Intraoperative patient anesthesia status assessment system based on EEG signals and ear temperature monitoring

Abstract

The present invention relates to the technical field of anesthesia state assessment, and specifically to an intraoperative patient anesthesia state assessment system based on EEG signals and ear temperature monitoring, comprising: a data acquisition unit for acquiring the EEG signals and external auditory canal temperature of a target patient based on EEG leads and ear temperature sensors, and synchronizing the EEG signals and external auditory canal temperature signals of the target patient based on a synchronous clock model; and an association modeling unit for dynamically performing association modeling on the target patient based on the EEG signals and external auditory canal temperature signals, so as to obtain a PAC feature matrix between the EEG signals and the external auditory canal temperature signals. By combining EEG signals and ear temperature monitoring data, the present invention can more comprehensively and accurately assess the patient's anesthesia depth. EEG signals can reflect the electrical activity of the brain, while ear temperature reflects the overall physiological state of the body, especially the effect of anesthetic drugs on body temperature. Comprehensively analyzing these two signals can effectively improve the accuracy of anesthesia state assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of anesthesia state assessment, and in particular to an intraoperative patient anesthesia state assessment system based on electroencephalogram (EEG) signals and ear temperature monitoring. Background Art

[0002] Traditional systems typically rely on a single physiological parameter, such as EEG or body temperature. This single-signal monitoring approach may not fully reflect the patient's anesthetic status. Furthermore, because traditional systems rely on a single signal, these signals are often subject to external interference (such as individual differences in anesthetic drugs and the external environment). This can lead to inaccurate assessments of the patient's anesthetic status, which in turn affects anesthetic control during surgery. Especially in a volatile surgical environment, traditional methods may not be able to capture changes in the patient's body in a timely and accurate manner. Traditional systems also typically use static models that are unable to capture dynamic changes in the patient's anesthetic state in real time. Anesthesia depth assessments may lag behind the patient's actual physiological changes, resulting in a slow response to anesthetic depth adjustments, which in turn affects the safety of the anesthetic process. Traditional systems often lack intelligent feedback mechanisms, and the anesthetic drug infusion rate is often manually monitored and adjusted, which carries the risk of human error. Especially in complex surgeries, it is difficult for anesthesiologists to simultaneously monitor all physiological parameters, and manual adjustments are easily affected by factors such as fatigue and misjudgment. The lack of real-time, automated feedback mechanisms can lead to over- or under-anesthesia, increasing the risk of anesthetic accidents. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide an intraoperative patient anesthesia status assessment system based on EEG signals and ear temperature monitoring.

[0004] The technical solution adopted to solve the above technical problems is: an intraoperative patient anesthesia status assessment system based on EEG signals and ear temperature monitoring, including:

[0005] A data acquisition unit, the data acquisition unit being used to acquire an EEG signal and an external auditory canal temperature of a target patient based on an EEG lead and an ear temperature sensor, and to synchronize the EEG signal and the external auditory canal temperature signal of the target patient based on a synchronous clock model;

[0006] an association modeling unit, configured to perform dynamic association modeling on the target patient based on the EEG signal and the external auditory canal temperature signal to obtain a PAC feature matrix between the EEG signal and the external auditory canal temperature signal;

[0007] A feature extraction unit, configured to construct a dynamic graph based on the channels of the EEG lead, the ear temperature sensor, and the PAC feature matrix, and perform spatiotemporal feature extraction on the dynamic graph based on a graph neural network to obtain spatiotemporal features of each node of the dynamic graph;

[0008] A feature fusion unit, configured to fuse features of each node of the dynamic graph using the spatiotemporal features according to an attention mechanism to obtain a fusion feature of the dynamic graph;

[0009] an anesthesia assessment unit, configured to perform anesthesia depth regression and anesthesia state classification on the target patient based on the fusion features of the dynamic image, so as to obtain a CSI index and anesthesia state level of the target patient;

[0010] An anesthesia control unit is used to feed back the CSI index and anesthesia state level of the target patient to the anesthesia infusion pump in real time, and dynamically adjust the anesthetic infusion rate according to the fuzzy PID controller.

[0011] Preferably, the expression of the synchronous clock model is as follows: ;

[0012] in, Indicates the signal time of synchronization, represents the acquisition time of EEG signals, represents the delayed acquisition time of the external auditory canal temperature signal, and , represents a Gaussian distribution, Indicates the standard deviation of the delay.

[0013] Preferably, dynamic correlation modeling is performed on the target patient according to the EEG signal and the external auditory canal temperature signal to obtain a PAC feature matrix between the EEG signal and the external auditory canal temperature signal, including:

[0014] The frequency band signals of the EEG signal are obtained by using a bandpass filter, wherein the frequency band signals include Wave, Wave, wave and wave, performing Hilbert transform on the signals of each frequency band to obtain the phase of the signals of each frequency band;

[0015] Calculating the low-frequency amplitude of the external auditory canal temperature signal according to short-time Fourier transform;

[0016] Calculating a coupling index between the phase of each frequency band signal and the low-frequency amplitude of the external auditory canal temperature signal according to the PAC;

[0017] A PAC characteristic matrix between the EEG signal and the external auditory canal temperature signal is obtained according to a coupling index between the phase of each frequency band signal and the low-frequency amplitude of the external auditory canal temperature signal.

[0018] Preferably, the calculation formula of the coupling index is as follows: ;

[0019] in, represents the phase-amplitude coupling index, represents the frequency band combination of phase-amplitude coupling, Indicates the phase of each frequency band of the EEG signal, represents the amplitude of the external auditory canal temperature, Indicates the total number of samples.

[0020] Preferably, constructing a dynamic graph according to each channel of the EEG lead, the ear temperature sensor and the PAC feature matrix includes:

[0021] Using each channel of the EEG lead and the ear temperature sensor as nodes of the dynamic graph, wherein the nodes of the dynamic graph include EEG lead nodes and ear temperature sensor nodes;

[0022] Using the low-frequency amplitude of the external auditory canal temperature signal and the PAC feature matrix as the feature vectors of each node of the dynamic graph;

[0023] Calculating the weight of each dynamic edge in the dynamic graph according to the feature vector of each node in the dynamic graph, and obtaining the adjacency matrix of the dynamic graph according to the weight of each dynamic edge in the dynamic graph;

[0024] The weights of each dynamic edge in the dynamic graph are updated according to a preset graph structure update frequency.

[0025] Preferably, the calculation formula of the weight of the dynamic edge is as follows: ;

[0026] in, Indicates the first Node and The weights of dynamic edges between nodes, represents the bandwidth parameter, and Indicates the first Node and The feature vector of the node.

[0027] Preferably, performing spatiotemporal feature extraction on the dynamic graph according to a graph neural network to obtain the spatiotemporal features of the dynamic graph includes:

[0028] Perform spatial graph convolution on the dynamic graph according to the adjacency matrix to obtain the spatial features of each node in the dynamic graph, wherein the calculation formula of the spatial features is as follows: ;

[0029] in, Indicates the The spatial characteristics of the layer represent the aggregation results of brain area information. represents the normalized adjacency matrix, represents the degree matrix of the normalized adjacency matrix, Indicates the layer of trainable spatial convolution kernels, Indicates the The input features of the layer represent the feature vectors of each node in the dynamic graph;

[0030] The time feature of each node in the dynamic graph is extracted based on the LSTM model to obtain the time feature of each node in the dynamic graph. The calculation formula of the spatial feature is as follows: ;

[0031] in, Indicates the The temporal features of the LSTM layer, Indicates the The memory state of the layer LSTM, Indicates the The trainable weight matrix of the layer LSTM;

[0032] The spatial features, temporal features and feature vectors of each node in the dynamic graph are residually connected to obtain the spatiotemporal features of each node in the dynamic graph.

[0033] Preferably, the feature fusion of each node of the dynamic graph is performed using the spatiotemporal features according to the attention mechanism to obtain the fusion features of the dynamic graph, including:

[0034] The attention weight of each EEG lead node in the dynamic graph is calculated according to the fusion features of each EEG lead node in the dynamic graph, wherein the calculation formula of the attention weight of the EEG lead node is as follows: ;

[0035] in, Indicates the The attention weight of each EEG lead node, represents the Sigmoid activation function, A trainable projection matrix representing the EEG lead nodes, represents the bias term of the EEG lead node, Represents the spatiotemporal characteristics of EEG lead nodes;

[0036] The attention weight of the ear temperature sensor node in the dynamic graph is calculated according to the fusion feature of the ear temperature sensor node in the dynamic graph, wherein the calculation formula of the attention weight of the ear temperature sensor node is as follows: ;

[0037] in, represents the attention weight of the ear temperature sensor node, Represents the trainable projection matrix of the ear temperature sensor node, represents the bias term of the ear temperature sensor node, Represents the spatiotemporal characteristics of the ear temperature sensor node;

[0038] The spatiotemporal features of each node in the dynamic graph are fused according to the attention weight of each EEG lead node in the dynamic graph and the attention weight of the ear temperature sensor node in the dynamic graph to obtain a fusion feature of the dynamic graph, wherein the calculation formula of the fusion feature is as follows: ;

[0039] in, Represents the fusion features of the dynamic graph, Indicates each EEG lead section in the dynamic graph, Represents the ear temperature sensor node in the dynamic graph.

[0040] Preferably, performing anesthesia depth regression and anesthesia state classification on the target patient according to the fusion features of the dynamic image to obtain the CSI index and anesthesia state level of the target patient includes:

[0041] Anesthesia depth regression is performed on the fusion features of the dynamic image to obtain the CSI index of the target patient, wherein the calculation formula of the CSI index is as follows: ;

[0042] in, represents the CSI index, represents the weight matrix for anesthesia depth regression, represents the bias term for the regression of anesthesia depth;

[0043] The fusion features of the dynamic image are used to classify the anesthesia state to obtain the anesthesia state level of the target patient, wherein the probability of the anesthesia state level is calculated as follows: ;

[0044] in, represents the probability of the level of anesthesia, represents the weight matrix for the classification of anesthetic states, Represents the bias term for the classification of anesthetic status.

[0045] Preferably, the CSI index and anesthesia state level of the target patient are fed back to the anesthesia infusion pump in real time, and the anesthetic infusion rate is dynamically adjusted according to the fuzzy PID controller, including:

[0046] defining a state space according to the CSI index and the anesthetic state level of the target patient;

[0047] An action space is defined based on an adjustment of the proportional gain, the integral gain, the differential gain, and the weight factor of the fuzzy rule of the fuzzy PID controller;

[0048] The reward function is determined according to the CSI index and the anesthesia state level of the target patient, wherein the calculation formula of the reward function is as follows: ;

[0049] in, represents the reward function, and Indicates the preset balance weight;

[0050] An intelligent agent is constructed according to the state space, the action space, and the reward function, and the intelligent agent is trained according to deep reinforcement learning.

[0051] The beneficial effects of the present invention are as follows: (1) The present invention can more comprehensively and accurately evaluate the patient's anesthesia depth by combining EEG signals and ear temperature monitoring data. EEG signals can reflect the brain's electrical activity, while ear temperature reflects the body's overall physiological state, especially the effect of anesthetic drugs on body temperature. Comprehensive analysis of these two signals can effectively improve the accuracy of anesthesia state assessment and avoid misjudgment caused by single signal monitoring in traditional methods. The EEG and ear temperature signals are dynamically correlated and modeled through the correlation modeling unit to extract the PAC feature matrix between the two. This modeling method can reveal the dynamic interaction between EEG signals and ear temperature, so that anesthesia assessment can reflect the changes in the patient's physiological state in real time, and improve the sensitivity and response speed of the assessment; (2) The present invention extracts spatiotemporal features through graph neural networks, extracts the spatiotemporal features of each node from the dynamic graph, and can effectively capture the spatial structure information and time change laws of EEG signals and ear temperature signals, and make full use of this information to conduct in-depth anesthesia state analysis, thereby improving the accuracy of anesthesia assessment, and by paying attention to The force mechanism performs feature fusion on spatiotemporal features and can dynamically adjust the fusion strategy according to the spatiotemporal features of each node, so that different types of features can be fully utilized. The attention mechanism can automatically select the most relevant features, thereby improving the accuracy of anesthesia assessment and ensuring that the patient's current anesthesia state can be accurately judged during anesthesia depth regression and state classification; (3) The present invention generates the patient's CSI and anesthesia state level in real time through the anesthesia assessment unit, and feeds this information back to the anesthesia control unit. Using the fuzzy PID controller, the system can dynamically adjust the infusion rate of anesthetics according to the patient's anesthesia state. Such an intelligent feedback mechanism can optimize the anesthesia process in real time, ensure that the patient maintains the best anesthesia state during the operation, avoid over-anesthesia or under-anesthesia, and can accurately evaluate and adjust the anesthesia state in real time, reduce errors in manual operation, and enhance the safety of the anesthesia process. Especially in complex operations, it may be difficult for anesthesiologists to continuously monitor all physiological parameters, and the system can provide real-time feedback in multiple dimensions, thereby reducing the occurrence of anesthesia accidents and ensuring the safety of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a schematic diagram of the system architecture of the overall system in an embodiment of the present invention.

[0053] Figure numerals: 1. Data acquisition unit; 2. Association modeling unit; 3. Feature extraction unit; 4. Feature fusion unit; 5. Anesthesia assessment unit; 6. Anesthesia control unit. DETAILED DESCRIPTION

[0054] Example 1, as Figure 1 As shown, the present invention proposes an intraoperative patient anesthesia status assessment system based on EEG signals and ear temperature monitoring, comprising:

[0055] Data acquisition unit 1, which is used to obtain the EEG signal and external auditory canal temperature of the target patient based on the EEG lead and the ear temperature sensor, and synchronize the EEG signal and external auditory canal temperature signal of the target patient according to the synchronous clock model;

[0056] The correlation modeling unit 2 is used to perform dynamic correlation modeling on the target patient according to the EEG signal and the external auditory canal temperature signal to obtain a PAC feature matrix between the EEG signal and the external auditory canal temperature signal;

[0057] Feature extraction unit 3, which is used to construct a dynamic graph based on each channel of the EEG lead, the ear temperature sensor, and the PAC feature matrix, and extract spatiotemporal features of the dynamic graph based on the graph neural network to obtain the spatiotemporal features of each node of the dynamic graph;

[0058] The feature fusion unit 4 is used to fuse the features of each node of the dynamic graph through the spatiotemporal features according to the attention mechanism to obtain the fusion features of the dynamic graph;

[0059] Anesthesia assessment unit 5, which is used to perform anesthesia depth regression and anesthesia state classification on the target patient based on the fusion features of the dynamic image, so as to obtain the CSI index and anesthesia state level of the target patient;

[0060] The anesthesia control unit 6 is used to feed back the CSI index and anesthesia state level of the target patient to the anesthesia infusion pump in real time, and dynamically adjust the anesthetic infusion rate according to the fuzzy PID controller.

[0061] In the present invention, EEG is a technology that measures and records the electrical activity of the cerebral cortex through electrodes. EEG leads refer to electrodes placed in different positions for collecting brain electrical signals. Through these electrodes, the patient's brain wave activity can be monitored, which is very important for assessing the depth of anesthesia. The ear temperature sensor is used to measure the temperature of the external auditory canal. The temperature of the external auditory canal is considered to be an effective indicator of human body temperature and has a certain correlation with the anesthesia state in the body. During anesthesia, changes in body temperature may reflect the patient's anesthesia depth. Therefore, ear temperature monitoring can serve as an important data source for anesthesia state assessment. The synchronous clock model is used to ensure that the data collected from different sensors can be aligned in time. PAC refers to the coupling relationship between the phase and amplitude of the EEG signal in the frequency domain. The PAC feature matrix is ​​a feature matrix extracted between the EEG signal and the ear temperature signal, representing the phase-amplitude coupling relationship between them. This feature helps to discover the correlation between the depth of anesthesia and the EEG signal and body temperature, providing a basis for anesthesia assessment. A dynamic graph is a graph structure used to represent spatiotemporal relationships that change over time. In this system, dynamic graphs are used to express how the interactions between different sensors (EEG, ear temperature) change over time. Each node represents a sensor or a specific signal channel, while the connections between nodes represent the correlation between signals. Graph neural networks are a type of neural network based on graph structures that can process graph data with nodes and edges. In this system, graph neural networks are used to extract spatiotemporal features from dynamic graphs and analyze the changing characteristics of each node (i.e., EEG leads or ear temperature sensors) in the temporal and spatial dimensions, which helps to explore the complex dependencies and interactions between signals. Spatiotemporal features refer to the characteristics of signals in time and space. Temporal features reflect the law of signal change over time, while spatial features represent the relationships between different sensors. The attention mechanism is a technology that mimics the human attention concentration process. It emphasizes important information and ignores irrelevant information by assigning different weights to different information (or nodes). CSI is an index used to represent the depth of anesthesia, usually calculated based on EEG data. It assesses the patient's state of consciousness by quantifying the characteristics of EEG signals (such as fluctuation frequency, amplitude, etc.). A lower CSI index usually indicates a deeper state of anesthesia. The PID controller is a feedback control algorithm commonly used in automatic control systems. The fuzzy PID controller combines fuzzy logic with PID control and can effectively adjust system parameters under uncertain or ambiguous conditions. It is used to dynamically adjust the infusion rate of anesthetic drugs according to the patient's anesthesia state to ensure that the anesthesia depth remains within an appropriate range.

[0062] In the second embodiment, the present invention proposes an intraoperative patient anesthesia status assessment system based on EEG signals and ear temperature monitoring. Compared with the first embodiment, this embodiment further includes: the expression of the synchronous clock model is as follows: ;

[0063] in, Indicates the signal time of synchronization, represents the acquisition time of EEG signals, represents the delayed acquisition time of the external auditory canal temperature signal, and , represents a Gaussian distribution, Indicates the standard deviation of the delay.

[0064] In an optional embodiment, dynamic correlation modeling is performed on the target patient based on the EEG signal and the external auditory canal temperature signal to obtain a PAC feature matrix between the EEG signal and the external auditory canal temperature signal, including:

[0065] The frequency band signals of the EEG signal are obtained by using a bandpass filter, wherein the frequency band signals include Wave, Wave, wave and Wave, perform Hilbert transform on the signal of each frequency band to obtain the phase of the signal of each frequency band;

[0066] The low-frequency amplitude of the external auditory canal temperature signal is calculated based on short-time Fourier transform;

[0067] The coupling index between the phase of each frequency band signal and the low-frequency amplitude of the external auditory canal temperature signal is calculated based on PAC;

[0068] The PAC feature matrix between the EEG signal and the external auditory canal temperature signal is obtained according to the coupling index between the phase of each frequency band signal and the low-frequency amplitude of the external auditory canal temperature signal.

[0069] It should be noted that a bandpass filter is an electronic filter that allows signals within a certain frequency range to pass through, while blocking signals below and above this range. In electroencephalogram (EEG) signal processing, bandpass filters are often used to extract signals in a specific frequency range; Hilbert transform is a mathematical transform that is often used in signal processing. Through Hilbert transform, a complex signal can be generated from a real signal, where the real part is the original signal and the imaginary part is its phase information; short-time Fourier transform is a method of time-frequency analysis of signals. It divides the signal into small segments and then performs Fourier transform on each segment to obtain the time-frequency spectrum of the signal; low-frequency amplitude refers to the amplitude of the signal in the low-frequency range (usually 0.5-4 Hz). For the external auditory canal temperature signal, the low-frequency amplitude represents the intensity of the low-frequency components of the temperature signal, which are usually related to physiological processes and anesthetic status.

[0070] In an optional embodiment, the coupling index is calculated as follows: ;

[0071] in, represents the phase-amplitude coupling index, represents the frequency band combination of phase-amplitude coupling, Indicates the phase of each frequency band of the EEG signal, represents the amplitude of the external auditory canal temperature, Indicates the total number of samples.

[0072] In an optional embodiment, a dynamic graph is constructed based on each channel of the EEG lead, the ear temperature sensor, and the PAC feature matrix, including:

[0073] Each channel of the EEG lead and the ear temperature sensor is used as a node of a dynamic graph, wherein the nodes of the dynamic graph include an EEG lead node and an ear temperature sensor node;

[0074] The low-frequency amplitude of the external auditory canal temperature signal and the PAC feature matrix are used as the feature vectors of each node in the dynamic graph;

[0075] Calculate the weight of each dynamic edge in the dynamic graph according to the eigenvector of each node in the dynamic graph, and obtain the adjacency matrix of the dynamic graph according to the weight of each dynamic edge in the dynamic graph;

[0076] The weights of each dynamic edge in the dynamic graph are updated according to the preset graph structure update frequency.

[0077] It should be noted that in a dynamic graph, edges are connections that represent relationships between nodes. Dynamic edges mean that the strength or existence of these connections changes over time. It can reflect how the relationship between EEG leads and ear temperature sensors changes over time.

[0078] In an optional embodiment, the calculation formula of the dynamic edge weight is as follows: ;

[0079] in, Indicates the first Node and The weights of dynamic edges between nodes, represents the bandwidth parameter, and Indicates the first Node and The feature vector of the node.

[0080] In an optional embodiment, spatiotemporal feature extraction is performed on the dynamic graph based on a graph neural network to obtain spatiotemporal features of the dynamic graph, including:

[0081] Perform spatial graph convolution on the dynamic graph according to the adjacency matrix to obtain the spatial features of each node in the dynamic graph. The calculation formula of the spatial features is as follows: ;

[0082] in, Indicates the The spatial characteristics of the layer represent the aggregation results of brain area information. represents the normalized adjacency matrix, represents the degree matrix of the normalized adjacency matrix, Indicates the layer of trainable spatial convolution kernels, Indicates the The input features of the layer represent the feature vectors of each node in the dynamic graph;

[0083] The time feature of each node in the dynamic graph is extracted based on the LSTM model to obtain the time feature of each node in the dynamic graph. The calculation formula of the spatial feature is as follows: ;

[0084] in, Indicates the The temporal features of the LSTM layer, Indicates the The memory state of the layer LSTM, Indicates the The trainable weight matrix of the layer LSTM;

[0085] The spatial features, temporal features and feature vectors of each node in the dynamic graph are residually connected to obtain the spatiotemporal features of each node in the dynamic graph.

[0086] It should be noted that residual connection is a structure in a neural network that transmits input information directly to the deep layer of the network through skip connections, thereby avoiding the gradient vanishing problem and improving training efficiency.

[0087] In an optional embodiment, the nodes of the dynamic graph are fused with features based on the spatiotemporal features according to the attention mechanism to obtain fused features of the dynamic graph, including:

[0088] The attention weight of each EEG lead node in the dynamic graph is calculated based on the fusion features of each EEG lead node in the dynamic graph. The calculation formula of the attention weight of the EEG lead node is as follows: ;

[0089] in, Indicates the The attention weight of each EEG lead node, represents the Sigmoid activation function, A trainable projection matrix representing the EEG lead nodes, represents the bias term of the EEG lead node, Represents the spatiotemporal characteristics of EEG lead nodes;

[0090] The attention weight of the ear temperature sensor node in the dynamic graph is calculated based on the fusion features of the ear temperature sensor node in the dynamic graph. The calculation formula of the attention weight of the ear temperature sensor node is as follows: ;

[0091] in, represents the attention weight of the ear temperature sensor node, Represents the trainable projection matrix of the ear temperature sensor node, represents the bias term of the ear temperature sensor node, Represents the spatiotemporal characteristics of the ear temperature sensor node;

[0092] According to the attention weight of each EEG lead node in the dynamic graph and the attention weight of the ear temperature sensor node in the dynamic graph, the spatiotemporal features of each node in the dynamic graph are fused to obtain the fusion feature of the dynamic graph. The calculation formula of the fusion feature is as follows: ;

[0093] in, Represents the fusion features of the dynamic graph, Indicates each EEG lead section in the dynamic graph, Represents the ear temperature sensor node in the dynamic graph.

[0094] In an optional embodiment, anesthesia depth regression and anesthesia state classification are performed on the target patient based on the fusion features of the dynamic image to obtain the CSI index and anesthesia state level of the target patient, including:

[0095] The anesthesia depth is regressed on the fusion features of the dynamic image to obtain the CSI index of the target patient. The calculation formula of the CSI index is as follows: ;

[0096] in, represents the CSI index, represents the weight matrix for anesthesia depth regression, represents the bias term for the regression of anesthesia depth;

[0097] The fusion features of the dynamic image are used to classify the anesthesia state to obtain the anesthesia state level of the target patient. The calculation formula of the anesthesia state level probability is as follows: ;

[0098] in, represents the probability of the level of anesthesia, represents the weight matrix for the classification of anesthetic states, Represents the bias term for the classification of anesthetic status.

[0099] In an optional embodiment, the CSI index and anesthesia state level of the target patient are fed back to the anesthesia infusion pump in real time, and the anesthetic infusion rate is dynamically adjusted according to the fuzzy PID controller, including:

[0100] The state space is defined according to the CSI index and anesthetic state level of the target patient;

[0101] An action space is defined based on an adjustment of the proportional gain, the integral gain, the differential gain, and the weight factor of the fuzzy rule of the fuzzy PID controller;

[0102] The reward function is determined according to the CSI index and anesthesia level of the target patient, where the calculation formula of the reward function is as follows: ;

[0103] in, represents the reward function, and Indicates the preset balance weight;

[0104] Construct an intelligent agent based on the state space, action space and reward function, and train the intelligent agent based on deep reinforcement learning.

[0105] It should be noted that state space refers to the set of all possible system states in reinforcement learning or control systems; action space refers to the set of all possible actions that the agent can take in each state; reward function is the standard used to evaluate the quality of the agent's behavior in reinforcement learning; the agent is a system that can perceive the environment and take actions to achieve its goals. In reinforcement learning, the agent learns the optimal strategy through interaction with the environment. The agent here perceives the environment through the state space (CSI index and anesthesia level of the target patient) and controls it through the action space (adjusting the parameters of the PID controller), and finally evaluates the effect of the action through the reward function; deep reinforcement learning is a method that combines deep learning and reinforcement learning to train agents to learn optimal strategies in complex environments. Through deep neural networks, agents can process and learn high-dimensional, complex state spaces and action spaces. During the training process, the agent conducts trial and error according to the reward function and gradually learns to optimize its own behavior.

[0106] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. An intraoperative patient anesthesia status assessment system based on EEG signals and ear temperature monitoring, characterized by: include: A data acquisition unit (1), the data acquisition unit (1) is used to obtain an EEG signal and an external auditory canal temperature of a target patient based on an EEG lead and an ear temperature sensor, and synchronize the EEG signal and the external auditory canal temperature signal of the target patient based on a synchronous clock model; An association modeling unit (2), the association modeling unit (2) is used to perform dynamic association modeling on the target patient according to the EEG signal and the external auditory canal temperature signal, so as to obtain a PAC feature matrix between the EEG signal and the external auditory canal temperature signal; A feature extraction unit (3), the feature extraction unit (3) is used to construct a dynamic graph based on each channel of the EEG lead, the ear temperature sensor and the PAC feature matrix, and perform spatiotemporal feature extraction on the dynamic graph based on a graph neural network to obtain spatiotemporal features of each node of the dynamic graph; A feature fusion unit (4), the feature fusion unit (4) is used to fuse the features of each node of the dynamic graph through the spatiotemporal features according to the attention mechanism to obtain a fusion feature of the dynamic graph; An anesthesia assessment unit (5), the anesthesia assessment unit (5) is used to perform anesthesia depth regression and anesthesia state classification on the target patient according to the fusion characteristics of the dynamic image, so as to obtain the CSI index and anesthesia state level of the target patient; An anesthesia control unit (6), the anesthesia control unit (6) is used to feed back the CSI index and anesthesia state level of the target patient to the anesthesia infusion pump in real time, and dynamically adjust the anesthetic infusion rate according to the fuzzy PID controller; The nodes of the dynamic graph are subjected to feature fusion by the spatiotemporal features according to the attention mechanism to obtain fusion features of the dynamic graph, including: The attention weight of each EEG lead node in the dynamic graph is calculated according to the fusion features of each EEG lead node in the dynamic graph, wherein the calculation formula of the attention weight of the EEG lead node is as follows: ; in, Indicates the The attention weight of each EEG lead node, represents the Sigmoid activation function, A trainable projection matrix representing the EEG lead nodes, represents the bias term of the EEG lead node, Represents the spatiotemporal characteristics of EEG lead nodes; The attention weight of the ear temperature sensor node in the dynamic graph is calculated according to the fusion feature of the ear temperature sensor node in the dynamic graph, wherein the calculation formula of the attention weight of the ear temperature sensor node is as follows: ; in, represents the attention weight of the ear temperature sensor node, Represents the trainable projection matrix of the ear temperature sensor node, represents the bias term of the ear temperature sensor node, Represents the spatiotemporal characteristics of the ear temperature sensor node; The spatiotemporal features of each node in the dynamic graph are fused according to the attention weight of each EEG lead node in the dynamic graph and the attention weight of the ear temperature sensor node in the dynamic graph to obtain a fusion feature of the dynamic graph, wherein the calculation formula of the fusion feature is as follows: ; in, Represents the fusion features of the dynamic graph, Indicates each EEG lead section in the dynamic graph, Represents the ear temperature sensor node in the dynamic graph; Anesthesia depth regression and anesthesia state classification are performed on the target patient according to the fusion features of the dynamic image to obtain the CSI index and anesthesia state level of the target patient, including: Anesthesia depth regression is performed on the fusion features of the dynamic image to obtain the CSI index of the target patient, wherein the calculation formula of the CSI index is as follows: ; in, represents the CSI index, represents the weight matrix for anesthesia depth regression, represents the bias term for the regression of anesthesia depth; The fusion features of the dynamic image are used to classify the anesthesia state to obtain the anesthesia state level of the target patient, wherein the probability of the anesthesia state level is calculated as follows: ; in, represents the probability of the level of anesthesia, represents the weight matrix for the classification of anesthetic states, represents the bias term for the classification of anesthetic status; Feedback the CSI index and anesthesia state level of the target patient to the anesthesia infusion pump in real time, and dynamically adjust the anesthetic infusion rate according to the fuzzy PID controller, including: defining a state space according to the CSI index and the anesthetic state level of the target patient; An action space is defined based on an adjustment of the proportional gain, the integral gain, the differential gain, and the weight factor of the fuzzy rule of the fuzzy PID controller; The reward function is determined according to the CSI index of the target patient and the probability of the anesthesia state level, wherein the calculation formula of the reward function is as follows: ; in, represents the reward function, and Indicates the preset balance weight; An intelligent agent is constructed according to the state space, the action space, and the reward function, and the intelligent agent is trained according to deep reinforcement learning.

2. The intraoperative patient anesthesia status assessment system based on EEG signal and ear temperature monitoring according to claim 1, characterized in that: The expression of the synchronous clock model is as follows: ; in, Indicates the signal time of synchronization, represents the acquisition time of EEG signal, represents the delayed acquisition time of the external auditory canal temperature signal, and , represents a Gaussian distribution, Indicates the standard deviation of the delay.

3. The intraoperative patient anesthesia status assessment system based on EEG signal and ear temperature monitoring according to claim 2, characterized in that: Performing dynamic correlation modeling on the target patient according to the EEG signal and the external auditory canal temperature signal to obtain a PAC feature matrix between the EEG signal and the external auditory canal temperature signal, including: The frequency band signals of the EEG signal are obtained by using a bandpass filter, wherein the frequency band signals include Wave, Wave, wave and wave, performing Hilbert transform on the signals of each frequency band to obtain the phase of the signals of each frequency band; Calculating the low-frequency amplitude of the external auditory canal temperature signal according to short-time Fourier transform; Calculating a coupling index between the phase of each frequency band signal and the low-frequency amplitude of the external auditory canal temperature signal according to the PAC; A PAC characteristic matrix between the EEG signal and the external auditory canal temperature signal is obtained according to a coupling index between the phase of each frequency band signal and the low-frequency amplitude of the external auditory canal temperature signal.

4. The intraoperative patient anesthesia status assessment system based on EEG signal and ear temperature monitoring according to claim 3, characterized in that: The calculation formula of the coupling index is as follows: ; in, represents the phase-amplitude coupling index, represents the frequency band combination of phase-amplitude coupling, Indicates the phase of each frequency band of the EEG signal, represents the amplitude of the external auditory canal temperature, Indicates the total number of samples.

5. The intraoperative patient anesthesia status assessment system based on EEG signal and ear temperature monitoring according to claim 4, characterized in that: Constructing a dynamic graph according to each channel of the EEG lead, the ear temperature sensor, and the PAC feature matrix, including: Using each channel of the EEG lead and the ear temperature sensor as nodes of the dynamic graph, wherein the nodes of the dynamic graph include EEG lead nodes and ear temperature sensor nodes; Using the low-frequency amplitude of the external auditory canal temperature signal and the PAC feature matrix as the feature vectors of each node of the dynamic graph; Calculating the weight of each dynamic edge in the dynamic graph according to the feature vector of each node in the dynamic graph, and obtaining the adjacency matrix of the dynamic graph according to the weight of each dynamic edge in the dynamic graph; The weights of each dynamic edge in the dynamic graph are updated according to a preset graph structure update frequency.

6. The intraoperative patient anesthesia status assessment system based on EEG signal and ear temperature monitoring according to claim 5, characterized in that: The calculation formula of the dynamic edge weight is as follows: ; in, Indicates the first Node and The weights of dynamic edges between nodes, represents the bandwidth parameter, and Indicates the first Node and The feature vector of the node.

7. The intraoperative patient anesthesia status assessment system based on EEG signal and ear temperature monitoring according to claim 6, characterized in that: Performing spatiotemporal feature extraction on the dynamic graph according to a graph neural network to obtain spatiotemporal features of the dynamic graph includes: Perform spatial graph convolution on the dynamic graph according to the adjacency matrix to obtain the spatial features of each node in the dynamic graph, wherein the calculation formula of the spatial features is as follows: ; in, Indicates the The spatial characteristics of the layer represent the aggregation results of brain area information. represents the normalized adjacency matrix, represents the degree matrix of the normalized adjacency matrix, Indicates the layer of trainable spatial convolution kernels, Indicates the The input features of the layer represent the feature vectors of each node in the dynamic graph; The time feature of each node in the dynamic graph is extracted based on the LSTM model to obtain the time feature of each node in the dynamic graph. The calculation formula of the spatial feature is as follows: ； in, Indicates the The temporal features of the LSTM layer, Indicates the The memory state of the layer LSTM, Indicates the The trainable weight matrix of the layer LSTM; The spatial features, temporal features and feature vectors of each node in the dynamic graph are residually connected to obtain the spatiotemporal features of each node in the dynamic graph.

Images