.33, indicating that the applicable degree of this rule is 33%;

[0070] S503. Determine the adjustment trend according to the activation degree of the rule, and be compatible with multi-rule conflicts. Specifically, synthesize the results of the activated rules to obtain a fuzzy adjustment trend. For example: the pacing frequency should be increased, and the sensing sensitivity should be increased. For multiple identical adjustment trends, select the triggered rule with a lower activation degree.

[0071] In an embodiment of the present invention, in S403, convert the adjustment trend of the first control parameter into an exact value, which specifically includes: multiply the activation degree of the rule triggered by the input variable by the adjustment value corresponding to the preset adjustment trend to obtain the final adjustment value. For example, for the pacing frequency, the adjustment trend of the triggered rule is "increase the pacing frequency", the activation degree is 0.7, and the preset adjustment value of "increase the pacing frequency" is +3 bpm. Then the final adjustment value is: 。

[0072] In an embodiment of the present invention, the local optimization is based on real-time pacemaker characteristic data, combined with a fuzzy control algorithm, to achieve fast adaptive adjustment of the first control parameter, ensuring the stability of the heart state and the real-time response of the pacemaker. Through the steps of input fuzzification, fuzzy inference, and defuzzification, the adjustment of the first control parameter has high sensitivity and low computational burden, and is applicable to low-power devices. The local optimization avoids the delay problem of cloud data transmission, improves the instant adaptation ability of the patient's physiological state, and forms a closed loop with the long-term optimization of the cloud, improving the robustness and personalized adaptation ability of the overall system.

[0073] In an embodiment of the present invention, the physiological state prediction model is constructed with a gated recurrent unit as the core. The composition structure of the physiological state prediction model includes: an input layer, a feature extraction layer, and an output layer; among them, the input layer is used to input the second feature sequence; the feature extraction layer uses the gated recurrent unit to process the second feature sequence and extract cross-time step correlation features; the output layer is used to perform dimensionality reduction mapping on the extracted cross-time step correlation features and finally generate a physiological health score.

[0074] In an embodiment of the present invention, the physiological state prediction model uses the mean square error function as the loss function. The mean square error function is used to calculate the average of the squared errors between the predicted physiological health score and the true physiological health score. The smaller the function result, the smaller the loss value, indicating that the model prediction is more accurate. By minimizing the function result, the predicted physiological health score approaches the true physiological health score.

[0075] In one embodiment of the present invention, a reinforcement learning model is constructed with the physiological health score as the core. The state variables of the reinforcement learning model include the parameters in the pacemaker feature data and the physiological health score. The action variables include the adjustment values of the pacing threshold, pacing frequency, and sensing sensitivity. The reward function is composed of the change value of the physiological health score and the first control parameter adjustment cost. During the training process, deep deterministic policy gradient is used to optimize the first control parameter under different physiological states to maximize the physiological health score. According to the trained reinforcement learning model, the first control parameter is adjusted in real time at the pacemaker device end to optimize the patient's cardiac health status. Among them, the calculation formula of the reward function is:

[0076] ;

[0077] Among them, represents the value of the reward function, represents the physiological health score at the current time step, represents the physiological health score at the previous time step, represents the first weight coefficient, , and respectively represent the second weight coefficient, the third weight coefficient, and the fourth weight coefficient, represents the adjustment value of the pacing threshold, represents the adjustment value of the pacing frequency, represents the adjustment value of the sensing sensitivity.

[0078] In one embodiment of the present invention, among the state variables of the reinforcement learning, the physiological health score of the previous time step is used as the current state variable, and the action variable adopts a dynamic step size adjustment mechanism. The step size adjustment is based on the change trend of the current physiological health score to avoid the inadaptability caused by too fast parameter adjustment.

[0079] In one embodiment of the present invention, compared with the local end, the cloud has stronger data storage and computing capabilities. It can mine long-term physiological trends through deep learning to improve the global optimality of parameter adjustment. In addition, the cloud uses deep deterministic policy gradient to optimize the reinforcement learning strategy, provides a dynamic and adaptive personalized pacing scheme, and periodically pushes the optimized strategy to the local end to improve the long-term adaptability and stability of the pacemaker.

[0080] In an embodiment of the present invention, the feedback mechanism module includes: local optimization feedback, cloud optimization feedback, and a closed-loop feedback mechanism. The local optimization feedback is adjusted in real time through a fuzzy control algorithm and a first preset rule, and the adjustment results are regularly uploaded to the cloud for long-term learning by the cloud; the cloud optimization feedback is used to output a physiological health score through a physiological state prediction model, and uses reinforcement learning to optimize the first control parameter to obtain an optimization strategy, and pushes it to the local end; the closed-loop feedback mechanism specifically includes: local real-time optimization, local feedback upload, cloud long-term learning, cloud push of the optimization strategy, and local update of the optimization strategy. The feedback mechanism reduces manual intervention and improves the intelligence level of the pacemaker, enabling patients to obtain accurate and efficient cardiac regulation solutions in different physiological states.

[0081] The above describes the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation manners. The above specific implementation manners are only illustrative and not restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this embodiment.

Claims

1. An electronic acquisition, storage and management system for pacemaker programming data, characterized in that, It includes a data acquisition module, a data preprocessing module, a data analysis module, and a feedback mechanism module: The data acquisition module is used to acquire the pacemaker characteristic data. Among them, the pacemaker characteristic data includes: pacing threshold, pacing frequency, sensing sensitivity, pacing mode, output energy, heart rate, heart rate variability, QT interval, and activity level; Among them, the pacemaker characteristic data is acquired at a first preset time interval within a first preset time period to obtain a first characteristic sequence; The j-th sequence unit of the first characteristic sequence represents the pacemaker characteristic data acquired at the j-th time point; The data preprocessing module is used to preprocess the first characteristic sequence and perform feature extraction to obtain a second characteristic sequence, including extracting the pacing frequency-energy comprehensive index feature, the cardiac beat stability coefficient feature, and the physiological recovery index feature; The data analysis module is used to perform real-time optimization of the pacemaker characteristic data at the local end and long-term optimization in the cloud. Among them, the local end represents the pacemaker device, and the cloud represents the remote server. At the local end, the physiological signals of the pacemaker are analyzed using a fuzzy control algorithm, and the first control parameter is adjusted in real time based on the first preset rule; in the cloud, a physiological state prediction model is constructed based on a gated recurrent unit to analyze the second characteristic sequence, and the first control parameter of the pacemaker is optimized in combination with a reinforcement learning algorithm; among them, the physiological signals include: heart rate, heart rate variability, QT interval, and activity level, and the first control parameters include: pacing threshold, pacing frequency, sensing sensitivity; Among them, the physiological state prediction model is constructed with a gated recurrent unit as the core, and the composition structure of the physiological state prediction model includes: an input layer, a feature extraction layer, and an output layer; The input layer is used to input the second characteristic sequence; the feature extraction layer uses a gated recurrent unit to process the second characteristic sequence and extract cross-time-step correlation features; the output layer is used to perform dimensionality reduction mapping on the extracted cross-time-step correlation features and finally generate a physiological health score; The feedback mechanism module is used to coordinate the local optimization and the cloud optimization to form a closed-loop control system.

2. The pacemaker programming data electronic acquisition, storage and management system according to claim 1, characterized in that, The pacing mode includes: single-chamber pacing, dual-chamber pacing, adaptive pacing, and is represented by numerical encoding. The heart rate variability is represented by the standard deviation of the heart rate, and the activity level is represented by the numerical encoding of the acceleration.

3. The pacemaker programming data electronic acquisition, storage and management system according to claim 1, characterized in that, The specific steps for preprocessing the first characteristic sequence include: S201, filling the missing values with the mean value in the first characteristic sequence; S202, denoising the first characteristic sequence using a band-pass filtering method; S203, based on the standard deviation and 3 rules perform statistical anomaly detection on the first feature sequence, screen out outliers, and use linear interpolation method for replacement processing.

4. The pacemaker programming data electronic acquisition, storage and management system according to claim 1, characterized in that, The specific steps of the feature extraction include: S301. Extract the pacing frequency-energy comprehensive index feature according to the first feature sequence. The calculation formula of the pacing frequency-energy comprehensive index feature is: , where represents the pacing frequency-energy comprehensive index feature, represents the current pacing frequency, represents the current output energy of the pacemaker, represents the current pacing threshold, represents the current sensing sensitivity, represents a minimum value between 0 and 1, represents the number of acquisition time points within the first preset time period, and i represents the index of the time point, represents the pacing frequency at the i-th time point, represents the average pacing frequency of T time points within the first preset time period; S302. Extract the heart beat stability coefficient feature according to the first feature sequence. The calculation formula for the heart beat stability coefficient feature is as follows: , where represents the heart beat stability coefficient feature, represents the first preset time interval, represents the heart rate at the i-th time point, represents the average heart rate of T time points within the first preset time period, represents the pacing mode, represents the reference sensing sensitivity, S represents the sensing sensitivity, represents the QT change weight coefficient, represents the QT interval, represents the standard QT value; S303. Extract the physiological recovery index feature according to the first feature sequence. The calculation formula of the physiological recovery index feature is as follows: , where represents the physiological recovery index feature, represents the heart rate at the (i - 1)-th time point, and represent the QT intervals at the (i - 1)-th time point and the i-th time point respectively, represents the activity level at the i-th time point, represents the average value of the activity levels at T time points within the first preset time period; S304, splicing the pacing frequency-energy comprehensive index feature, the cardiac beat stability coefficient feature, and the physiological recovery index feature after the sequence unit of the first characteristic sequence, and performing normalization to obtain the second characteristic sequence.

5. The pacemaker programming data electronic acquisition, storage and management system according to claim 1, characterized in that, The specific steps of the fuzzy control algorithm include: S401, the input fuzzyfication step, performing multi-layer fuzzy mapping on the physiological signals; S402, the fuzzy inference step, analyzing the adjustment trend of the first control parameter under the fuzzy logic framework according to the first preset rule, where the first preset rule is used to define the fuzzy adjustment boundaries corresponding to different physiological states; S403, Defuzzification step. After the adjustment trend is clear, convert the control quantity in the fuzzy space into an executable value and apply it to the first control parameter of the pacemaker.

6. The pacemaker programming data electronic acquisition, storage and management system according to claim 5, wherein The specific steps of S402 include: S501, Calculate the membership degree of the physiological signal based on the triangular membership function. The membership degree represents the degree of belonging of the input variable at this division level; S502, Calculate the activation degree of the rule triggered by the input variable according to the membership degree of the input variable; S503, Determine the adjustment trend of the first control parameter according to the activation degree of the rule.

7. The pacemaker programming data electronic acquisition, storage and management system according to claim 1, characterized in that Construct a reinforcement learning model with the physiological health score as the core. The state variables of the reinforcement learning model include the parameters in the pacemaker feature data and the physiological health score, the action variables include the adjustment values of the pacing threshold, pacing frequency, and sensing sensitivity, and the reward function is composed of the change value of the physiological health score and the adjustment cost of the first control parameter; during the training process, use deep deterministic policy gradient to optimize the first control parameter under different physiological states; among them, the calculation formula of the reward function is: ； Among them, represents the value of the reward function, represents the physiological health score at the current time step, represents the physiological health score at the previous time step, represents the first weight coefficient, 、 and represent the second weight coefficient, the third weight coefficient, and the fourth weight coefficient respectively, represents the adjustment value of the pacing threshold, represents the adjustment value of the pacing frequency, represents the adjustment value of the sensing sensitivity.

8. The pacemaker programmed data electronic acquisition, storage and management system according to claim 7, characterized in that, The feedback mechanism module includes: local optimization feedback, cloud optimization feedback, and closed-loop feedback mechanism. The local optimization feedback is adjusted in real time through the fuzzy control algorithm and the first preset rule, and the adjustment results are regularly uploaded to the cloud for long-term learning by the cloud; the cloud optimization feedback is used to output the physiological health score through the physiological state prediction model, and use reinforcement learning to optimize the first control parameter to obtain the optimization strategy and push it to the local end; the closed-loop feedback mechanism specifically includes: local real-time optimization, local upload feedback, cloud long-term learning, cloud push optimization strategy, and local update optimization strategy.

Images