Patient anesthesia awakening state monitoring system based on artificial intelligence

Abstract

The invention discloses a patient anesthesia awakening state monitoring system based on artificial intelligence. The system comprises a data acquisition module, a data preprocessing module, an anesthesia awakening abnormity detection module, an anesthesia awakening monitoring module, a hyper-parameter optimization module and a patient anesthesia awakening management module. The invention relates to the technical field of anesthesia wake-up data processing, in particular to a patient anesthesia wake-up state monitoring system based on artificial intelligence, and innovatively proposes the scheme that abnormal conditions in the anesthesia process are preferentially recognized, it is ensured that the wake-up state of a patient can be accurately divided in different anesthesia stages, and the complete wake-up time is accurately predicted; a comprehensive density calculation method combining local and global densities and a sub-tree merging strategy are introduced, so that accurate detection and classification of anesthesia awakening abnormal data are realized; through an excellent selection factor, a particle random divergence strategy and a low-fitness particle individual optimization method, the global optimal solution search capability is enhanced, so that an optimal parameter combination of a monitoring model is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of anesthesia recovery data processing, and specifically refers to a patient anesthesia recovery state monitoring system based on artificial intelligence. Background Art

[0002] During the surgical anesthesia process, accurately monitoring the patient's anesthesia recovery state is crucial to ensure that the patient can safely and timely recover from the anesthesia state. Traditional monitoring methods may have certain limitations. To address this challenge, a patient anesthesia recovery state monitoring system based on artificial intelligence has emerged. This system uses artificial intelligence algorithms for analysis, can more accurately judge the patient's recovery progress and potential risks, give early warnings in a timely manner and take corresponding measures, which helps to improve the safety and quality of the anesthesia recovery process; however, there are technical problems of lagging risk identification in the traditional anesthesia recovery management system; there are technical problems in the clustering algorithm of the existing anesthesia recovery abnormal detection model, such as being sensitive to noise points and outliers, having poor adaptability to complex data distributions, and lacking comprehensive consideration of global and local information, resulting in inaccurate output results of the prediction model; there are technical problems in the traditional monitoring model where the parameter settings are improper and the parameter optimization algorithm has weak global optimal solution acquisition ability, resulting in inaccurate final monitoring results. Summary of the Invention

[0003] In view of the above situation, to overcome the defects of the prior art, the present invention provides a patient anesthesia awakening state monitoring system based on artificial intelligence. Aiming at the technical problem of lagging risk identification in the traditional anesthesia awakening management system, this solution innovatively proposes to prioritize the identification of abnormal situations during anesthesia and feedback the abnormal detection results to the anesthesia awakening monitoring module, effectively avoiding potential safety hazards caused by reaction lag, ensuring accurate classification of the patient's awakening state at different anesthesia stages, and accurately predicting the complete awakening time. The system significantly improves the safety and real-time performance during anesthesia, optimizes the entire anesthesia awakening management process, reduces errors, shortens the patient's anesthesia recovery time, and thus improves the work efficiency of medical staff and the overall experience of patients. Aiming at the technical problems existing in the clustering algorithm in the existing anesthesia awakening abnormal detection model, such as being sensitive to noise points and outliers, having poor adaptability to complex data distributions, and lacking comprehensive consideration of global and local information, resulting in inaccurate output results of the prediction model, this solution innovatively introduces a comprehensive density calculation method combining local and global densities and a subtree merging strategy, improving the robustness of the clustering algorithm and its adaptability to complex data distributions, significantly enhancing the accuracy and reliability of the monitoring model, and thus achieving precise detection and classification of anesthesia awakening abnormal data, further supporting the accuracy and safety of patient anesthesia awakening state monitoring. Aiming at the technical problems existing in the traditional monitoring model, such as improper parameter settings and weak global optimal solution acquisition ability of the parameter optimization algorithm, resulting in inaccurate final monitoring results, this solution enhances the search for the global optimal solution and improves the robustness of the algorithm through excellent selection factors, particle random divergence strategies, and methods for optimizing particle individuals with low fitness, thereby obtaining the optimal parameter values and solving the problem of inaccurate output results of the final monitoring model.

[0004] The technical solution adopted by the present invention is as follows: The patient anesthesia awakening state monitoring system based on artificial intelligence provided by the present invention includes a data acquisition module, a data preprocessing module, an anesthesia awakening abnormal detection module, an anesthesia awakening monitoring module, a hyperparameter optimization module, and a patient anesthesia awakening management module;

[0005] The data acquisition module obtains the original data for anesthesia awakening state monitoring by collecting data from the hospital internal database;

[0006] The data preprocessing module optimizes the original data for anesthesia awakening state monitoring to obtain the preliminary data for anesthesia awakening state monitoring;

[0007] The anesthesia awakening abnormal detection module establishes an anesthesia awakening abnormal detection model through a clustering algorithm, inputs the data into the anesthesia awakening abnormal detection model, and real-time detects abnormal situations of the patient during anesthesia to obtain the anesthesia awakening abnormal detection results;

[0008] The anesthesia recovery monitoring module monitors the patient's anesthesia status and predicts the complete recovery time by establishing an anesthesia recovery monitoring model and training the model, thereby obtaining the anesthesia recovery monitoring results of the patient.

[0009] The hyperparameter optimization module obtains the optimal parameter combination of the anesthesia recovery monitoring model by improving the particle swarm optimization algorithm.

[0010] Based on the output results of the anesthesia recovery anomaly detection module and the anesthesia recovery monitoring module, the patient anesthesia recovery management module provides medical staff with information on anesthesia anomalies, the current anesthesia recovery status, and the patient's complete recovery time during the anesthesia process.

[0011] Further, the data acquisition module specifically collects the original data of anesthesia recovery status monitoring from the hospital's internal database; the original data of anesthesia recovery status monitoring includes historical anesthesia recovery anomaly detection data, real-time anesthesia recovery anomaly detection data, historical anesthesia recovery monitoring data, and real-time anesthesia recovery monitoring data.

[0012] Further, the data preprocessing module specifically performs data cleaning, data synchronization, rule engine screening, standardization processing, data segmentation, and feature selection on the original data of anesthesia recovery status monitoring to obtain preliminary data of anesthesia recovery status monitoring; data cleaning processes missing values, outliers, and duplicate values; data synchronization aligns the timestamps of all data; rule engine screening defines the normal range of each physiological signal according to medical standards and marks data that does not conform to the normal range as potentially abnormal; standardization processing standardizes the data based on the maximum-minimum normalization method; data segmentation processes the time window of the data in segments; feature selection uses the correlation analysis method to screen out features related to anesthesia recovery anomaly detection and anesthesia recovery monitoring.

[0013] Further, the anesthesia recovery anomaly detection module includes establishing an anesthesia recovery anomaly detection model and real-time detection of anesthesia recovery anomalies; specifically, it includes the following steps:

[0014] Establishing an anesthesia recovery anomaly detection model specifically includes the following steps:

[0015] Calculating the local density of sample data points, where the sample data points are the data values of historical anesthesia recovery anomaly detection data and real-time anesthesia recovery anomaly detection data, and the formula used is as follows:

[0016] ;

[0017] In the formula, represents the local density function of the sample data point, represents the i-th sample data point, represents the j-th sample data point, represents the sample data point of the set of K nearest neighbors, where K represents the number of nearest neighbors, represents the sample data point and the sample data point of the Euclidean distance, represents the adjustment coefficient of the local density, represents the sample data point of the local density;

[0018] The calculation of the global density of the sample data point uses the following formula:

[0019] ;

[0020] In the formula, represents the global density function of the sample data point, N represents the total number of sample data points, represents traversing all sample points other than , represents the sample data point of the local density, represents the sample data point of the global density;

[0021] The calculation of the comprehensive density of the sample data point is specifically to use the local density and global density of the sample data point to comprehensively evaluate the importance of the sample point; the formula used is as follows:

[0022] ;

[0023] In the formula, represents the comprehensive density function of the sample data point, represents the weight parameter that balances the local and global densities in the comprehensive density calculation, represents the sample data point of the comprehensive density;

[0024] Determine the density peak point. Specifically, if meets the density peak point selection condition formula, then mark as the density peak point; the formula used is as follows:

[0025] ;

[0026] In the formula, represents the average value of the distances between all sample point pairs, represents the standard deviation of the distances between all sample point pairs, and represent the weight parameters that control the threshold, denotes the distance from the i-th sample data point to its K-th nearest neighbor sample point, denotes the sample data point comprehensive density;

[0027] Cluster label assignment, specifically, assign a unique cluster label to each density peak point. For each non-density peak point, select the nearest neighbor point with a density higher than the non-density peak point as the nearest parent node , and the non-density peak point inherits the cluster label;

[0028] Construct a density peak tree. Specifically, each node in the tree corresponds to a sample data point, and the path between nodes in the tree represents the similarity relationship between sample data points. The root node of the tree is the density peak point, and the ordinary nodes of the tree are non-density peak points; it includes the following steps:

[0029] Calculate the relationship attenuation value between nodes, and the formula used is as follows:

[0030] ;

[0031] In the formula, denotes the relationship attenuation value between node and node , denotes the regularization term parameter;

[0032] Calculate the node retention value, and the formula used is as follows:

[0033] ;

[0034] ;

[0035] In the formula, denotes the retention value of node , denotes the importance weight of the root node, denotes the attenuation factor of the node path, denotes node and the Euclidean distance between them, m represents the number of nodes on the path, and denote two adjacent nodes on the path, denotes node and the root node the path distance between them, denotes from node to the root node the set of all nodes on the path, denotes node and node Attenuation value of the relationship

[0036] Calculate the similarity value between subtrees, and the formula used is as follows:

[0037] ;

[0038] In the formula, represents the similarity value between subtrees, represents the retention value of node ;

[0039] Calculate the merging value between subtrees, and the formula used is as follows:

[0040] ;

[0041] In the formula, represents the merging value of two subtrees, represents the maximum similarity value between two subtrees, represents the average similarity value between two subtrees, represents the adjustment parameter for subtree merging, represents subtree and Euclidean distance between the root nodes;

[0042] Obtain the final clustering result, specifically by gradually merging subtrees from high to low according to the merging value between subtrees until the merging value is lower than the set threshold, and finally forming a complete clustering;

[0043] Real-time detection of abnormal anesthesia awakening, specifically using the historical anesthesia awakening abnormal detection data and real-time anesthesia awakening abnormal detection data in the preliminary data of anesthesia awakening state monitoring as the input data of the anesthesia awakening abnormal detection model to obtain the anesthesia awakening abnormal detection result.

[0044] Furthermore, the anesthesia awakening monitoring module includes establishing an anesthesia awakening monitoring model, training the anesthesia awakening monitoring model, and real-time monitoring of the patient's anesthesia awakening; specifically including the following steps:

[0045] Establish an anesthesia awakening monitoring model, specifically including the following steps:

[0046] Multi-modal feature extraction, and the formula used is as follows:

[0047] ;

[0048] In the formula, represents the unit operation function, represents the hidden state at the i-th time step, represents the input data at the i-th time step, represents the The hidden state at a time step, indicating it is the cell unit state at the time step;

[0049] The calculation of attention weights uses the following formula:

[0050] ;

[0051] In the formula, represents the attention weight value at the i-th time step, represents the attention weight matrix, represents the attention bias term parameter;

[0052] To obtain the task branch shared features, the following formula is used:

[0053] ;

[0054] In the formula, represents the global feature weighted by the attention mechanism, represents the total number of time steps;

[0055] To obtain the output result of the task branch, the specific steps are as follows:

[0056] For patient anesthesia state classification, the following formula is used:

[0057] ;

[0058] In the formula, represents the patient anesthesia state classification result, represents the patient anesthesia state task branch weight matrix, represents the bias term parameter of the patient anesthesia state task branch;

[0059] For patient awakening time prediction, the following formula is used:

[0060] ;

[0061] In the formula, represents the patient anesthesia awakening time prediction result, represents the weight matrix of the awakening time prediction branch, represents the bias term parameter of the awakening time prediction branch;

[0062] Training of the anesthesia awakening monitoring model, specifically using the historical anesthesia awakening monitoring data in the preliminary data of the anesthesia awakening state monitoring to train the anesthesia awakening monitoring model to obtain the trained anesthesia awakening monitoring model;

[0063] Real-time monitoring of the patient's anesthesia recovery, specifically using the real-time anesthesia recovery monitoring data and the anesthesia recovery abnormality detection results in the preliminary anesthesia recovery state monitoring data as the input data of the trained anesthesia recovery monitoring model to obtain the patient's anesthesia recovery monitoring results.

[0064] Further, the hyperparameter optimization module specifically obtains the optimal parameter combination of the anesthesia recovery monitoring model through an improved particle swarm optimization algorithm, including the following steps:

[0065] Initialize parameters, specifically by constructing the initial parameters of the algorithm; the initial parameters of the algorithm include the number of particles N and the maximum number of iterations ;

[0066] Initialize the particle position, and the formula used is as follows:

[0067] ;

[0068] In the formula, represents the initial position of the i-th particle, and represent the lower and upper limits of the particle swarm position respectively;

[0069] Calculate the fitness value, specifically calculate the fitness value f of the particles in the particle swarm i ; Use the performance of the anesthesia recovery monitoring model established based on the individual position as the fitness value of the individual, sort the individuals from the best to the worst according to the fitness value, and obtain the position of the individual with the highest global fitness value ;

[0070] Obtain the excellent selection factor, and the formula used is as follows:

[0071] ;

[0072] In the formula, represents the excellent selection factor at the t-th iteration, represents the maximum value of the excellent selection factor, represents the minimum value of the excellent selection factor, t represents the current iteration number, represents a random number uniformly distributed within the range of ;

[0073] Update the particle velocity, and the formula used is as follows:

[0074] ;

[0075] In the formula, represents the velocity of the i-th particle in the (t + 1)-th iteration, represents the velocity of the i-th particle in the t-th iteration, Denote the position of the $i$-th particle in the $t$-th iteration, Denote the local optimal position of the particle individual, 、 and Denote random numbers in the range of $[0, 1]$, Denote the inertial weight of the particle, Denote the individual learning factor, which is used to control the speed of the particle moving towards the individual optimal position, Denote the swarm learning factor, which is used to control the speed of the particle moving towards the global optimal position, Denote the excellent particle learning factor;

[0076] Update the particle position, and the formula used is as follows:

[0077] ;

[0078] In the formula, Denote the position of the $i$-th particle in the $(t + 1)$-th iteration;

[0079] Obtain the position step factor. Specifically, use the particle random divergence strategy to obtain the particle divergence position step. The formula used is as follows:

[0080] ;

[0081] In the formula, Denote the current iteration position step factor, Denote the standard deviation of the fitness of the current population, Denote the mean of the fitness of the current population, and Both denote random numbers that follow a normal distribution, Denote the parameter that controls the step size;

[0082] Update the local optimal position of the particle individual, and the formula used is as follows:

[0083] ;

[0084] In the formula, Denote the updated local optimal position of the particle individual, Denote the fitness function;

[0085] Optimize the particle individuals with low fitness. Specifically, compare the fitness values of the local optimal positions of all particles, re - sort the particle individuals from the best to the worst according to the fitness values, and select 10%N particle individuals with low fitness values to optimize. The formula used is as follows:

[0086] ;

[0087] ;

[0088] In the formula, represents the position of the particle individual with low fitness after optimization, represents the selected particle individual with low fitness, represents the dynamic optimization control parameter, represents the maximum value of the dynamic optimization control parameter, represents the minimum value of the dynamic optimization control parameter;

[0089] Search determination, specifically, by constructing a search termination condition, the search determination of the optimal particle individual position is carried out to obtain the setting of the optimal particle individual position data;

[0090] The search termination condition includes threshold termination and iteration termination;

[0091] The threshold termination is specifically to set a fitness threshold. When the particle fitness value f i is higher than the fitness threshold, the search is completed;

[0092] The iteration termination specifically means that when the maximum number of iterations is reached, the iteration is terminated and the global optimal position of the particle is obtained;

[0093] The global optimal position of the particle specifically refers to the optimal parameter combination of the anesthesia awakening monitoring model.

[0094] Furthermore, the patient anesthesia awakening management module specifically feeds back the current anesthesia awakening state of the patient and the complete awakening time of the patient to the medical staff according to the patient anesthesia awakening monitoring results, and gives a real-time alarm according to the anesthesia awakening abnormality detection results to ensure that the medical staff can take corresponding measures in time.

[0095] The beneficial effects obtained by the present invention using the above scheme are as follows:

[0096] (1) Aiming at the technical problem of lagging risk identification in the traditional anesthesia awakening management system, this scheme innovatively proposes to prioritize the identification of abnormal situations during anesthesia and feedback the abnormality detection results to the anesthesia awakening monitoring module, effectively avoiding potential safety hazards caused by reaction lag, ensuring accurate classification of the patient's awakening state at different anesthesia stages, accurately predicting the complete awakening time, significantly improving the safety and real-time performance during anesthesia, optimizing the entire anesthesia awakening management process, reducing errors, shortening the patient's anesthesia recovery time, and thus improving the work efficiency of medical staff and the overall experience of patients.

[0097] (2) Aiming at the problems existing in the clustering algorithm of the existing anesthetic awakening abnormality detection model, such as being sensitive to noise points and outliers, having poor adaptability to complex data distributions, and lacking comprehensive consideration of global and local information, which lead to inaccurate output results of the prediction model. This solution innovatively introduces a comprehensive density calculation method combining local and global densities and a subtree merging strategy, improving the robustness of the clustering algorithm and its adaptability to complex data distributions, significantly enhancing the accuracy and reliability of the monitoring model, thus achieving accurate detection and classification of anesthetic awakening abnormal data, and further supporting the accuracy and safety of patient anesthetic awakening state monitoring.

[0098] (3) Aiming at the technical problems existing in the traditional monitoring model, such as improper parameter settings and weak global optimal solution acquisition ability of the parameter optimization algorithm, which lead to inaccurate final monitoring results. This solution enhances the global optimal solution search and improves the robustness of the algorithm by excellent selection factors, particle random divergence strategies, and methods for optimizing particle individuals with low fitness, thereby obtaining the optimal parameter values and solving the problem of inaccurate output results of the final monitoring model. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 It is a schematic diagram of the modules of the patient anesthetic awakening state monitoring system based on artificial intelligence provided by the present invention;

[0100] Figure 2 It is a schematic diagram of the process of the anesthetic awakening abnormality detection module;

[0101] Figure 3 It is a schematic diagram of the process of the anesthetic awakening monitoring module;

[0102] Figure 4 It is a schematic diagram of the process of the hyperparameter optimization module;

[0103] Figure 5 It is a schematic diagram of the process of establishing an anesthetic awakening abnormality detection model in the anesthetic awakening abnormality detection module;

[0104] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0105] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0106] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. These are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention.

[0107] Embodiment 1. Refer to Figure 1 , the patient anesthesia awakening state monitoring system based on artificial intelligence provided by the present invention includes a data acquisition module, a data preprocessing module, an anesthesia awakening abnormality detection module, an anesthesia awakening monitoring module, a hyperparameter optimization module, and a patient anesthesia awakening management module;

[0108] The data acquisition module obtains the original data of anesthesia awakening state monitoring by collecting data from the hospital internal database, and sends the data to the data preprocessing module;

[0109] The data preprocessing module receives the data sent by the data acquisition module, performs data cleaning, data synchronization, rule engine screening, standardization processing, data segmentation, and feature selection on the original data of anesthesia awakening state monitoring, obtains the preliminary data of anesthesia awakening state monitoring, and sends the data to the anesthesia awakening abnormality detection module and the anesthesia awakening monitoring module;

[0110] The anesthesia awakening abnormality detection module receives the data sent by the data preprocessing module, establishes an anesthesia awakening abnormality detection model through a clustering algorithm, inputs the data into the anesthesia awakening abnormality detection model, real-time detects the abnormal conditions of the patient during the anesthesia process, obtains the anesthesia awakening abnormality detection result, and sends the data to the anesthesia awakening monitoring module and the patient anesthesia awakening management module;

[0111] The anesthesia awakening monitoring module receives the data sent by the data preprocessing module, the anesthesia awakening abnormality detection module, and the hyperparameter optimization module, trains the anesthesia awakening monitoring model by establishing the anesthesia awakening monitoring model, realizes the monitoring of the patient's anesthesia state and the prediction of the complete awakening time, obtains the patient anesthesia awakening monitoring result, and sends the data to the patient anesthesia awakening management module;

[0112] The hyperparameter optimization module obtains the optimal parameter combination of the anesthesia awakening monitoring model through an improved particle swarm optimization algorithm, and sends the data to the anesthesia awakening monitoring module;

[0113] The patient anesthesia awakening management module receives the data of the anesthesia awakening abnormality detection module and the anesthesia awakening monitoring module, and based on the output results of the two modules, provides the medical staff with the anesthesia abnormal conditions, the current anesthesia awakening state, and the patient's complete awakening time of the patient during the anesthesia process.

[0114] By performing the above operations, in view of the technical problem of lagging risk identification in the traditional anesthesia recovery management system, this solution innovatively proposes to prioritize the identification of abnormal conditions during anesthesia and feedback the abnormal detection results to the anesthesia recovery monitoring module, effectively avoiding potential safety hazards caused by reaction lag, ensuring the accurate classification of the patient's recovery state at different anesthesia stages, and accurately predicting the complete recovery time. The system significantly improves the safety and real-time performance during anesthesia, optimizes the entire anesthesia recovery management process, reduces errors, shortens the patient's anesthesia recovery time, and thus improves the work efficiency of medical staff and the overall experience of patients.

[0115] Example 2. Refer to Figure 1 , this example is based on the above example. The data acquisition module specifically obtains the original data of anesthesia recovery state monitoring from the hospital's internal database; the original data of anesthesia recovery state monitoring includes historical anesthesia recovery abnormal detection data, real-time anesthesia recovery abnormal detection data, historical anesthesia recovery monitoring data, and real-time anesthesia recovery monitoring data; both the historical anesthesia recovery abnormal detection data and the real-time anesthesia recovery abnormal detection data include physiological data and the change rate of physiological signals; the historical anesthesia recovery abnormal detection data also includes the type of anesthesia recovery abnormal detection; the type of anesthesia recovery abnormal detection is divided into normal, abnormal heart rate, respiratory depression, low blood oxygen, and other abnormalities; the type of anesthesia recovery abnormal detection is used as a data label, which is not considered during clustering and is only used when selecting the cluster label; both the historical anesthesia recovery monitoring data and the real-time anesthesia recovery monitoring data include physiological data and body movement data; the historical anesthesia recovery monitoring data also includes the anesthesia recovery state; the anesthesia recovery state is divided into deep anesthesia, light anesthesia, edge of recovery, and complete recovery; the physiological data includes electroencephalogram data, heart rate data, respiratory rate data, blood oxygen saturation, and blood pressure data.

[0116] Example 3. Refer to Figure 1, this embodiment is based on the above embodiment. The data preprocessing module specifically performs data cleaning, data synchronization, rule engine screening, standardization processing, data segmentation, and feature selection on the original data of anesthesia awakening state monitoring to obtain preliminary data of anesthesia awakening state monitoring. The data cleaning is to handle missing values, outliers, and duplicate values. The data synchronization is to align the timestamps of all data. The rule engine screening is to define the normal range of each physiological signal according to medical standards and mark the data that does not conform to the normal range as potentially abnormal. The standardization processing is to standardize the data based on the maximum-minimum normalization method. The data segmentation is to segment the time window of the data. The acquisition data time window of the historical anesthesia awakening anomaly detection data and the real-time anesthesia awakening anomaly detection data is set to 40 seconds, and the window is updated every 1 second. The acquisition data time window of the historical anesthesia awakening monitoring and the real-time anesthesia awakening monitoring is 2 minutes, and the window is updated every 10 seconds. The feature selection is to use the correlation analysis method to screen out the features related to anesthesia awakening anomaly detection and anesthesia awakening monitoring.

[0117] Embodiment 4, refer to Figure 1 、 Figure 2 and Figure 5 , this embodiment is based on the above embodiment. The anesthesia awakening anomaly detection module includes establishing an anesthesia awakening anomaly detection model and real-time anesthesia awakening anomaly detection, and specifically includes the following steps:

[0118] Establishing an anesthesia awakening anomaly detection model specifically includes the following steps:

[0119] Calculating the local density of sample data points. The sample data points are the data values of historical anesthesia awakening anomaly detection data and real-time anesthesia awakening anomaly detection data. The formula used is as follows:

[0120] ;

[0121] In the formula, represents the local density function of the sample data point, represents the i-th sample data point, represents the j-th sample data point, represents the sample data point 's set of K nearest neighbors. K represents the number of nearest neighbors, represents the sample data point and the sample data point 's Euclidean distance, represents the adjustment coefficient of the local density, represents the local density of the sample data point ;

[0122] Calculating the global density of sample data points. The formula used is as follows:

[0123] ;

[0124] In the formula, represents the global density function of the sample data points, N represents the total number of sample data points, represents traversing all sample points other than , represents the local density of the sample data point ; represents the global density of the sample data point ;

[0125] The comprehensive density calculation of the sample data points is specifically to comprehensively evaluate the importance of the sample points through the local density and global density of the sample data points; the formula used is as follows:

[0126] ;

[0127] In the formula, represents the comprehensive density function of the sample data points, represents the weight parameter for balancing the local and global densities in the comprehensive density calculation, represents the sample data point ;

[0128] Determine the density peak points. Specifically, if meets the density peak point selection condition formula, then mark as the density peak point; the formula used is as follows:

[0129] ;

[0130] In the formula, represents the average value of the distances between all sample point pairs, represents the standard deviation of the distances between all sample point pairs, and represent the weight parameters for controlling the threshold, represents the distance from the i-th sample data point to its K-th nearest neighbor sample point, represents the sample data point ;

[0131] Cluster label assignment. Specifically, assign a unique cluster label to each density peak point. For each non-density peak point, select the nearest neighbor point with a density higher than the non-density peak point as the nearest parent node , and the non-density peak point inherits 's cluster label;

[0132] Construct a density peak tree. Specifically, each node in the tree corresponds to a sample data point, and the path between nodes in the tree represents the similarity relationship between sample data points. The root node of the tree is the density peak point, and the ordinary nodes of the tree are non-density peak points. The steps are as follows:

[0133] Calculate the relationship attenuation value between nodes. The formula used is as follows:

[0134] ;

[0135] In the formula, represents the relationship attenuation value between node and node , represents the regularization term parameter;

[0136] Calculate the node retention value. The formula used is as follows:

[0137] ;

[0138] ;

[0139] In the formula, represents the retention value of node , represents the importance weight of the root node, represents the attenuation factor of the node path, represents node and 's Euclidean distance. m represents the number of nodes on the path, and represent two adjacent nodes on the path, represents node and the root node 's path distance, represents from node to the root node 's path's all node set, represents node and node 's relationship attenuation value;

[0140] Calculate the similarity value between subtrees. The formula used is as follows:

[0141] ;

[0142] In the formula, represents the similarity value between subtrees, represents the retention value of node ;

[0143] Calculate the merge value between subtrees. The formula used is as follows:

[0144] ;

[0145] In the formula, represents the combined value of two subtrees, represents the maximum similarity value between two subtrees, represents the average similarity value between two subtrees, represents the adjustment parameter for subtree combination, represents subtree and represents the Euclidean distance of the root nodes between

[0146] Obtain the final clustering result, specifically, gradually combine subtrees from high to low according to the combined value between subtrees until the combined value is lower than the set threshold, and finally form a complete clustering;

[0147] Real-time detection of abnormal anesthesia awakening, specifically, using the historical anesthesia awakening abnormal detection data and real-time anesthesia awakening abnormal detection data in the preliminary anesthesia awakening state monitoring data as the input data of the anesthesia awakening abnormal detection model to obtain the anesthesia awakening abnormal detection result.

[0148] By performing the above operations, for the problems existing in the clustering algorithm in the existing anesthesia awakening abnormal detection model, such as being sensitive to noise points and outliers, having poor adaptability to complex data distributions, and lacking comprehensive consideration of global and local information, which leads to inaccurate output results of the prediction model. This solution innovatively introduces a comprehensive density calculation method combining local and global densities and a subtree combination strategy, improves the robustness of the clustering algorithm and the adaptability to complex data distributions, significantly improves the accuracy and reliability of the monitoring model, thus realizing the accurate detection and classification of anesthesia awakening abnormal data, and further supporting the accuracy and safety of patient anesthesia awakening state monitoring.

[0149] Example Five, refer to Figure 1 and Figure 3 , this example is based on the above example, and the anesthesia awakening monitoring module includes establishing an anesthesia awakening monitoring model, training the anesthesia awakening monitoring model, and real-time monitoring of patient anesthesia awakening; specifically includes the following steps:

[0150] Establish an anesthesia awakening monitoring model, specifically including the following steps:

[0151] Multi-modal feature extraction, the formula used is as follows:

[0152] ;

[0153] In the formula, represents the unit operation function, represents the hidden state at the i-th time step, represents the input data at the i-th time step, represents the hidden state at the -th time step, and represents the cell unit state at the

[0154] The calculation of the attention weights uses the following formula:

[0155] ;

[0156] In the formula, represents the attention weight value at the i-th time step, represents the attention weight matrix, and represents the attention bias term parameter;

[0157] To obtain the task branch shared features, the following formula is used:

[0158] ;

[0159] In the formula, represents the global feature weighted by the attention mechanism, represents the total number of time steps;

[0160] To obtain the output result of the task branch, the following specific steps are included:

[0161] For the classification of the patient's anesthesia state, the following formula is used:

[0162] ;

[0163] In the formula, represents the classification result of the patient's anesthesia state, represents the weight matrix of the patient's anesthesia state task branch, and represents the bias term parameter of the patient's anesthesia state task branch;

[0164] For the prediction of the patient's awakening time, the following formula is used:

[0165] ;

[0166] In the formula, represents the prediction result of the patient's anesthesia awakening time, represents the weight matrix of the awakening time prediction branch, and represents the bias term parameter of the awakening time prediction branch;

[0167] The training of the anesthesia awakening monitoring model is specifically to train the anesthesia awakening monitoring model using the historical anesthesia awakening monitoring data in the preliminary data of the anesthesia awakening state monitoring to obtain the trained anesthesia awakening monitoring model;

[0168] Real-time monitoring of the patient's anesthesia recovery, specifically using the real-time anesthesia recovery monitoring data and the anesthesia recovery abnormality detection results in the preliminary data of the anesthesia recovery state monitoring as the input data of the trained anesthesia recovery monitoring model to obtain the patient's anesthesia recovery monitoring results, where the patient's anesthesia recovery monitoring results include the current anesthesia recovery state and the patient's complete awakening time.

[0169] Example 6, refer to Figure 1 and Figure 4 , based on the above example, the hyperparameter optimization module specifically obtains the optimal parameter combination of the anesthesia recovery monitoring model through an improved particle swarm optimization algorithm, including the following steps:

[0170] Initialize parameters, specifically by constructing the initial parameters of the algorithm; the initial parameters of the algorithm include the number of particles N and the maximum number of iterations ;

[0171] Initialize the particle position, and the formula used is as follows:

[0172] ;

[0173] In the formula, represents the initial position of the i-th particle, and represent the lower and upper limits of the particle swarm position respectively;

[0174] Calculate the fitness value, specifically calculate the fitness value f of the particles in the particle swarm i ; Use the performance of the anesthesia recovery monitoring model established based on the individual position as the fitness value of the individual, sort the individuals from the best to the worst according to the fitness value, and obtain the position of the individual with the highest global fitness value ;

[0175] Obtain the excellent selection factor, and the formula used is as follows:

[0176] ;

[0177] In the formula, represents the excellent selection factor at the t-th iteration, represents the maximum value of the excellent selection factor, represents the minimum value of the excellent selection factor, t represents the current iteration number, represents a random number uniformly distributed within the range of ;

[0178] Update the particle velocity, and the formula used is as follows:

[0179] ;

[0180] In the formula, represents the velocity of the i-th particle in the (t + 1)-th iteration, represents the velocity of the i-th particle in the t-th iteration, represents the position of the i-th particle in the t-th iteration, represents the local optimal position of the particle individual, , and represent random numbers within the range [0, 1], represents the inertia weight of the particle, represents the individual learning factor, which is used to control the velocity of the particle moving towards the individual optimal position, represents the swarm learning factor, which is used to control the velocity of the particle moving towards the global optimal position, represents the excellent particle learning factor;

[0181] Update the particle position, and the formula used is as follows:

[0182] ;

[0183] In the formula, represents the position of the i-th particle in the (t + 1)-th iteration;

[0184] Obtain the position step factor. Specifically, use the particle random divergence strategy to obtain the particle divergence position step. The formula used is as follows:

[0185] ;

[0186] In the formula, represents the current iteration position step factor, represents the standard deviation of the fitness of the current population, represents the mean of the fitness of the current population, and both represent random numbers that follow a normal distribution, represents the parameter for controlling the step length;

[0187] Update the local optimal position of the particle individual. The formula used is as follows:

[0188] ;

[0189] In the formula, represents the updated local optimal position of the particle individual, represents the fitness function;

[0190] Optimize the particle individuals with low fitness. Specifically, compare the fitness values of the local optimal positions of all particles, re - sort the particle individuals from the best to the worst according to the fitness values, select 10%N particle individuals with low fitness values, and optimize the particle individuals. The formula used is as follows:

[0191] ;

[0192] ;

[0193] In the formula, represents the position of the particle individual with low fitness after optimization, represents the selected particle individual with low fitness, represents the dynamic optimization control parameter, represents the maximum value of the dynamic optimization control parameter, represents the minimum value of the dynamic optimization control parameter;

[0194] Search determination. Specifically, by constructing search termination conditions, conduct search determination for the position of the optimal particle individual to obtain the setting of the optimal particle individual position data;

[0195] The search termination conditions include threshold termination and iteration termination;

[0196] The threshold termination is specifically to set a fitness threshold. When the fitness value f of the particle i is higher than the fitness threshold, the search is completed;

[0197] The iteration termination specifically means that when the maximum number of iterations is reached, the iteration is terminated and the global optimal position of the particle is obtained;

[0198] The global optimal position of the particle specifically refers to the optimal parameter combination of the anesthesia awakening monitoring model.

[0199] By performing the above operations, aiming at the technical problems in the traditional monitoring model such as improper parameter setting and weak ability to obtain the global optimal solution in the parameter optimization algorithm, resulting in inaccurate final monitoring results. This solution enhances the search for the global optimal solution and improves the robustness of the algorithm through excellent selection factors, particle random divergence strategies, and methods for optimizing particle individuals with low fitness, thereby obtaining the optimal parameter values and solving the problem of inaccurate output results of the final monitoring model.

[0200] Example Seven, refer to Figure 1 , this example is based on the above example. The patient anesthesia awakening management module specifically feeds back the current anesthesia awakening state of the patient and the patient's complete awakening time to the medical staff according to the patient anesthesia awakening monitoring results, and gives a real - time alarm according to the anesthesia awakening abnormal detection results to ensure that the medical staff can take corresponding measures in time.

[0201] It should be noted that, in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0202] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0203] The above description of the present invention and its embodiments is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. In general, if those of ordinary skill in the art are inspired by it and, without departing from the purpose of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. The patient anesthesia awakening status monitoring system based on artificial intelligence is characterized by: It includes data acquisition module, data preprocessing module, anesthesia recovery abnormality detection module, anesthesia recovery monitoring module, hyperparameter optimization module and patient anesthesia recovery management module; The data acquisition module acquires raw data of anesthesia recovery status monitoring by collecting data from the hospital's internal database; The data preprocessing module optimizes the raw data of the anesthesia recovery state monitoring to obtain preliminary data of the anesthesia recovery state monitoring; The anesthesia awakening anomaly detection module establishes an anesthesia awakening anomaly detection model by introducing a comprehensive density calculation method combining local and global density and a subtree merging strategy clustering algorithm, and inputs data into the anesthesia awakening anomaly detection model to detect abnormal conditions of patients during anesthesia in real time and obtain an anesthesia awakening anomaly detection result; The anesthesia recovery monitoring module establishes an anesthesia recovery monitoring model through multimodal feature extraction and multi-task branching, performs model training on the anesthesia recovery monitoring model, and inputs the anesthesia recovery abnormality detection result and the anesthesia recovery state monitoring preliminary data into the anesthesia recovery monitoring model to realize the monitoring of the patient's anesthesia state and the prediction of the complete recovery time, and obtain the patient's anesthesia recovery monitoring result; The hyperparameter optimization module improves the particle swarm optimization algorithm by using excellent selection factors, random particle divergence strategy and individual particle optimization method with low fitness to obtain the optimal parameter combination of the anesthesia awakening monitoring model; The patient anesthesia awakening management module provides medical staff with the patient's anesthesia abnormality during anesthesia, the current anesthesia awakening status and the patient's complete awakening time based on the output results of the anesthesia awakening abnormality detection module and the anesthesia awakening monitoring module.

2. The patient anesthesia awakening status monitoring system based on artificial intelligence according to claim 1 is characterized in that: The anesthesia awakening abnormality detection module includes establishing an anesthesia awakening abnormality detection model and anesthesia awakening abnormality real-time detection; specifically includes the following steps: Establishing an anesthesia recovery abnormality detection model includes the following steps: The local density calculation of the sample data points, which are the data values ​​of the historical anesthesia awakening abnormality detection data and the real-time anesthesia awakening abnormality detection data, uses the following formula: ； In the formula, represents the local density function of the sample data points, represents the i-th sample data point, represents the jth sample data point, Represents sample data points The K nearest neighbor set of , K represents the number of nearest neighbors, Represents sample data points and sample data points The Euclidean distance of represents the adjustment coefficient of local density, Represents sample data points The local density of The global density calculation of the sample data points uses the following formula: ； In the formula, represents the global density function of the sample data points, N represents the total number of sample data points, Represents traversal All sample points except , Represents sample data points The local density of Represents sample data points The global density of The comprehensive density calculation of the sample data points is specifically to use the local density and global density of the sample data points to comprehensively evaluate the importance of the sample points; the formula used is as follows: ； In the formula, represents the comprehensive density function of the sample data points, represents the weight parameter for balancing local and global density in the comprehensive density calculation, Represents sample data points The comprehensive density of Determine the density peak point, specifically if If the density peak point selection condition formula is met, then mark is the density peak point; the formula used is as follows: ； In the formula, represents the average distance between all pairs of sample points. represents the standard deviation of the distances between all pairs of sample points, and represents the weight parameter of the control threshold, Represents the distance from the i-th sample data point to its K-th nearest neighbor sample point, Represents sample data points The comprehensive density of Cluster label assignment, specifically assigning a unique cluster label to each density peak point, and for each non-density peak point, selecting the nearest neighbor point with a higher density than the non-density peak point as the nearest parent node , non-density peak point inheritance The cluster labels of Constructing a density peak tree, specifically, each node in the tree corresponds to a sample data point, the paths between the nodes in the tree represent the similarity relationship between the sample data points, the root node of the tree is the density peak point, and the ordinary nodes of the tree are non-density peak points; including the following steps: Calculate the relationship attenuation value between nodes using the following formula: ； In the formula, Representation Node With Node The relationship attenuation value, represents the regularization term parameter; Calculate the node retention value using the following formula: ； ； In the formula, Representation Node The retention value of represents the importance weight of the root node, represents the attenuation factor of the node path, Representation Node and The Euclidean distance, m represents the number of nodes on the path, and Represents two adjacent nodes on the path. Representation Node With the root node The path distance between Represents a slave node To the root node The set of all nodes on the path, Representation Node With Node The attenuation value of the relationship between them; The similarity value between subtrees is calculated using the following formula: ； In the formula, represents the similarity value between subtrees, Representation Node The retention value of Calculate the merge value between subtrees using the following formula: ； In the formula, represents the merged value of two subtrees, represents the maximum similarity between two subtrees, represents the average similarity between two subtrees, represents the tuning parameter for subtree merging, Represents a subtree and The Euclidean distance between the root nodes; The final clustering result is obtained by gradually merging subtrees from high to low according to the merging values ​​between subtrees until the merging value is lower than the set threshold, and finally forming a complete clustering; The real-time detection of anesthesia recovery anomaly is specifically to use the historical anesthesia recovery anomaly detection data and the real-time anesthesia recovery anomaly detection data in the preliminary anesthesia recovery status monitoring data as the input data of the anesthesia recovery anomaly detection model to obtain the anesthesia recovery anomaly detection result.

3. The patient anesthesia awakening status monitoring system based on artificial intelligence according to claim 1 is characterized in that: The anesthesia awakening monitoring module includes establishing an anesthesia awakening monitoring model, training the anesthesia awakening monitoring model and real-time monitoring of patient anesthesia awakening; specifically includes the following steps: Establishing an anesthesia recovery monitoring model includes the following steps: Multimodal feature extraction, the formula used is as follows: ； In the formula, Indicates the unit runs the function, represents the hidden state at the i-th time step, represents the input data of the i-th time step, Indicates that it is The hidden state of time steps, Indicates that it is The cell state at each time step; The attention weight is calculated using the following formula: ； In the formula, represents the attention weight value of the i-th time step, represents the attention weight matrix, represents the attention bias parameter; Get the shared features of task branches. The formula used is as follows: ； In the formula, represents the global features after the weighting of the attention mechanism, Indicates the total number of time steps; Obtain the output results of the task branch, which specifically includes the following steps: The patient's anesthesia status is classified using the following formula: ； In the formula, Indicates the classification result of the patient's anesthesia status, represents the patient anesthesia status task branch weight matrix, The bias parameter representing the patient's anesthesia status task branch; The patient's awakening time is predicted using the following formula: ； In the formula, It indicates the prediction result of the patient's anesthesia awakening time. represents the weight matrix of the wake-up time prediction branch, Represents the bias parameter of the wake-up time prediction branch; Anesthesia awakening monitoring model training, specifically using the historical anesthesia awakening monitoring data in the anesthesia awakening state monitoring preliminary data to train the anesthesia awakening monitoring model to obtain a trained anesthesia awakening monitoring model; The real-time monitoring of the patient's anesthesia awakening includes taking the real-time anesthesia awakening monitoring data in the preliminary anesthesia awakening status monitoring data and the anesthesia awakening abnormality detection result as input data of the trained anesthesia awakening monitoring model to obtain the patient's anesthesia awakening monitoring result.

4. The patient anesthesia awakening status monitoring system based on artificial intelligence according to claim 1 is characterized in that: The hyperparameter optimization module specifically obtains the optimal parameter combination of the anesthesia awakening monitoring model by improving the particle swarm optimization algorithm, including the following steps: Initialization parameters, specifically, by constructing the initial parameters of the algorithm; the initial parameters of the algorithm include the number of particles N and the maximum number of iterations ; Initialize the particle position using the following formula: ； In the formula, represents the initial position of the ith particle, and Respectively represent the lower and upper limits of the particle group position; Calculate the fitness value, specifically calculate the fitness value f of the particles in the particle swarm i ; The performance of the anesthesia awakening monitoring model established based on the individual position is used as the individual fitness value, and the individuals are sorted from best to worst according to the fitness value to obtain the individual position with the highest global fitness value ; To obtain the excellent selection factor, the formula used is as follows: ； In the formula, represents the excellent selection factor for the t-th iteration, represents the maximum value of the excellent selection factor, represents the minimum value of the excellent selection factor, t represents the current number of iterations, Indicates between A random number uniformly distributed in the range; Update particle velocity using the following formula: ； In the formula, represents the velocity of the i-th particle in the t+1th iteration, represents the velocity of the i-th particle in the t-th iteration, represents the position of the i-th particle in the t-th iteration, represents the local optimal position of individual particles, , and represents a random number in the range [0,1], represents the inertia weight of the particle, represents the individual learning factor, which is used to control the speed at which particles move to the individual optimal position. represents the group learning factor, which is used to control the speed at which particles move to the global optimal position. represents the excellent particle learning factor; Update the particle position using the following formula: ； In the formula, represents the position of the i-th particle in the t+1-th iteration; Get the position step factor, specifically use the particle random divergence strategy to get the particle divergence position step. The formula used is as follows: ； In the formula, Indicates the step size factor of the current iteration position, represents the standard deviation of the current population fitness, represents the mean of the current population fitness, and All of them represent random numbers that follow a normal distribution. represents the parameter controlling the step size; Update the local optimal position of individual particles. The formula used is as follows: ； In the formula, represents the updated local optimal position of the individual particle, represents the fitness function; Optimize individual particles with low fitness. Specifically, compare the fitness values ​​of the local optimal positions of all particles, re-sort the individual particles from best to worst according to the fitness values, select 10%N particles with low fitness values, and optimize the individual particles. The formula used is as follows: ； ； In the formula, Indicates the individual position of particles with low fitness after optimization, Indicates that the particle individuals with low fitness are selected. represents the dynamic optimization control parameter, represents the maximum value of the dynamic optimization control parameter, Indicates the minimum value of the dynamic optimization control parameter; Search determination, specifically, by constructing a search termination condition, searching and determining the optimal individual particle position, and obtaining the optimal individual particle position data setting; The search termination conditions include threshold termination and iteration termination; The threshold termination is specifically to set a fitness threshold, when the particle fitness value f i When it is above the fitness threshold, the search is completed; The iteration termination specifically refers to terminating the iteration and obtaining the global optimal position of the particle when the maximum number of iterations is reached; The global optimal position of the particle specifically refers to the optimal parameter combination of the anesthesia awakening monitoring model.

5. The patient anesthesia awakening status monitoring system based on artificial intelligence according to claim 1 is characterized in that: The patient anesthesia awakening management module specifically feeds back the patient's current anesthesia awakening status and the patient's complete awakening time to medical staff based on the patient's anesthesia awakening monitoring results, and issues real-time alarms based on the anesthesia awakening abnormal detection results to ensure that medical staff take corresponding measures in a timely manner.

6. The patient anesthesia awakening status monitoring system based on artificial intelligence according to claim 1 is characterized by: The data acquisition module specifically acquires the original data of anesthesia recovery status monitoring from the hospital's internal database; the original data of anesthesia recovery status monitoring includes historical anesthesia recovery abnormality detection data, real-time anesthesia recovery abnormality detection data, historical anesthesia recovery monitoring data and real-time anesthesia recovery monitoring data.

7. The patient anesthesia awakening status monitoring system based on artificial intelligence according to claim 1 is characterized in that: The data preprocessing module specifically performs data cleaning, data synchronization, rule engine screening, standardization processing, data segmentation and feature selection on the raw data of anesthesia recovery status monitoring to obtain preliminary data of anesthesia recovery status monitoring; the rule engine screening is to define the normal range of each physiological signal according to medical standards, and mark the data that does not meet the normal range as potential abnormalities; the data segmentation is the time window for segmented data processing.

Images