Anesthesia depth assessment method and system based on patient vital sign data

Abstract

The present invention discloses a method and system for assessing the depth of anesthesia based on patient vital sign data, and relates to the field of medical anesthesia technology. A system for assessing the depth of anesthesia based on patient vital sign data comprises: a pre-anesthesia assessment module and a post-anesthesia assessment module. The present invention combines preoperative vital sign data analysis, real-time anesthesia depth assessment, anesthesia metabolism prediction and subsequent anesthesia management, and utilizes a densely connected convolutional network, an attention mechanism, and dynamic time series analysis to accurately assess the depth of anesthesia and adjust the anesthesia plan in real time, thereby optimizing the patient's anesthesia management, ensuring the safety and personalization of the depth of anesthesia during surgery, and improving the anesthesia effect and patient safety.

Description

Technical Field

[0001] The present invention relates to the field of medical anesthesia technology, and in particular to a method and system for evaluating anesthesia depth based on patient vital sign data. Background Art

[0002] Anesthesia depth assessment is a key link in anesthesia management, which aims to ensure that patients can achieve adequate sedation during anesthesia while avoiding the risks of excessive anesthesia. The main purpose of preoperative anesthesia depth assessment is to evaluate the patient's overall health status, ensure the rational selection of anesthetic drugs, and predict potential reactions during anesthesia in order to develop a personalized anesthesia plan; while postoperative anesthesia depth assessment focuses on monitoring the metabolic process of anesthetic drugs, the patient's recovery status, and whether anesthesia-related complications occur to ensure the patient's safe recovery.

[0003] However, most existing technologies rely on monitoring methods based on a single physical sign, which has certain limitations. For example, EEG activity is used to assess the depth of anesthesia, but it is affected by the type of drug and individual differences among patients and may not accurately reflect the actual depth of anesthesia; and traditional monitoring methods based on blood pressure and heart rate may not fully reflect the patient's anesthetic status.

[0004] Therefore, it is particularly important to conduct a comprehensive assessment based on multidimensional physical sign data before and after surgery. By flexibly adjusting the focus of the assessment according to different needs before and after surgery, a more accurate assessment of the depth of anesthesia can be achieved, which can improve the accuracy of the assessment and thus ensure the safety and effectiveness of anesthesia management. Summary of the Invention

[0005] The present invention aims to provide a method and system for evaluating the depth of anesthesia based on patient vital sign data, so as to accurately evaluate the depth of anesthesia before and after surgery.

[0006] A method for assessing anesthesia depth based on patient vital sign data comprises the following steps:

[0007] Obtain the patient's preoperative vital signs data; the patient's preoperative vital signs data includes preoperative basic data, preoperative vital signs data, preoperative professional examination data, and patient real-time monitoring data; among which, the patient's real-time monitoring data is based on the patient's actual condition changes; based on the patient's preoperative vital signs data and the preoperative anesthesia depth assessment model, analyze the patient's real-time anesthesia depth assessment results;

[0008] The preoperative anesthesia depth assessment model combines data preprocessing, special signal processing, indicator evaluation, and result output layers, using a densely connected convolutional network and an adaptive attention mechanism to effectively analyze patients' preoperative vital signs and accurately assess anesthesia depth.

[0009] During the patient's surgery, the patient's real-time anesthesia depth assessment results are continuously collected to obtain the patient's intraoperative anesthesia depth assessment data set; the patient's intraoperative anesthesia depth assessment data set contains N patient real-time anesthesia depth assessment results G n , n=1, 2, ..., N; metabolic assessment prediction is performed based on the patient's intraoperative anesthesia depth assessment dataset and the patient's anesthesia metabolic assessment model to obtain the patient's postoperative anesthesia depth assessment result;

[0010] The patient anesthesia metabolic assessment model can accurately assess the patient's postoperative anesthesia depth through multi-level feature extraction and optimized metabolic prediction, combined with dynamic time series analysis and regression optimization;

[0011] Subsequent anesthesia management is performed based on the patient's postoperative anesthesia depth assessment results.

[0012] As a preferred technical solution of the present invention, the anesthesia depth preoperative assessment model includes a data preprocessing layer, a special signal processing layer, an index assessment layer and a result output layer;

[0013] The data preprocessing layer is used to divide the patient's preoperative vital sign data into special signal sign data and basic vital sign data; preprocess the basic vital sign data to obtain preprocessed basic vital sign data; wherein the special signal sign data is image feature data;

[0014] The special signal processing layer is used to perform feature analysis on special signal vital sign data to obtain the features of special signal vital sign data;

[0015] The special signal processing layer is based on the improvement of the traditional convolutional network using dense connections;

[0016] The indicator evaluation layer is used to evaluate the depth of anesthesia based on the characteristics of special signal sign data and pre-processed basic sign data, and obtain the patient's real-time anesthesia depth evaluation results;

[0017] The result output layer is used to output the patient's real-time anesthesia depth assessment results.

[0018] As a preferred technical solution of the present invention, the specific steps of performing feature analysis in the feature signal processing layer include:

[0019] In the feature signal processing layer, there are M feature recognition units D m , m=1, 2, ..., M; in the feature recognition unit D m In it, there are dense feature recognition blocks and adaptive attention blocks; set D m =[X 1m , X 2m , X 3m ]; where X 1m Indicates that in the feature recognition unit D mThe number of 1*1 convolutional layers in the dense feature recognition block, X 2m Indicates that in the feature recognition unit D m The number of 3*3 convolutional layers in the dense feature recognition block, X 3m In the feature recognition unit D m Channel attention coefficient of the adaptive attention block in

[15] ;

[0020] In the feature recognition unit D m In the example, the output data is the special signal sign data iterative feature T m , specific steps:

[0021] Receive special signal sign data iterative features T1, T2, ..., T m-1 Perform feature recognition to obtain special signal sign data iterative feature T m ;

[0022] Until all feature recognition units D are traversed m , obtain the special signal sign data characteristics;

[0023] The specific steps for training the feature signal processing layer include:

[0024] Collect several groups of special signal processing training samples, each of which contains target parameter matching factors and verified special signal sign data; combine several groups of special signal processing training samples to obtain a special signal processing training set; the target parameter matching factor is used to match the feature recognition unit D m Channel attention coefficient of the adaptive attention block in

[15] ;

[0025] The model is trained using a special signal processing training set to obtain an initial feature signal processing layer. The initial feature signal processing layer is evaluated. If the initial feature signal processing layer passes the model evaluation, the initial feature signal processing layer is used as the feature signal processing layer in the preoperative assessment model of anesthesia depth. Otherwise, the model training is continued using the special signal processing training set.

[0026] As a preferred technical solution of the present invention, the specific steps of performing anesthesia depth assessment at the index assessment layer include:

[0027] The improved anesthesia depth assessment matrix based on TOPSIS method is used for depth identification, wherein the anesthesia depth assessment matrix is used Calculate the patient's real-time anesthesia depth assessment results; perform time slice data segmentation on the special signal sign data characteristics and pre-processed basic sign data to obtain feature item data ;

[0028] Where i is the basic time slice, i=1, 2, ..., I, I is the total time for obtaining the patient's preoperative vital sign data; J is the total number of feature items in the special signal sign data features and preprocessed basic vital sign data, j=1, 2, ..., J; Represents the characteristics of special signal vital sign data and the jth feature item data of the i-th basic time slice in the preprocessed basic vital sign data; Represents the feature item weighting factor.

[0029] As a preferred technical solution of the present invention, the patient anesthesia metabolism assessment model includes a main feature extraction layer, an optimized metabolism prediction layer and a postoperative result output layer;

[0030] The main feature extraction layer is used to evaluate the real-time anesthesia depth of patients in the patient intraoperative anesthesia depth assessment dataset G n Perform time series feature extraction to obtain the patient's real-time anesthesia depth assessment result feature G n 'and the time series characteristics of the patient's intraoperative anesthesia depth assessment data;

[0031] The optimized metabolic prediction layer is used to calculate the patient's intraoperative anesthesia depth assessment data time series characteristics and the patient's real-time anesthesia depth assessment result characteristics G n 'And the anesthesia metabolism regression function is used to optimize the metabolic prediction analysis and obtain the patient's postoperative anesthesia depth assessment results;

[0032] The postoperative result output layer is used to output the patient's postoperative anesthesia depth assessment results.

[0033] As a preferred technical solution of the present invention, the specific steps of extracting temporal features in the main feature extraction layer include:

[0034] The main feature extraction layer includes a multi-point data connection layer and a dynamic time feature extraction layer;

[0035] Real-time anesthesia depth assessment results of patients in multi-point data connection layer G n Perform adaptive convolution feature extraction to obtain the patient's real-time anesthesia depth assessment result feature G n ';

[0036] In the dynamic time feature extraction layer, the feature G is evaluated based on the real-time anesthesia depth of all patients. n 'Perform time series feature extraction to obtain the time series features of the patient's intraoperative anesthesia depth assessment data.

[0037] As a preferred technical solution of the present invention, the specific steps of constructing the anesthesia metabolism regression function include:

[0038] An initial anesthesia metabolism regression function is constructed based on linear regression analysis and historical anesthesia metabolism data; the historical anesthesia metabolism data contains a historical anesthesia depth assessment data set of several patients;

[0039] Construct K prediction function adjustment factor individuals H k , k=1, 2, ..., K; wherein each prediction function adjustment factor individual contains an adjustment strategy for adjusting the initial anesthesia metabolism regression function; K prediction function adjustment factor individuals H k Combine to obtain the iterative population of the prediction function adjustment factor; set the number of iterations p, p=1, 2, ..., P, P is the maximum number of iterations;

[0040] Prediction function adjustment factor individual H k The fitness is S k , represents the prediction function adjustment factor individual H k Improvement in the predictive ability of the initial anesthesia metabolism regression function;

[0041] In the process of population iteration, each time the prediction function adjustment factor iteration population is screened, the prediction function adjustment factor individuals with poor fitness and good fitness are called the crossover mutation prediction function adjustment factor population;

[0042] Using the formula B(p)=B0*e -λp (1+β*avg(p)) controls the variability of the cross-variation prediction function to adjust the factor population;

[0043] Among them, B(p) is the variability, B0 is the initial variability, λ is the decay rate, β is the factor that controls the influence of individual distribution on variability, and avg(p) represents the mean fitness of the iterative population of the prediction function adjustment factor when the number of iterations is p;

[0044] The cross-mutation prediction function adjustment factor population is updated according to B(p) to obtain a new cross-mutation prediction function adjustment factor population; the new cross-mutation prediction function adjustment factor population is combined with the prediction function adjustment factor iteration population to obtain a new prediction function adjustment factor iteration population; the new prediction function adjustment factor iteration population is used to perform the next population iteration;

[0045] When the maximum number of iterations is reached, the prediction function adjustment factor individual corresponding to the current maximum fitness is output, which is the optimal prediction function adjustment factor individual. The initial anesthesia metabolism regression function is optimized based on the optimal prediction function adjustment factor individual to obtain the anesthesia metabolism regression function.

[0046] An anesthesia depth assessment system based on patient vital sign data, comprising:

[0047] The anesthesia preoperative assessment module includes a data acquisition unit and a preoperative analysis unit; the data acquisition unit is used to obtain the patient's preoperative vital signs data; the patient's preoperative vital signs data contains preoperative basic data, preoperative vital signs data, preoperative professional examination data and patient real-time monitoring data; among which, the patient's real-time monitoring data is based on the patient's actual state changes; the preoperative analysis unit is used to analyze the patient's preoperative vital signs data and the anesthesia depth preoperative assessment model to obtain the patient's real-time anesthesia depth assessment result; the anesthesia depth preoperative assessment model combines data preprocessing, special signal processing, indicator evaluation and result output layer, and uses a densely connected convolutional network and an adaptive attention mechanism to effectively analyze the patient's preoperative vital signs data and accurately assess the anesthesia depth;

[0048] The postoperative anesthesia assessment module includes a data acquisition unit and a postoperative analysis unit; the data acquisition unit is used to continuously collect the patient's real-time anesthesia depth assessment results during the patient's surgery to obtain the patient's intraoperative anesthesia depth assessment data set; the patient's intraoperative anesthesia depth assessment data set contains N patient real-time anesthesia depth assessment results G n , n=1, 2, …, N; the postoperative analysis unit is used to perform metabolic assessment and prediction based on the patient's intraoperative anesthesia depth assessment dataset and the patient's anesthesia metabolic assessment model to obtain the patient's postoperative anesthesia depth assessment result; the patient's anesthesia metabolic assessment model can accurately assess the patient's postoperative anesthesia depth through multi-level feature extraction and optimized metabolic prediction, combined with dynamic time series analysis and regression optimization; subsequent anesthesia management is performed based on the patient's postoperative anesthesia depth assessment result.

[0049] The present invention has the following advantages:

[0050] 1. By combining preoperative vital sign data analysis, real-time anesthesia depth assessment, anesthesia metabolism prediction and subsequent anesthesia management, the present invention utilizes densely connected convolutional networks, attention mechanisms and dynamic time series analysis to accurately assess anesthesia depth and adjust anesthesia plans in real time, thereby optimizing the patient's anesthesia management, ensuring the safety and personalization of anesthesia depth during surgery, and improving anesthesia effects and patient safety.

[0051] 2. The present invention combines the patient's preoperative vital sign data and the preoperative anesthesia depth assessment model to accurately assess the depth of anesthesia in real time, which helps to maintain the optimal depth of anesthesia during surgery and avoid over-anesthesia or under-anesthesia, thereby reducing surgical risks and improving patient safety. In the special signal processing layer, the model improves the traditional convolutional network through a densely connected convolutional network. The dense connection can effectively utilize feature information at different levels, enhance the network's expression ability, and improve the ability to extract and analyze complex signal features. The special signal sign data is combined with the basic sign data, so that complex signals such as image data can be effectively processed, thereby improving the accuracy of anesthesia depth assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a schematic structural diagram of an anesthesia depth assessment system based on patient vital sign data adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0054] Example 1, a method for assessing anesthesia depth based on patient vital sign data, comprising the following steps:

[0055] Obtaining the patient's preoperative vital signs data; the patient's preoperative vital signs data includes preoperative basic data, preoperative vital signs data, preoperative professional examination data, and patient real-time monitoring data; wherein, the patient's real-time monitoring data is based on the patient's actual status changes;

[0056] Basic preoperative data are usually collected before surgery and do not change significantly throughout the entire surgical process and after the surgery, including the patient's age, gender, height, weight, past medical history, etc. They usually do not change over time, but are related to the patient's basic health status; preoperative vital signs data may change over time, including blood pressure, heart rate, respiratory rate, body temperature, etc. Preoperative vital signs data are usually closely related to time and are therefore dynamically affected by the patient's current health status; preoperative professional examination data are usually obtained by professional examinations, such as blood tests, imaging tests, electrocardiograms, etc.; real-time patient monitoring data are based on changes in the patient's actual status and are collected in real time. Usually, with various factors during the operation, including electrocardiograms, blood oxygen saturation, ventilator parameters, etc., these data are collected in real time and are continuously updated over time for real-time assessment of the patient's status;

[0057] Based on the analysis of the patient's preoperative physical sign data and the preoperative anesthesia depth assessment model, the patient's real-time anesthesia depth assessment results are obtained;

[0058] The preoperative anesthesia depth assessment model combines data preprocessing, special signal processing, indicator evaluation, and result output layers, using a densely connected convolutional network and an adaptive attention mechanism to effectively analyze patients' preoperative vital signs and accurately assess anesthesia depth.

[0059] The anesthesia depth preoperative assessment model includes a data preprocessing layer, a special signal processing layer, an indicator evaluation layer, and a result output layer;

[0060] The data preprocessing layer is used to divide the patient's preoperative vital sign data into special signal sign data and basic vital sign data; preprocess the basic vital sign data to obtain preprocessed basic vital sign data; wherein the special signal sign data is image feature data;

[0061] The special signal processing layer is used to perform feature analysis on special signal vital sign data to obtain the features of special signal vital sign data;

[0062] The special signal processing layer is based on the improvement of the traditional convolutional network using dense connections;

[0063] The indicator evaluation layer is used to evaluate the depth of anesthesia based on the characteristics of special signal sign data and pre-processed basic sign data, and obtain the patient's real-time anesthesia depth evaluation results;

[0064] The result output layer is used to output the patient's real-time anesthesia depth assessment results;

[0065] By combining the patient's preoperative vital sign data and the preoperative anesthesia depth assessment model, the anesthesia depth can be assessed in real time and accurately, which helps maintain the optimal anesthesia depth during surgery and avoid over-anesthesia or under-anesthesia, thereby reducing surgical risks and improving patient safety. In the special signal processing layer, the model improves the traditional convolutional network through a densely connected convolutional network. The dense connection can effectively utilize feature information at different levels, enhance the network's expression ability, and improve the ability to extract and analyze complex signal features. Combining special signal sign data with basic sign data allows complex signals such as image data to be effectively processed, thereby improving the accuracy of anesthesia depth assessment. For example, image data from signals such as electroencephalogram (EEG) and electrocardiogram (ECG) can be deeply analyzed to better predict the patient's anesthesia status.

[0066] The specific steps of feature analysis in the feature signal processing layer include:

[0067] In the feature signal processing layer, there are M feature recognition units D m , m=1, 2, ..., M; in the feature recognition unit D m In it, there are dense feature recognition blocks and adaptive attention blocks; set D m =[X 1m , X 2m , X 3m ]; where X 1m Indicates that in the feature recognition unit D m The number of 1*1 convolutional layers in the dense feature recognition block, X 2m Indicates that in the feature recognition unit D m The number of 3*3 convolutional layers in the dense feature recognition block, X 3m In the feature recognition unit Dm Channel attention coefficient of the adaptive attention block in

[15] ;

[0068] In the feature recognition unit D m In the example, the output data is the special signal sign data iterative feature T m , specific steps:

[0069] Receive special signal sign data iterative features T1, T2, ..., T m-1 Perform feature recognition to obtain special signal sign data iterative feature T m ;

[0070] Until all feature recognition units D are traversed m , obtain the special signal sign data characteristics;

[0071] The specific steps for training the feature signal processing layer include:

[0072] Collect several groups of special signal processing training samples, each of which contains target parameter matching factors and verified special signal sign data; combine several groups of special signal processing training samples to obtain a special signal processing training set; the target parameter matching factor is used to match the feature recognition unit D m Channel attention coefficient of the adaptive attention block in

[15] ;

[0073] The model is trained using the special signal processing training set to obtain an initial feature signal processing layer; the initial feature signal processing layer is evaluated; if the initial feature signal processing layer passes the model evaluation, the initial feature signal processing layer is used as the feature signal processing layer in the preoperative assessment model for depth of anesthesia; otherwise, the model training is continued using the special signal processing training set;

[0074] The dense feature recognition block in the feature recognition unit is composed of 1×1 convolutional layers and 3×3 convolutional layers, which can effectively capture multi-level local and global features. The 1×1 convolutional layer can be used to reduce the number of channels and improve computational efficiency, while the 3×3 convolutional layer helps capture richer contextual information. This combination improves the model's expressiveness when processing complex signals and enhances its sensitivity to signals related to anesthesia depth assessment. By introducing the channel attention coefficient through the adaptive attention block, the model can dynamically adjust the attention of different channels according to the characteristics of the input data. The channel attention mechanism can effectively distinguish the importance of different features when processing anesthesia depth-related features, thereby optimizing feature representation and improving the model's adaptability and accuracy to different patient vital signs. Through layer-by-layer iterative processing, the model can gradually dig deeper and refine more abstract and efficient feature representations. The output features of each feature recognition unit will gradually form more refined anesthesia depth-related signal features, thereby improving the accuracy of anesthesia depth assessment.

[0075] The specific steps for evaluating the depth of anesthesia at the indicator evaluation level include:

[0076] The improved anesthesia depth assessment matrix based on TOPSIS method is used for depth identification, wherein the anesthesia depth assessment matrix is used Calculate the patient's real-time anesthesia depth assessment results; perform time slice data segmentation on the special signal sign data characteristics and pre-processed basic sign data to obtain feature item data ;

[0077] Where i is the basic time slice, i=1, 2, ..., I, I is the total time for obtaining the patient's preoperative vital sign data; J is the total number of feature items in the special signal sign data features and preprocessed basic vital sign data, j=1, 2, ..., J; Represents the characteristics of special signal vital sign data and the jth feature item data of the i-th basic time slice in the preprocessed basic vital sign data; represents the weighting factor of the feature item;

[0078] The TOPSIS method is used to identify the anesthesia depth assessment matrix. By calculating the relative proximity of each feature indicator to the ideal solution, the anesthesia depth can be quantified and assessed more scientifically, improving the objectivity and accuracy of the assessment. By introducing feature weights, the indicator weights can be dynamically adjusted according to the importance of specific features. In this way, when assessing the anesthesia depth, key features can be highlighted, the interference of secondary features on the results can be reduced, and the sensitivity and accuracy of the model can be ensured. Special signal features and pre-processed basic vital sign data are segmented and synthesized to form a feature item data matrix. This multi-feature fusion method can comprehensively reflect the patient's anesthesia-related vital sign information and avoid misjudgment that may be caused by a single feature. The matrix-based anesthesia depth assessment results are data-driven, reducing the influence of subjective factors on anesthesia depth assessment and providing clinicians with more reliable auxiliary decision support.

[0079] During the patient's surgery, the patient's real-time anesthesia depth assessment results are continuously collected to obtain the patient's intraoperative anesthesia depth assessment data set; the patient's intraoperative anesthesia depth assessment data set contains N patient real-time anesthesia depth assessment results G n , n=1, 2, ..., N; metabolic assessment prediction is performed based on the patient's intraoperative anesthesia depth assessment dataset and the patient's anesthesia metabolic assessment model to obtain the patient's postoperative anesthesia depth assessment result;

[0080] The patient anesthesia metabolic assessment model can accurately assess the patient's postoperative anesthesia depth through multi-level feature extraction and optimized metabolic prediction, combined with dynamic time series analysis and regression optimization;

[0081] The patient anesthesia metabolic assessment model includes a main feature extraction layer, an optimized metabolic prediction layer, and a postoperative result output layer;

[0082] The main feature extraction layer is used to evaluate the real-time anesthesia depth of patients in the patient intraoperative anesthesia depth assessment dataset G n Perform time series feature extraction to obtain the patient's real-time anesthesia depth assessment result feature G n 'and the time series characteristics of the patient's intraoperative anesthesia depth assessment data;

[0083] The optimized metabolic prediction layer is used to calculate the patient's intraoperative anesthesia depth assessment data time series characteristics and the patient's real-time anesthesia depth assessment result characteristics G n 'And the anesthesia metabolism regression function is used to optimize the metabolic prediction analysis and obtain the patient's postoperative anesthesia depth assessment results;

[0084] The postoperative result output layer is used to output the patient's postoperative anesthesia depth assessment results;

[0085] The specific steps for extracting temporal features in the main feature extraction layer include:

[0086] The main feature extraction layer includes a multi-point data connection layer and a dynamic time feature extraction layer;

[0087] Real-time anesthesia depth assessment results of patients in multi-point data connection layer G n Perform adaptive convolution feature extraction to obtain the patient's real-time anesthesia depth assessment result feature G n ';

[0088] In the dynamic time feature extraction layer, the feature G is evaluated based on the real-time anesthesia depth of all patients. n 'Perform time series feature extraction to obtain the time series features of the patient's intraoperative anesthesia depth assessment data;

[0089] Through dynamic time series analysis and regression optimization, the patient anesthesia metabolic assessment model can efficiently analyze the intraoperative anesthesia depth assessment data, and combine the results of multi-level feature extraction to accurately assess the patient's postoperative anesthesia depth, which helps to reduce postoperative anesthesia-related complications; in the main feature extraction layer, the patient's real-time anesthesia depth assessment results are multi-point connected and dynamic time feature extracted, which can not only capture single-point features, but also mine the global pattern of time series data, providing more comprehensive information for metabolic assessment prediction; the optimized metabolic prediction layer combines the time series features of the patient's intraoperative anesthesia depth assessment data, the real-time assessment result features and the anesthesia metabolic regression function to optimize the metabolic process and effectively reflect the individual differences of patients, thereby improving the accuracy and reliability of postoperative anesthesia depth assessment; in the dynamic time feature extraction layer, using the time series model, the model can capture the dynamic change trend of intraoperative anesthesia depth over time, and provide more timely input information for postoperative anesthesia depth prediction. This method can better adapt to the changes in anesthesia depth during long-term surgery;

[0090] The specific steps to construct the anesthesia metabolism regression function include:

[0091] An initial anesthesia metabolism regression function is constructed based on linear regression analysis and historical anesthesia metabolism data; the historical anesthesia metabolism data contains a historical anesthesia depth assessment data set of several patients;

[0092] Construct K prediction function adjustment factor individuals H k , k=1, 2, ..., K; wherein each prediction function adjustment factor individual contains an adjustment strategy for adjusting the initial anesthesia metabolism regression function; K prediction function adjustment factor individuals H k Combination, to obtain the iterative population of the prediction function adjustment factor; set the number of iterations p, p = 1, 2, ..., P, P is the maximum number of iterations; the maximum number of iterations is set by professional technicians according to actual conditions;

[0093] Prediction function adjustment factor individual H k The fitness is S k , represents the prediction function adjustment factor individual H k Improvement in the predictive ability of the initial anesthesia metabolism regression function;

[0094] In the process of population iteration, each time the prediction function adjustment factor iteration population is screened, the prediction function adjustment factor individuals with poor fitness and good fitness are called the crossover mutation prediction function adjustment factor population;

[0095] Using the formula B(p)=B0*e -λp (1+β*avg(p)) controls the variability of the cross-variation prediction function to adjust the factor population;

[0096] Where B(p) is the variability, B0 is the initial variability, λ is the decay rate, β is the factor that controls the influence of individual distribution on variability, and avg(p) represents the mean fitness of the prediction function adjustment factor iteration population when the number of iterations is p. The initial variability is set by professional technicians based on actual conditions.

[0097] The cross-mutation prediction function adjustment factor population is updated according to B(p) to obtain a new cross-mutation prediction function adjustment factor population; the new cross-mutation prediction function adjustment factor population is combined with the prediction function adjustment factor iteration population to obtain a new prediction function adjustment factor iteration population; the new prediction function adjustment factor iteration population is used to perform the next population iteration;

[0098] When the maximum number of iterations is reached, the prediction function adjustment factor individual corresponding to the current maximum fitness is output, which is the optimal prediction function adjustment factor individual. The initial anesthesia metabolism regression function is optimized based on the optimal prediction function adjustment factor individual to obtain the anesthesia metabolism regression function;

[0099] By introducing a prediction function adjustment factor and an iterative optimization mechanism, the predictive capability of the initial anesthesia metabolism regression function has been significantly improved. The model can more accurately fit historical anesthesia metabolism data, thereby improving the accuracy of postoperative anesthesia depth assessment for patients. Utilizing cross-mutation and fitness screening strategies, the adjustment factor population can dynamically evolve, gradually optimizing the initial anesthesia metabolism regression function. This adaptive optimization method enables the model to continuously approach the optimal prediction result. Dynamically controlling the variability through the formula increases exploration capabilities in the early stages of optimization, while reducing randomness in the later iterations to focus on development. This balancing mechanism improves the algorithm's convergence speed and the ultimate optimization effect. By introducing a control factor and leveraging the effect of the population fitness mean on variability, the diversity of the prediction function adjustment factor population is effectively maintained, preventing the population from falling into a single mode and improving global optimization capabilities. In each iteration, individual prediction function adjustment factors with better and worse fitness undergo cross-mutation updates, ensuring dynamic adjustment of the model optimization direction. This mechanism can accelerate the screening process of the optimal adjustment factor and improve the optimization efficiency of the model. The optimized anesthesia metabolism regression function can quickly provide high-precision prediction results in clinical applications, providing doctors with reliable decision-making support for postoperative anesthesia depth management, and improving the safety and efficiency of postoperative patient recovery.

[0100] Subsequent anesthesia management is carried out based on the results of the patient's postoperative anesthesia depth assessment. For example, after the patient's surgery is completed, the postoperative anesthesia depth assessment results are monitored in real time to determine whether the patient has reached the awakening standard, prevent delayed awakening due to residual anesthetics, and improve the safety of the patient's postoperative awakening. When the assessment results show that the patient's anesthesia depth gradually decreases to a safe level, the doctor can remove the endotracheal tube in a timely manner and arrange for the patient to enter the postoperative intensive care unit; based on the anesthesia depth assessment results, assess whether the patient has fully recovered consciousness, and formulate a postoperative analgesia strategy based on the patient's pain feedback. If the assessment results show that the anesthetic is metabolized too quickly, it may be necessary to supplement analgesics in time to relieve the patient's pain. By assessing the patient's metabolism and recovery, the dosage of analgesics can be adjusted to avoid side effects caused by excessive medication; the postoperative anesthesia depth assessment results can significantly improve the safety, personalization and efficiency of postoperative management, and provide patients with better recovery guarantees.

[0101] Example 2, a system for assessing the depth of anesthesia based on patient vital signs data, see Figure 1 Shown, including:

[0102] The anesthesia preoperative assessment module includes a data acquisition unit and a preoperative analysis unit; the data acquisition unit is used to obtain the patient's preoperative vital signs data; the patient's preoperative vital signs data contains preoperative basic data, preoperative vital signs data, preoperative professional examination data and patient real-time monitoring data; among which, the patient's real-time monitoring data is based on the patient's actual state changes; the preoperative analysis unit is used to analyze the patient's preoperative vital signs data and the anesthesia depth preoperative assessment model to obtain the patient's real-time anesthesia depth assessment result; the anesthesia depth preoperative assessment model combines data preprocessing, special signal processing, indicator evaluation and result output layer, and uses a densely connected convolutional network and an adaptive attention mechanism to effectively analyze the patient's preoperative vital signs data and accurately assess the anesthesia depth;

[0103] The postoperative anesthesia assessment module includes a data acquisition unit and a postoperative analysis unit; the data acquisition unit is used to continuously collect the patient's real-time anesthesia depth assessment results during the patient's surgery to obtain the patient's intraoperative anesthesia depth assessment data set; the patient's intraoperative anesthesia depth assessment data set contains N patient real-time anesthesia depth assessment results G n , n=1, 2, …, N; the postoperative analysis unit is used to perform metabolic assessment and prediction based on the patient's intraoperative anesthesia depth assessment dataset and the patient's anesthesia metabolic assessment model to obtain the patient's postoperative anesthesia depth assessment result; the patient's anesthesia metabolic assessment model can accurately assess the patient's postoperative anesthesia depth through multi-level feature extraction and optimized metabolic prediction, combined with dynamic time series analysis and regression optimization; subsequent anesthesia management is performed based on the patient's postoperative anesthesia depth assessment result.

[0104] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.

Claims

1. A system for assessing the depth of anesthesia based on patient vital sign data, characterized in that: include: Anesthesia preoperative assessment module, including a data acquisition unit and a preoperative analysis unit; The data acquisition unit is used to obtain the patient's preoperative vital sign data; The patient's preoperative vital signs data contains preoperative basic data, preoperative vital signs data, preoperative professional examination data, and real-time patient monitoring data; among them, the real-time patient monitoring data is based on the patient's actual status changes; the preoperative analysis unit is used to analyze the patient's preoperative vital signs data and the preoperative anesthesia depth assessment model to obtain the patient's real-time anesthesia depth assessment result; the preoperative anesthesia depth assessment model combines data preprocessing, special signal processing, indicator evaluation and result output layer, and uses a densely connected convolutional network and adaptive attention mechanism to effectively analyze the patient's preoperative vital signs data and accurately assess the anesthesia depth; The postoperative anesthesia assessment module includes a data acquisition unit and a postoperative analysis unit; the data acquisition unit is used to continuously collect the patient's real-time anesthesia depth assessment results during the patient's surgery to obtain the patient's intraoperative anesthesia depth assessment data set; the patient's intraoperative anesthesia depth assessment data set contains N patient real-time anesthesia depth assessment results G n , n=1, 2, …, N; the postoperative analysis unit is used to perform metabolic assessment and prediction based on the patient's intraoperative anesthesia depth assessment dataset and the patient's anesthesia metabolic assessment model to obtain the patient's postoperative anesthesia depth assessment result; the patient's anesthesia metabolic assessment model can accurately assess the patient's postoperative anesthesia depth through multi-level feature extraction and optimized metabolic prediction, combined with dynamic time series analysis and regression optimization; subsequent anesthesia management is performed based on the patient's postoperative anesthesia depth assessment result; The anesthesia depth preoperative assessment model includes a data preprocessing layer, a special signal processing layer, an indicator evaluation layer, and a result output layer; The data preprocessing layer is used to divide the patient's preoperative vital sign data into special signal sign data and basic vital sign data; preprocess the basic vital sign data to obtain preprocessed basic vital sign data; wherein the special signal sign data is image feature data; The special signal processing layer is used to perform feature analysis on special signal vital sign data to obtain the features of special signal vital sign data; The special signal processing layer is based on the improvement of the traditional convolutional network using dense connections; The specific steps of feature analysis in the feature signal processing layer include: In the feature signal processing layer, there are M feature recognition units D m , m=1, 2, ..., M; in the feature recognition unit D m In it, there are dense feature recognition blocks and adaptive attention blocks; set D m =[X 1m , X 2m , X 3m ]; where X 1m Indicates that in the feature recognition unit D m The number of 1*1 convolutional layers in the dense feature recognition block, X 2m Indicates that in the feature recognition unit D m The number of 3*3 convolutional layers in the dense feature recognition block, X 3m In the feature recognition unit D m Channel attention coefficient of the adaptive attention block in [15]; The patient anesthesia metabolic assessment model includes a main feature extraction layer, an optimized metabolic prediction layer, and a postoperative result output layer; The specific steps for extracting temporal features in the main feature extraction layer include: The main feature extraction layer includes a multi-point data connection layer and a dynamic time feature extraction layer; Real-time anesthesia depth assessment results of patients in multi-point data connection layer G n Perform adaptive convolution feature extraction to obtain the patient's real-time anesthesia depth assessment result feature G n '; In the dynamic time feature extraction layer, the feature G is evaluated based on the real-time anesthesia depth of all patients. n 'Perform time series feature extraction to obtain the time series features of the patient's intraoperative anesthesia depth assessment data; The optimized metabolic prediction layer is used to calculate the patient's intraoperative anesthesia depth assessment data time series characteristics and the patient's real-time anesthesia depth assessment result characteristics G n 'And the anesthesia metabolism regression function is used to optimize the metabolic prediction analysis and obtain the patient's postoperative anesthesia depth assessment results; The specific steps to construct the anesthesia metabolism regression function include: An initial anesthesia metabolism regression function is constructed based on linear regression analysis and historical anesthesia metabolism data; the historical anesthesia metabolism data contains a historical anesthesia depth assessment data set of several patients; Construct K prediction function adjustment factor individuals H k , k=1, 2, ..., K; wherein each prediction function adjustment factor individual contains an adjustment strategy for adjusting the initial anesthesia metabolism regression function; K prediction function adjustment factor individuals H k Combine to obtain the iterative population of the prediction function adjustment factor; set the number of iterations p, p=1, 2, ..., P, P is the maximum number of iterations; Prediction function adjustment factor individual H k The fitness is S k , represents the prediction function adjustment factor individual H k Improvement in the predictive ability of the initial anesthesia metabolism regression function; In the process of population iteration, each time the prediction function adjustment factor iteration population is screened, the prediction function adjustment factor individuals with poor fitness and good fitness are called the crossover mutation prediction function adjustment factor population; Using the formula B(p)=B0*e -λp (1+β*avg(p)) controls the variability of the cross-variation prediction function to adjust the factor population; Among them, B(p) is the variability, B0 is the initial variability, λ is the decay rate, β is the factor that controls the influence of individual distribution on variability, and avg(p) represents the mean fitness of the iterative population of the prediction function adjustment factor when the number of iterations is p; The cross-mutation prediction function adjustment factor population is updated according to B(p) to obtain a new cross-mutation prediction function adjustment factor population; the new cross-mutation prediction function adjustment factor population is combined with the prediction function adjustment factor iteration population to obtain a new prediction function adjustment factor iteration population; the new prediction function adjustment factor iteration population is used to perform the next population iteration; When the maximum number of iterations is reached, the prediction function adjustment factor individual corresponding to the current maximum fitness is output, which is the optimal prediction function adjustment factor individual. The initial anesthesia metabolism regression function is optimized based on the optimal prediction function adjustment factor individual to obtain the anesthesia metabolism regression function.

2. The anesthesia depth assessment system based on patient vital sign data according to claim 1, characterized in that: The indicator evaluation layer is used to evaluate the depth of anesthesia based on the characteristics of special signal sign data and pre-processed basic sign data, and obtain the patient's real-time anesthesia depth evaluation results; The result output layer is used to output the patient's real-time anesthesia depth assessment results.

3. The anesthesia depth assessment system based on patient vital sign data according to claim 2, characterized in that: In the feature recognition unit D m In the example, the output data is the special signal sign data iterative feature T m , specific steps: Receive special signal sign data iterative features T1, T2, ..., T m-1 Perform feature recognition to obtain special signal sign data iterative feature T m ; Until all feature recognition units D are traversed m , obtain the special signal sign data characteristics; The specific steps for training the feature signal processing layer include: Collect several groups of special signal processing training samples, each of which contains target parameter matching factors and verified special signal sign data; combine several groups of special signal processing training samples to obtain a special signal processing training set; the target parameter matching factor is used to match the feature recognition unit D m Channel attention coefficient of the adaptive attention block in [15]; The model is trained using a special signal processing training set to obtain an initial feature signal processing layer. The initial feature signal processing layer is evaluated. If the initial feature signal processing layer passes the model evaluation, the initial feature signal processing layer is used as the feature signal processing layer in the preoperative assessment model of anesthesia depth. Otherwise, the model training is continued using the special signal processing training set.

4. The anesthesia depth assessment system based on patient vital sign data according to claim 3, characterized in that: The specific steps for evaluating the depth of anesthesia at the indicator evaluation level include: The improved anesthesia depth assessment matrix based on TOPSIS method is used for depth identification, wherein the anesthesia depth assessment matrix is used Calculate the patient's real-time anesthesia depth assessment results; perform time slice data segmentation on the special signal sign data characteristics and pre-processed basic sign data to obtain feature item data ; Where i is the basic time slice, i=1, 2, ..., I, I is the total time for obtaining the patient's preoperative vital sign data; J is the total number of feature items in the special signal sign data features and preprocessed basic vital sign data, j=1, 2, ..., J; Represents the characteristics of special signal vital sign data and the jth feature item data of the i-th basic time slice in the preprocessed basic vital sign data; Represents the feature item weighting factor.

5. The anesthesia depth assessment system based on patient vital sign data according to claim 4, characterized in that: The postoperative result output layer is used to output the patient's postoperative anesthesia depth assessment results.

Images