Anesthesia depth monitoring system and method based on skin electric signal analysis

Abstract

The invention discloses an anesthesia depth monitoring system and method based on skin electric signal analysis, the system comprises a dynamic skin electric analysis subsystem, a dynamic signal enhancement and adaptive noise reduction processing subsystem, a polymorphic fusion depth network subsystem and a reinforcement learning driving adaptive regulator, and the method is realized based on the system. According to the invention, an accurate anesthesia depth monitoring scheme is provided through multi-modal analysis of skin electric signals; based on a self-adaptive drug adjustment algorithm and reinforcement learning, real-time adjustment of the anesthesia depth is realized, and too deep or too shallow anesthesia is prevented; individual anesthesia management is performed in combination with individual differences of patients, so that the anesthesia safety is improved; compared with a traditional electroencephalogram method, the GSR signal collection mode is noninvasive and more comfortable, and long-term monitoring of a patient is facilitated; through innovative signal processing, deep learning and adaptive adjustment mechanisms, a more intelligent, accurate and personalized anesthesia depth monitoring solution can be provided.

Description

Technical Field

[0001] The present invention relates to the field of anesthesia depth monitoring, and in particular to an anesthesia depth monitoring system and method based on skin electrical signal analysis. Background Art

[0002] Anesthesia depth refers to the degree to which a patient's consciousness, pain perception, and physiological functions are suppressed under the action of anesthetic drugs. Traditional anesthesia depth monitoring methods mainly rely on clinical observations and physiological indicators, such as heart rate and blood pressure, but lack direct assessment of brain function status, resulting in low accuracy in anesthesia depth monitoring and certain safety risks.

[0003] Although existing technologies use technologies such as electroencephalogram (EEG) or bispectral index (BIS) to monitor the depth of anesthesia, the technology is expensive, the operation is complicated, and it causes great discomfort to the patients.

[0004] Therefore, there is an urgent need for a safer, more accurate and non-invasive anesthesia depth monitoring method and system. Summary of the invention

[0005] In order to solve the above-described technical problems, the present invention provides an anesthesia depth monitoring system and method based on skin electrical signal analysis.

[0006] Furthermore, an anesthesia depth monitoring system based on skin electrical signal analysis includes a neural dynamic skin electrical analysis subsystem, a dynamic signal enhancement and adaptive noise reduction processing subsystem, a polymorphic fusion deep network subsystem, and a reinforcement learning driven adaptive controller; The neural dynamic skin electrodermal analysis subsystem uses a multi-dimensional dynamic analysis method based on skin electrodermal response, combining real-time changes and neural complexity to monitor the depth of anesthesia; The dynamic signal enhancement and adaptive noise reduction processing subsystem removes noise and artifacts in the skin electrical signal and enhances the effective information of the signal; The polymorphic fusion deep network subsystem combines convolutional neural networks and long short-term memory networks to extract high-precision features from multimodal signals and accurately assess the depth of anesthesia; The reinforcement learning driven adaptive controller ensures that the anesthesia depth is maintained at an optimal state by dynamically adjusting the anesthetic drug infusion amount.

[0007] Further, an anesthesia depth monitoring system based on skin electrical signal analysis, the neurodynamic skin electrical analysis subsystem includes a skin electrical response dynamic sensitivity analysis module, a skin electrical signal complexity analysis module, and a neurodynamic index comprehensive analysis module; The dynamic sensitivity analysis module of the skin electrical response quantifies the reaction speed of the sympathetic nervous system by calculating the instantaneous rate of change of skin conductance, and the formula is: ; in, It stands for nerve sensitivity index and is used to capture changes in the activity of the sympathetic nervous system during anesthesia. Represents skin conductance at time interval Changes within; The complexity analysis module of the skin electrical signal calculates the entropy value of the signal through a multi-scale method to reveal the nonlinear dynamics in the skin electrical signal and the complex pattern of anesthesia depth. The formula is: ; in, represents the neural complexity entropy, represents the probability distribution of the skin electrical signal state, i represents the different states of the signal, and n represents the total number of states into which the signal is divided; The neurodynamic index comprehensive analysis module combines NSI and NCE to calculate the neurodynamic index, which reflects the activity state of the patient's sympathetic nervous system and indirectly infers the depth of anesthesia. The formula is: ; in, represents the neural dynamic index, and They all represent weighting coefficients, which are adjusted according to different anesthesia stages to comprehensively reflect the changes in skin electrical signals.

[0008] Further, in an anesthesia depth monitoring system based on skin electrical signal analysis, the dynamic signal enhancement and adaptive noise reduction processing subsystem includes an adaptive pattern enhancement module, a timing artifact removal module, and a frequency domain signal optimization module; The adaptive mode enhancement module optimizes the signal characteristics through collaborative filters based on the multi-dimensional signals collected in real time. The formula is: ; in, represents the optimized signal after the adaptive mode enhancement, represents the original signal, represents the noise component, α and β represent adjustment coefficients, which dynamically optimize and enhance the quality of the signal; The temporal artifact removal module automatically corrects artifacts caused by patient activity or external interference by learning the temporal dependency of the signal. The formula is: ; in, represents the signal after removing artifacts, represents the artifact component; The frequency domain signal optimization module performs frequency domain analysis on the signal to remove invalid high-frequency noise while retaining low-frequency information to improve the overall quality of the signal. The formula is: ; Said represents the optimized signal at time t after frequency domain signal optimization, j represents the frequency component index in the frequency domain, and only the signal components within the effective frequency range are considered. Represents the value of the original signal S(t) corresponding to the frequency component j, Represents the low-frequency range, which is the frequency domain interval of the effective signal.

[0009] Further, an anesthesia depth monitoring system based on skin electrical signal analysis, the dynamic signal enhancement and adaptive noise reduction processing subsystem includes an input layer and data preprocessing module, a convolution layer module, a long short-term memory network module, a fusion layer module, and a fully connected layer module; The input layer and data preprocessing module receive multimodal data collected from different sensors and convert them into high-dimensional feature vectors after preprocessing. The formula is: ; Among them, HRV stands for heart rate variability, Resp stands for respiratory rate, Represents a high-dimensional feature vector; The convolution layer module learns local changes in the signal through the convolution kernel, and the convolution formula is: ; in, represents the output feature map of the lth layer of the convolutional neural network, represents the convolution kernel, represents the activation function, represents the output of the previous layer, represents the bias term, which is used to adjust the model; The long short-term memory network module models the temporal changes of skin electrical signals and heart rate signals and captures the temporal characteristics of the signals. The formula is: ; in, Represents the cell state at the current moment, represents the forget gate, represents the input gate, Represents the cell state at the previous moment, Represents the current candidate memory; The fusion layer module combines the extracted local features and temporal features to obtain the final anesthesia depth assessment result; The fully connected layer outputs the classification result of anesthesia depth; The classification results include light anesthesia, deep anesthesia, and loss of consciousness.

[0010] Further, in an anesthesia depth monitoring system based on skin electrical signal analysis, the reinforcement learning driven adaptive controller includes a reward function module, a drug regulation strategy module, and a reinforcement learning algorithm optimization module; The reward function module provides feedback on the real-time assessment of the depth of anesthesia and the amount of drug adjustment, and the formula is: ; in, represents the reward value at time step t, and represents the adjustment coefficient, represents the adjusted amount of drug infusion, represents the target depth of anesthesia, Represents the current depth of anesthesia; The drug adjustment strategy module adjusts the drug infusion volume in real time according to the difference between the current anesthesia depth and the target anesthesia depth. The formula is: ; in, represents the adjusted drug infusion volume, Represents the current drug infusion volume, that is, the amount of drug used by the system at the last moment or step. represents the learning rate, which automatically adjusts the drug regulation amplitude; The reinforcement learning algorithm optimization module selects the best drug regulation action according to the patient's anesthesia depth and physiological signals at each moment, optimizes the use of anesthetic drugs, and makes adjustments through feedback. It continuously interacts with the environment through reinforcement learning to learn the optimal drug regulation strategy.

[0011] Furthermore, an anesthesia depth monitoring system based on skin electrical signal analysis also includes a hardware subsystem and a software processing subsystem; The hardware subsystem includes a skin electrical sensor, a heart rate sensor, and a respiratory rate sensor; The sampling frequency of the skin electrical sensor is 500Hz, and the material is a flexible electrode, the electrode uses a conductive gel, and the electrode is placed on the patient's palm, sole or chest, and is adjusted according to the body shape and anesthesia site of different patients; The PPG sampling frequency of the heart rate sensor is 100Hz, using multi-frequency infrared light source technology; The respiratory rate sensor selects a micro airflow sensor or a pressure sensor to monitor the patient's respiratory rate in real time, with a sampling rate of 1 Hz, which is suitable for periodically changing respiratory signals; The software processing subsystem includes a data acquisition module, a wireless transmission module, a data processing unit, a drug infusion module, and an automatic adjustment mechanism; The data of the data acquisition module is transmitted to the data processing unit in real time via the wireless transmission module; The wireless transmission module adopts low-power Bluetooth technology to avoid signal loss or delay; The data processing unit supports real-time multitasking and can synchronously receive data from multiple signal sources for parallel processing during anesthesia; The drug infusion module includes an intelligent pump, a drug infusion pipeline, an infusion pressure sensor and a flow sensor, and automatically adjusts the drug infusion volume according to the real-time assessed depth of anesthesia; The automatic adjustment mechanism automatically adjusts the drug flow of the infusion module according to the deviation between the anesthesia depth and the target depth through a feedback control system, and customizes the drug infusion strategy through input parameters to ensure the anesthesia effect.

[0012] A method for monitoring the depth of anesthesia based on skin electrical signal analysis comprises the following steps: S1: Signal preprocessing: Acquire the physiological signals of skin electrical signals, heart rate signals, and respiratory rate through the sensors of the hardware subsystem; S2: Signal denoising: Use the multimodal signal collaborative filtering, time artifact removal and frequency domain signal optimization technology of the neural dynamic skin electrode analysis subsystem and the dynamic signal enhancement and adaptive noise reduction processing subsystem to eliminate noise and artifacts in the signal and retain valid information; S3: Anesthesia depth assessment and feature extraction: Using deep learning technology and multimodal signal fusion strategy, the anesthesia depth is accurately assessed through the neural dynamic skin electrodermal analysis system; S4: Drug regulation and reinforcement learning: Reinforcement learning drives the adaptive controller to dynamically adjust the drug infusion volume to ensure that the anesthesia depth is always maintained within the target range; S41: According to the deviation between the anesthesia depth and the target depth, the effect of each drug infusion is adjusted through the reward function module and the drug regulation strategy module, and the infusion amount is automatically adjusted to minimize the anesthesia depth error in the reward function; S5: Real-time monitoring and feedback mechanism: Real-time monitoring of the patient's anesthesia depth, timely adjustment of drug infusion volume through the feedback mechanism. During the anesthesia process, the doctor views the real-time data through the graphical interface and intervenes according to the system's suggestions. When the anesthesia depth exceeds the preset range, the doctor is reminded to intervene through sound alarms and graphical interface warnings, and the drug infusion is automatically adjusted to ensure the safety of the patient; S6: Clinical verification and system optimization: Continuously optimize the anesthesia depth assessment model through verification and feedback of clinical data.

[0013] Further, a method for monitoring the depth of anesthesia based on skin electrical signal analysis, said S2 comprises the following steps: S21: Separate the noise and effective signal of each sensor signal through multi-modal signal collaborative filtering to improve the signal quality; S211: At a certain moment , the signals collected from a sensors are ,Multimodal signal collaborative filtering uses weighted average method, combined with the characteristics of each sensor signal, to calculate the final collaborative filtering signal. The formula is: ; in, represents the final collaborative filtering signal, Representative The weight of the sensor signal, Automatically adjusts by calculating noise level, signal strength and correlation in real time, represents the signal obtained by the bth sensor at time t, and a represents the total number of sensors participating in collaborative filtering in the system; S212: Automatically adjust weights based on the signal-to-noise ratio of each signal through an adaptive learning process : When the noise is strong, increase the filtering strength to eliminate unnecessary noise; When the signal quality is good, reduce the filtering intensity to maintain the original characteristics of the signal; S22: Identify and remove artifacts by modeling the timing signal; S221: Set at time , the signal collected from the sensor is , the normal model established for historical data is , calculate the artifact component , the formula is: ; S222: Modeling the time series signal in combination with a deep learning algorithm, learning the time series characteristics of the signal, identifying in real time and dynamically removing artifacts caused by movement, body position changes, etc. through a time series artifact removal module, correcting the signal, and optimizing the time series artifact removal module through a feedback mechanism; S23: Perform frequency domain analysis on the signal through the frequency domain signal optimization module to remove high-frequency noise and retain the low-frequency effective components; S231: Perform fast Fourier transform on the signal to convert the signal from time domain to frequency domain. The formula is: ; in, represents the frequency domain representation of the signal, represents the Fourier transform operation; S232: Design a bandpass filter to remove high-frequency noise components and retain only the low-frequency signal reflecting the depth of anesthesia. Convert the optimized frequency domain signal back to the time domain through inverse Fourier transform. The formula is: ; in, represents the frequency domain signal after removing high-frequency noise, represents the time domain signal after denoising; S233: According to the spectrum characteristics of each signal, the frequency band of high-frequency noise is automatically identified, and the filter parameters are dynamically adjusted to remove interference signals and retain useful physiological information; S24: performing quality assessment on the final processed signal and optimizing the signal quality by signal-to-noise ratio; S241: By calculating the signal-to-noise ratio , which measures the signal quality, is expressed as: ; Among them, Signal Power represents the power of the effective component of the signal, and Noise Power represents the power of the noise component; S242: Denoised and optimized signal and As input for subsequent anesthesia depth assessment, the deep learning model performs feature extraction and classification, and transmits the optimized data to the doctor in real time to provide instant feedback and monitoring.

[0014] Further, a method for monitoring the depth of anesthesia based on skin electrical signal analysis, said S3 comprises the following steps: S31: The dynamic sensitivity analysis module of the skin electrical response calculates the nerve sensitivity index in real time every second, and reflects the fluctuation of the depth of anesthesia according to the change trend; S32: Quantify the complexity of the signal through the neural complexity entropy of the skin electrical signal complexity analysis module, infer the impact of anesthesia on the sympathetic nervous system, and provide real-time feedback; S33: The neurodynamic index of the neurodynamic index comprehensive analysis module is used as a comprehensive evaluation indicator of anesthesia depth, which is automatically adjusted according to the patient's real-time physiological changes to help anesthesiologists accurately evaluate anesthesia depth; S34: Extract local features of skin electrical signals and heart rate signals through convolutional neural networks, and model temporal dependencies through long short-term memory networks to capture the changes in anesthesia depth over time; S35: The output features of the convolutional neural network and the long short-term memory network are fused to finally generate an evaluation result of the anesthesia depth, which is dynamically adjusted based on the deep learning algorithm.

[0015] The beneficial effects of the present invention are: a system and method for monitoring the depth of anesthesia based on skin electrical signal analysis, which provides an accurate anesthesia depth monitoring solution through multimodal analysis of skin electrical signals; based on an adaptive drug adjustment algorithm and reinforcement learning, real-time adjustment of the anesthesia depth is achieved to prevent excessive or shallow anesthesia; personalized anesthesia management is performed in combination with individual differences of patients to improve anesthesia safety; compared with traditional EEG methods, the GSR signal acquisition method is non-invasive and more comfortable, which is convenient for long-term monitoring of patients; through innovative signal processing, deep learning and adaptive adjustment mechanisms, a more intelligent, accurate and personalized anesthesia depth monitoring solution can be provided, providing new technical support for clinical anesthesia management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a structural diagram of an anesthesia depth monitoring system based on skin electrical signal analysis.

[0017] Figure 2 It is a flow chart of an anesthesia depth monitoring method based on skin electrical signal analysis. DETAILED DESCRIPTION

[0018] The present invention is further described below, but the protection scope of the present invention is not limited to the following description.

[0019] As attached Figure 1 As shown, an anesthesia depth monitoring system based on skin electrical signal analysis includes a neural dynamic skin electrical analysis subsystem, a dynamic signal enhancement and adaptive noise reduction processing subsystem, a polymorphic fusion deep network subsystem, and a reinforcement learning driven adaptive controller; The neural dynamic skin electrodermal analysis subsystem uses a multi-dimensional dynamic analysis method based on skin electrodermal response, combining real-time changes and neural complexity to monitor the depth of anesthesia; Among them, the neurodynamic skin electrodermal analysis subsystem is based on the neurodynamic skin electrodermal analysis method NDSECA. By combining the analysis of the skin electrodermal response (GSR) and the dynamic changes of the nervous system, it can directly reflect the depth of anesthesia. The neurodynamic skin electrodermal analysis method NDSECA can not only capture the instantaneous changes of the skin electrodermal signal, but also deeply explore the dynamic response of the nervous system, providing a new technical path for the accurate monitoring of the depth of anesthesia. The dynamic signal enhancement and adaptive noise reduction processing subsystem removes noise and artifacts in the skin electrical signal and enhances the effective information of the signal; The dynamic signal enhancement and adaptive noise reduction processing subsystem is based on a dynamic signal enhancement and adaptive noise reduction processing algorithm DSE-ADA, which combines the ideas of adaptive signal enhancement, time series mode self-learning and multi-domain optimization. The polymorphic fusion deep network subsystem combines convolutional neural networks and long short-term memory networks to extract high-precision features from multimodal signals and accurately assess the depth of anesthesia, achieving the dual advantages of local feature extraction and temporal dependency modeling in a deep learning architecture; The reinforcement learning driven adaptive controller ensures that the anesthesia depth is maintained at an optimal state by dynamically adjusting the anesthetic drug infusion amount.

[0020] Further, an anesthesia depth monitoring system based on skin electrical signal analysis, the neurodynamic skin electrical analysis subsystem includes a skin electrical response dynamic sensitivity analysis module, a skin electrical signal complexity analysis module, and a neurodynamic index comprehensive analysis module; The dynamic sensitivity analysis module of the skin electrical response quantifies the reaction speed of the sympathetic nervous system by calculating the instantaneous rate of change of skin conductance, and the formula is: ; in, It stands for nerve sensitivity index and is used to capture changes in the activity of the sympathetic nervous system during anesthesia. Represents skin conductance at time interval Changes within; Among them, the skin electrode response (SCR) is controlled by the activity of the sympathetic nervous system and therefore reflects the changes in the depth of anesthesia; The complexity analysis module of the skin electrical signal calculates the entropy value of the signal through a multi-scale method to reveal the nonlinear dynamics in the skin electrical signal and the complex pattern of anesthesia depth. The formula is: ; in, represents the neural complexity entropy, represents the probability distribution of the skin electrical signal state, i represents the different states of the signal, and n represents the total number of states into which the signal is divided. For example, if the signal is divided into 5 different amplitude ranges, then n=5; Among them, neural complexity entropy, based on the nonlinear dynamic mode (NDM) analysis method, is used to quantify the complexity of skin electrical signals. Skin electrical signals are highly nonlinear and self-similar, and can reflect the complex neural activities during anesthesia. The neurodynamic index comprehensive analysis module combines NSI and NCE to calculate the neurodynamic index, which reflects the activity state of the patient's sympathetic nervous system and indirectly infers the depth of anesthesia. The formula is: ; in, represents the neural dynamic index, They all represent weighting coefficients, which are adjusted according to different anesthesia stages to comprehensively reflect the changes in skin electrical signals.

[0021] Further, in an anesthesia depth monitoring system based on skin electrical signal analysis, the dynamic signal enhancement and adaptive noise reduction processing subsystem includes an adaptive pattern enhancement module, a timing artifact removal module, and a frequency domain signal optimization module; The adaptive mode enhancement module optimizes the signal characteristics through collaborative filters based on the multi-dimensional signals collected in real time. The formula is: ; in, represents the optimized signal after the adaptive mode enhancement, represents the original signal, represents the noise component, α and β represent adjustment coefficients, which dynamically optimize and enhance the quality of the signal; Among them, the adaptive mode enhancement module uses the Multi-modal Collaborative Filtering (MMCF) algorithm in the early stage of signal processing to automatically identify and enhance effective signal modes by analyzing the timing characteristics of the signal and the external environmental noise, based on multi-dimensional signals collected in real time, such as skin electricity, heart rate, respiratory rate, etc.; The temporal artifact removal module automatically corrects artifacts caused by patient activity or external interference by learning the temporal dependency of the signal. The formula is: ; in, represents the signal after removing artifacts, represents the artifact component; In the frequency domain, the noise components of the signal are mainly concentrated in the high-frequency area, while the useful information is usually concentrated in the low-frequency area; The frequency domain signal optimization module performs frequency domain analysis on the signal to remove invalid high-frequency noise while retaining low-frequency information to improve the overall quality of the signal. The formula is: ; Said represents the optimized signal at time t after frequency domain signal optimization, j represents the frequency component index in the frequency domain, and only the signal components within the effective frequency range are considered. Represents the original signal At the value corresponding to the frequency component j, Represents the low-frequency range, which is the frequency domain interval of the effective signal.

[0022] Furthermore, an anesthesia depth monitoring system based on skin electrical signal analysis, the dynamic signal enhancement and adaptive noise reduction processing subsystem includes an input layer and data preprocessing module, a convolutional layer module CNN, a long short-term memory network module LSTM, a fusion layer module Fusion Layer, and a fully connected layer module; The input layer and data preprocessing module receive multimodal data collected from different sensors and convert them into high-dimensional feature vectors after preprocessing. The formula is: ; Among them, HRV stands for heart rate variability, Resp stands for respiratory rate, Represents a high-dimensional feature vector; The convolution layer module learns local changes in the signal through the convolution kernel, and the convolution formula is: ; in, represents the output feature map of the lth layer of the convolutional neural network, represents the convolution kernel, represents the activation function, represents the output of the previous layer, represents the bias term, which is used to adjust the model; Among them, the convolutional layer module CNN is used to extract the local features of each signal, especially the dynamic patterns in the skin electrode signal GSR and the heart rate signal; The long short-term memory network module models the temporal changes of skin electrical signals and heart rate signals and captures the temporal characteristics of the signals. The formula is: ; in, Represents the cell state at the current moment, represents the forget gate, represents the input gate, Represents the cell state at the previous moment, Represents the current candidate memory; The fusion layer module combines the extracted local features and temporal features to obtain the final anesthesia depth assessment result; The fully connected layer outputs the classification result of anesthesia depth; The classification results include light anesthesia, deep anesthesia, and loss of consciousness.

[0023] Further, in an anesthesia depth monitoring system based on skin electrical signal analysis, the reinforcement learning driven adaptive controller includes a reward function module, a drug regulation strategy module, and a reinforcement learning algorithm optimization module; The reward function module provides feedback on the real-time assessment of the depth of anesthesia and the amount of drug adjustment, and the formula is: ; in, represents the reward value at time step t, and represents the adjustment coefficient, represents the adjusted amount of drug infusion, represents the target depth of anesthesia, Represents the current depth of anesthesia; The drug adjustment strategy module adjusts the drug infusion volume in real time according to the difference between the current anesthesia depth and the target anesthesia depth. The formula is: ; in, represents the adjusted drug infusion volume, Represents the current drug infusion volume, that is, the amount of drug used by the system at the last moment or step. represents the learning rate, which automatically adjusts the drug regulation amplitude; The reinforcement learning algorithm optimization module selects the best drug regulation action according to the patient's anesthesia depth and physiological signals at each moment, optimizes the use of anesthetic drugs, and makes adjustments through feedback. It continuously interacts with the environment through reinforcement learning to learn the optimal drug regulation strategy.

[0024] Furthermore, an anesthesia depth monitoring system based on skin electrical signal analysis also includes a hardware subsystem and a software processing subsystem; The hardware subsystem includes a skin electrical sensor, a heart rate sensor, and a respiratory rate sensor; The galvanic skin sensor, i.e., the GSR sensor: the galvanic skin sensor uses a flexible electrode material, which can improve the stability of contact with the skin and reduce the influence of motion artifacts. The electrode part is designed with conductive gel to reduce the contact impedance with the skin and ensure stable signal acquisition. The sampling frequency is set to 500 Hz to ensure that the changes in the galvanic skin signal can be captured in time to reflect the dynamic response of the sympathetic nervous system. This frequency can accurately capture the short-term fluctuations of the sympathetic nerves and adapt to the rapid changes during anesthesia. The electrodes are placed on the palms, soles or chest of the patient and can be adjusted according to the body shape and anesthesia site of different patients. The heart rate sensor, namely PPG technology: the heart rate sensor adopts multi-frequency infrared light source LED technology, which enhances its sensing ability to blood flow changes and reduces the interference of ambient light on data collection; the amplitude of the PPG signal is closely related to the volume change of blood flow, and can accurately monitor the heart rate; the sampling frequency of the PPG sensor is 100Hz, which ensures real-time monitoring of heart rate changes and fusion analysis with skin electrical signals.

[0025] The respiratory rate sensor selects a micro airflow sensor or a pressure sensor to monitor the patient's respiratory rate in real time, with a sampling rate of 1 Hz, which is suitable for periodically changing respiratory signals, and accurately evaluates the impact of anesthesia based on the relationship between the patient's respiratory rate and the depth of anesthesia; The software processing subsystem includes a data acquisition module, a wireless transmission module, a data processing unit, a drug infusion module, and an automatic adjustment mechanism; The data of the data acquisition module is transmitted to the data processing unit in real time via the wireless transmission module; The wireless transmission module adopts low-power Bluetooth technology to avoid signal loss or delay; The data processing unit supports real-time multitasking and can synchronously receive data from multiple signal sources for parallel processing during anesthesia; The drug infusion module includes an intelligent pump, a drug infusion pipeline, an infusion pressure sensor and a flow sensor, and automatically adjusts the drug infusion volume according to the real-time assessed depth of anesthesia; The automatic adjustment mechanism automatically adjusts the drug flow rate of the infusion module according to the deviation between the anesthesia depth and the target depth through a feedback control system, and customizes the drug infusion strategy through input parameters to ensure the anesthesia effect; The input parameters include weight, age, type of surgery, etc.

[0026] As attached Figure 2 As shown, a method for monitoring the depth of anesthesia based on skin electrical signal analysis comprises the following steps: S1: Signal preprocessing: Acquire the physiological signals of skin electrical signals, heart rate signals, and respiratory rate through the sensors of the hardware subsystem; Among them, the signal is usually affected by various noise sources, such as motion artifacts, electromagnetic interference, etc., so the signal must be effectively preprocessed to ensure the accuracy and stability of subsequent anesthesia depth assessment; S2: Signal denoising: Use the multimodal signal collaborative filtering, time artifact removal and frequency domain signal optimization technology of the neural dynamic skin electrode analysis subsystem and the dynamic signal enhancement and adaptive noise reduction processing subsystem to eliminate noise and artifacts in the signal and retain valid information; S3: Anesthesia depth assessment and feature extraction: Using deep learning technology and multimodal signal fusion strategy, the anesthesia depth is accurately assessed through the neural dynamic skin electrodermal analysis system; S4: Drug regulation and reinforcement learning: Reinforcement learning drives the adaptive controller to dynamically adjust the drug infusion volume to ensure that the anesthesia depth is always maintained within the target range; S41: According to the deviation between the anesthesia depth and the target depth, the effect of each drug infusion is adjusted through the reward function module and the drug regulation strategy module, and the infusion amount is automatically adjusted to minimize the anesthesia depth error in the reward function; S5: Real-time monitoring and feedback mechanism: Real-time monitoring of the patient's anesthesia depth, timely adjustment of drug infusion volume through the feedback mechanism. During the anesthesia process, the doctor views the real-time data through the graphical interface and intervenes according to the system's suggestions. When the anesthesia depth exceeds the preset range, the doctor is reminded to intervene through sound alarms and graphical interface warnings, and the drug infusion is automatically adjusted to ensure the safety of the patient; S6: Clinical verification and system optimization: Continuously optimize the anesthesia depth assessment model through verification and feedback of clinical data.

[0027] Further, a method for monitoring the depth of anesthesia based on skin electrical signal analysis, said S2 comprises the following steps: S21: Separate the noise and effective signal of each sensor signal through multi-modal signal collaborative filtering to improve the signal quality; During anesthesia, physiological signals such as skin electrical signals, heart rate and respiratory rate have certain correlations in the process of changes in anesthesia depth. For example, when the anesthesia depth is deep, the patient's sympathetic nerve activity will change significantly, and skin electrical signals and heart rate signals usually fluctuate synchronously. Therefore, multimodal signal collaborative filtering can filter multiple signal sources through weighted averaging, thereby optimizing signal quality. S211: At a certain moment , the signals collected from a sensors are ,Multimodal signal collaborative filtering uses weighted average method, combined with the characteristics of each sensor signal, to calculate the final collaborative filtering signal. The formula is: ; in, represents the final collaborative filtering signal, Representative The weight of the sensor signal, Automatically adjusts by calculating noise level, signal strength and correlation in real time, represents the signal obtained by the bth sensor at time t, and a represents the total number of sensors participating in collaborative filtering in the system; S212: Automatically adjust weights based on the signal-to-noise ratio of each signal through an adaptive learning process : When the noise is strong, increase the filtering strength to eliminate unnecessary noise; When the signal quality is good, reduce the filtering intensity to maintain the original characteristics of the signal; It can also be understood as dynamically optimizing the enhanced signal by adjusting the coefficients of the original signal S(t) and the noise component N(t). quality; S22: Identify and remove artifacts by modeling the timing signal; Artifacts are usually caused by changes in patient position, movement, or poor sensor contact. Traditional artifact removal methods usually make corrections based on thresholds, which sometimes mistakenly remove valid signals. S221: Set at time , the signal collected from the sensor is , the normal model established for historical data is , calculate the artifact component , the formula is: ; S222: Modeling the time series signal in combination with a deep learning algorithm, learning the time series characteristics of the signal, identifying in real time and dynamically removing artifacts caused by movement, body position changes, etc. through a time series artifact removal module, correcting the signal, and optimizing the time series artifact removal module through a feedback mechanism; S23: Perform frequency domain analysis on the signal through the frequency domain signal optimization module to remove high-frequency noise and retain the low-frequency effective components; S231: Perform fast Fourier transform on the signal to convert the signal from time domain to frequency domain. The formula is: ; in, represents the frequency domain representation of the signal, represents the Fourier transform operation; S232: Design a bandpass filter to remove high-frequency noise components and retain only the low-frequency signal reflecting the depth of anesthesia. Convert the optimized frequency domain signal back to the time domain through inverse Fourier transform. The formula is: ; in, represents the frequency domain signal after removing high-frequency noise, represents the time domain signal after denoising; S233: According to the spectrum characteristics of each signal, the frequency band of high-frequency noise is automatically identified, and the filter parameters are dynamically adjusted to remove interference signals and retain useful physiological information; S24: performing quality assessment on the final processed signal and optimizing the signal quality by signal-to-noise ratio; S241: By calculating the signal-to-noise ratio , which measures the signal quality, is expressed as: ; Among them, Signal Power represents the power of the effective component of the signal, and Noise Power represents the power of the noise component; S242: Denoised and optimized signal and As input for subsequent anesthesia depth assessment, the deep learning model performs feature extraction and classification, and transmits the optimized data to the doctor in real time to provide instant feedback and monitoring.

[0028] Further, a method for monitoring the depth of anesthesia based on skin electrical signal analysis, said S3 comprises the following steps: S31: The dynamic sensitivity analysis module of the skin electrical response calculates the nerve sensitivity index in real time every second, and reflects the fluctuation of the depth of anesthesia according to the change trend; S32: Quantify the complexity of the signal through the neural complexity entropy of the skin electrical signal complexity analysis module, infer the impact of anesthesia on the sympathetic nervous system, and provide real-time feedback; S33: The neurodynamic index of the neurodynamic index comprehensive analysis module is used as a comprehensive evaluation indicator of anesthesia depth, which is automatically adjusted according to the patient's real-time physiological changes to help anesthesiologists accurately evaluate anesthesia depth; S34: Extract local features of skin electrical signals and heart rate signals through convolutional neural networks, and model temporal dependencies through long short-term memory networks to capture the changes in anesthesia depth over time; S35: The output features of the convolutional neural network and the long short-term memory network are fused to finally generate an evaluation result of the anesthesia depth, which is dynamically adjusted based on the deep learning algorithm.

[0029] The benefits of the present invention are: 1. Improve the accuracy and real-time performance of anesthesia depth monitoring: Traditional anesthesia depth monitoring methods rely on physiological indicators, such as heart rate, blood pressure or EEG technology, which have certain limitations. By introducing skin electrical signals GSR as an objective physiological signal, combined with advanced signal processing and feature extraction algorithms, more accurate anesthesia depth assessment results can be obtained in real-time monitoring, thereby reducing the uncertainty in the clinical anesthesia process; 2. Personalized anesthesia management: Through comprehensive analysis of multimodal physiological signals, such as skin electrical signals, heart rate variability, respiratory rate, etc., the dosage of anesthetic drugs can be dynamically adjusted according to individual differences of patients and real-time changes in anesthesia depth, ensuring that the anesthesia depth is maintained within the safest and most effective range, reducing the risk of drug overdose or underdose; 3. Improve the non-invasiveness and comfort of anesthesia depth monitoring: Compared with the traditional EEG monitoring method, the skin electrical signal GSR monitoring method is non-invasive, convenient and does not require complex equipment, avoiding the discomfort caused by EEG monitoring for patients. At the same time, it has better portability and ease of use, and is suitable for various clinical scenarios; 4. Reduce anesthesia risks and ensure surgical safety: By establishing an accurate anesthesia depth assessment model and combining it with adaptive adjustment technology, it is possible to timely detect the changing trend of anesthesia depth, avoid too shallow or too deep anesthesia, ensure the safety of patients during surgery, and effectively reduce the occurrence of anesthesia-related complications; 5. Optimize the use of anesthetic drugs and save medical resources: The real-time feedback mechanism helps anesthesiologists adjust the use of drugs in a timely manner, ensure that the depth of anesthesia remains within the appropriate range, avoid excessive use of anesthetic drugs, reduce drug waste and side effects, and improve the efficiency of anesthetic drug use.

[0030] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An anesthesia depth monitoring system based on skin electrical signal analysis, characterized in that: Including neural dynamic skin electricity analysis subsystem, dynamic signal enhancement and adaptive noise reduction processing subsystem, polymorphic fusion deep network subsystem, reinforcement learning driven adaptive controller; The neural dynamic skin electrodermal analysis subsystem uses a multi-dimensional dynamic analysis method based on skin electrodermal response, combining real-time changes and neural complexity to monitor the depth of anesthesia; The dynamic signal enhancement and adaptive noise reduction processing subsystem removes noise and artifacts in the skin electrical signal and enhances the effective information of the signal; The polymorphic fusion deep network subsystem combines convolutional neural networks and long short-term memory networks to extract high-precision features from multimodal signals and accurately assess the depth of anesthesia; The reinforcement learning driven adaptive controller ensures that the anesthesia depth is maintained at an optimal state by dynamically adjusting the anesthetic drug infusion amount.

2. The anesthesia depth monitoring system based on skin electrical signal analysis according to claim 1 is characterized in that: The neural dynamic skin electrical analysis subsystem includes a skin electrical response dynamic sensitivity analysis module, a skin electrical signal complexity analysis module, and a neural dynamic index comprehensive analysis module; The dynamic sensitivity analysis module of the skin electrical response quantifies the reaction speed of the sympathetic nervous system by calculating the instantaneous rate of change of skin conductance, and the formula is: ； in, It stands for nerve sensitivity index and is used to capture changes in the activity of the sympathetic nervous system during anesthesia. Represents skin conductance at time interval Changes within; The complexity analysis module of the skin electrical signal calculates the entropy value of the signal through a multi-scale method to reveal the nonlinear dynamics in the skin electrical signal and the complex pattern of anesthesia depth. The formula is: ； in, represents the neural complexity entropy, represents the probability distribution of the skin electrical signal state, i represents the different states of the signal, and n represents the total number of states into which the signal is divided; The neurodynamic index comprehensive analysis module combines NSI and NCE to calculate the neurodynamic index, which reflects the activity state of the patient's sympathetic nervous system and indirectly infers the depth of anesthesia. The formula is: ； in, represents the neural dynamic index, and They all represent weighting coefficients, which are adjusted according to different anesthesia stages to comprehensively reflect the changes in skin electrical signals.

3. The anesthesia depth monitoring system based on skin electrical signal analysis according to claim 1 is characterized in that: The dynamic signal enhancement and adaptive noise reduction processing subsystem includes an adaptive pattern enhancement module, a timing artifact removal module, and a frequency domain signal optimization module; The adaptive mode enhancement module optimizes the signal characteristics through collaborative filters based on the multi-dimensional signals collected in real time. The formula is: ； in, represents the optimized signal after the adaptive mode enhancement, represents the original signal, represents the noise component, α and β represent adjustment coefficients, which dynamically optimize and enhance the quality of the signal; The temporal artifact removal module automatically corrects artifacts caused by patient activity or external interference by learning the temporal dependency of the signal. The formula is: ； in, represents the signal after removing artifacts, represents the artifact component; The frequency domain signal optimization module performs frequency domain analysis on the signal to remove invalid high-frequency noise while retaining low-frequency information to improve the overall quality of the signal. The formula is: ； Said represents the optimized signal at time t after frequency domain signal optimization, j represents the frequency component index in the frequency domain, and only the signal components within the effective frequency range are considered. Represents the value of the original signal S(t) corresponding to the frequency component j, Represents the low-frequency range, which is the frequency domain interval of the effective signal.

4. The anesthesia depth monitoring system based on skin electrical signal analysis according to claim 1, characterized in that: The dynamic signal enhancement and adaptive noise reduction processing subsystem includes an input layer and data preprocessing module, a convolutional layer module, a long short-term memory network module, a fusion layer module, and a fully connected layer module; The input layer and data preprocessing module receive multimodal data collected from different sensors and convert them into high-dimensional feature vectors after preprocessing. The formula is: ； Among them, HRV stands for heart rate variability, Resp stands for respiratory rate, Represents a high-dimensional feature vector; The convolution layer module learns local changes in the signal through the convolution kernel, and the convolution formula is: ； in, represents the output feature map of the lth layer of the convolutional neural network, represents the convolution kernel, represents the activation function, represents the output of the previous layer, represents the bias term, which is used to adjust the model; The long short-term memory network module models the temporal changes of skin electrical signals and heart rate signals and captures the temporal characteristics of the signals. The formula is: ； in, Represents the cell state at the current moment, represents the forget gate, represents the input gate, Represents the cell state at the previous moment, Represents the current candidate memory; The fusion layer module combines the extracted local features and temporal features to obtain the final anesthesia depth assessment result; The fully connected layer outputs the classification result of the anesthesia depth; The classification results include light anesthesia, deep anesthesia, and loss of consciousness.

5. The anesthesia depth monitoring system based on skin electrical signal analysis according to claim 1, characterized in that: The reinforcement learning driven adaptive controller includes a reward function module, a drug regulation strategy module, and a reinforcement learning algorithm optimization module; The reward function module provides feedback on the real-time assessment of the depth of anesthesia and the amount of drug adjustment, and the formula is: ； in, represents the reward value at time step t, and represents the adjustment coefficient, represents the adjusted amount of drug infusion, represents the target depth of anesthesia, Represents the current depth of anesthesia; The drug adjustment strategy module adjusts the drug infusion volume in real time according to the difference between the current anesthesia depth and the target anesthesia depth. The formula is: ； in, represents the adjusted drug infusion volume, Represents the current drug infusion volume, that is, the amount of drug used by the system at the last moment or step. represents the learning rate, which automatically adjusts the drug regulation amplitude; The reinforcement learning algorithm optimization module selects the best drug regulation action according to the patient's anesthesia depth and physiological signals at each moment, optimizes the use of anesthetic drugs, and makes adjustments through feedback. It continuously interacts with the environment through reinforcement learning to learn the optimal drug regulation strategy.

6. The anesthesia depth monitoring system based on skin electrical signal analysis according to claim 1, characterized in that: It also includes hardware subsystems and software processing subsystems; The hardware subsystem includes a skin electrical sensor, a heart rate sensor, and a respiratory rate sensor; The sampling frequency of the skin electrical sensor is 500Hz, and the material is a flexible electrode, the electrode uses a conductive gel, and the electrode is placed on the patient's palm, sole or chest, and is adjusted according to the body shape and anesthesia site of different patients; The PPG sampling frequency of the heart rate sensor is 100Hz, using multi-frequency infrared light source technology; The respiratory rate sensor selects a micro airflow sensor or a pressure sensor to monitor the patient's respiratory rate in real time, with a sampling rate of 1 Hz, which is suitable for periodically changing respiratory signals; The software processing subsystem includes a data acquisition module, a wireless transmission module, a data processing unit, a drug infusion module, and an automatic adjustment mechanism; The data of the data acquisition module is transmitted to the data processing unit in real time via the wireless transmission module; The wireless transmission module adopts low-power Bluetooth technology to avoid signal loss or delay; The data processing unit supports real-time multitasking and can synchronously receive data from multiple signal sources for parallel processing during anesthesia; The drug infusion module includes an intelligent pump, a drug infusion pipeline, an infusion pressure sensor and a flow sensor, and automatically adjusts the drug infusion volume according to the real-time assessed depth of anesthesia; The automatic adjustment mechanism automatically adjusts the drug flow of the infusion module according to the deviation between the anesthesia depth and the target depth through a feedback control system, and customizes the drug infusion strategy through input parameters to ensure the anesthesia effect.

7. A method for monitoring the depth of anesthesia based on skin electrical signal analysis, implemented by the system for monitoring the depth of anesthesia based on skin electrical signal analysis according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1: Signal preprocessing: Acquire the physiological signals of skin electrical signals, heart rate signals, and respiratory rate through the sensors of the hardware subsystem; S2: Signal denoising: Use the multimodal signal collaborative filtering, time artifact removal and frequency domain signal optimization technology of the neural dynamic skin electrode analysis subsystem and the dynamic signal enhancement and adaptive noise reduction processing subsystem to eliminate noise and artifacts in the signal and retain valid information; S3: Anesthesia depth assessment and feature extraction: Using deep learning technology and multimodal signal fusion strategy, the anesthesia depth is accurately assessed through the neural dynamic skin electrodermal analysis system; S4: Drug regulation and reinforcement learning: Reinforcement learning drives the adaptive controller to dynamically adjust the drug infusion volume to ensure that the anesthesia depth is always maintained within the target range; S41: According to the deviation between the anesthesia depth and the target depth, the effect of each drug infusion is adjusted through the reward function module and the drug regulation strategy module, and the infusion amount is automatically adjusted to minimize the anesthesia depth error in the reward function; S5: Real-time monitoring and feedback mechanism: Real-time monitoring of the patient's anesthesia depth, timely adjustment of drug infusion volume through the feedback mechanism. During the anesthesia process, the doctor views the real-time data through the graphical interface and intervenes according to the system's suggestions. When the anesthesia depth exceeds the preset range, the doctor is reminded to intervene through sound alarms and graphical interface warnings, and the drug infusion is automatically adjusted to ensure the safety of the patient; S6: Clinical verification and system optimization: Continuously optimize the anesthesia depth assessment model through verification and feedback of clinical data.

8. The method for monitoring the depth of anesthesia based on skin electrical signal analysis according to claim 7, characterized in that: The S2 comprises the following steps: S21: Separate the noise and effective signal of each sensor signal through multi-modal signal collaborative filtering to improve the signal quality; S211: At a certain moment , the signals collected from a sensors are ,Multimodal signal collaborative filtering uses weighted average method, combined with the characteristics of each sensor signal, to calculate the final collaborative filtering signal. The formula is: ； in, represents the final collaborative filtering signal, Representative The weight of the sensor signal, Automatically adjusts by calculating noise level, signal strength and correlation in real time, represents the signal obtained by the b-th sensor at time t, and a represents the total number of sensors participating in collaborative filtering in the system; S212: Automatically adjust weights based on the signal-to-noise ratio of each signal through an adaptive learning process : When the noise is strong, increase the filtering strength to eliminate unnecessary noise; When the signal quality is good, reduce the filtering intensity to maintain the original characteristics of the signal; S22: Identify and remove artifacts by modeling the timing signal; S221: Set at time , the signal collected from the sensor is , the normal model established for historical data is , calculate the artifact component , the formula is: ； S222: Modeling the time series signal in combination with a deep learning algorithm, learning the time series characteristics of the signal, identifying in real time and dynamically removing artifacts caused by movement, body position changes, etc. through a time series artifact removal module, correcting the signal, and optimizing the time series artifact removal module through a feedback mechanism; S23: Perform frequency domain analysis on the signal through the frequency domain signal optimization module to remove high-frequency noise and retain the low-frequency effective components; S231: Perform fast Fourier transform on the signal to convert the signal from time domain to frequency domain. The formula is: ； in, represents the frequency domain representation of the signal, represents the Fourier transform operation; S232: Design a bandpass filter to remove high-frequency noise components and retain only the low-frequency signal reflecting the depth of anesthesia. Convert the optimized frequency domain signal back to the time domain through inverse Fourier transform. The formula is: ； in, represents the frequency domain signal after removing high-frequency noise, represents the time domain signal after denoising; S233: According to the spectrum characteristics of each signal, the frequency band of high-frequency noise is automatically identified, and the filter parameters are dynamically adjusted to remove interference signals and retain useful physiological information; S24: performing quality assessment on the final processed signal and optimizing the signal quality by signal-to-noise ratio; S241: By calculating the signal-to-noise ratio of the signal , which measures the signal quality, is expressed as: ； Among them, Signal Power represents the power of the effective component of the signal, and Noise Power represents the power of the noise component; S242: Denoised and optimized signal and As input for subsequent anesthesia depth assessment, the deep learning model performs feature extraction and classification, and transmits the optimized data to the doctor in real time to provide instant feedback and monitoring.

9. The method for monitoring the depth of anesthesia based on skin electrical signal analysis according to claim 7, characterized in that: The S3 comprises the following steps: S31: The dynamic sensitivity analysis module of the skin electrical response calculates the nerve sensitivity index in real time every second, and reflects the fluctuation of the depth of anesthesia according to the change trend; S32: Quantify the complexity of the signal through the neural complexity entropy of the skin electrical signal complexity analysis module, infer the impact of anesthesia on the sympathetic nervous system, and provide real-time feedback; S33: The neurodynamic index of the neurodynamic index comprehensive analysis module is used as a comprehensive evaluation indicator of anesthesia depth, which is automatically adjusted according to the patient's real-time physiological changes to help anesthesiologists accurately evaluate anesthesia depth; S34: Extract local features of skin electrical signals and heart rate signals through convolutional neural networks, and model temporal dependencies through long short-term memory networks to capture the changes in anesthesia depth over time; S35: The output features of the convolutional neural network and the long short-term memory network are fused to finally generate an evaluation result of the anesthesia depth, which is dynamically adjusted based on the deep learning algorithm.

Images