A method for evaluating the status of patients after anesthesia

Abstract

The present invention discloses a method for evaluating the status of a patient after anesthesia, and relates to the technical field of medical data analysis. The system for evaluating the status of a patient after anesthesia constructed by the method includes: a data acquisition module, a data preprocessing module, a comprehensive evaluation module, an evaluation verification module, and a record optimization module. The present invention provides a comprehensive evaluation of the patient's physical, psychological, and neurological status by combining objective data (vital signs, anesthesia operation data, and disease history) with subjective data (descriptions of patients and experts), ensuring the comprehensiveness and accuracy of the evaluation results, and effectively handling the uncertainty and ambiguity of the data through the combination of multiple models (objective fuzzy evaluation model and subjective weight model), thereby improving the accuracy and reliability of the patient status evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data analysis, and in particular to a method for evaluating the status of a patient after anesthesia. Background Art

[0002] In the management of patients after anesthesia, it is crucial to assess their recovery status. Traditional assessment methods mainly rely on the subjective judgment of clinicians, which often lacks systematicity and comprehensiveness and may lead to deviations in assessment results. This method is not only time-consuming but also easily affected by individual experience, thus affecting patient safety and satisfaction.

[0003] With the rapid development of data science and artificial intelligence technology, more and more hospitals are beginning to adopt data-based assessment models. These models can integrate multiple data sources, such as real-time monitoring data, electronic medical records, and patient feedback, for comprehensive analysis. By applying advanced technologies such as machine learning and fuzzy logic, the accuracy and reliability of the assessment can be improved.

[0004] In addition, the patient's psychological state also plays an important role in postoperative recovery. In recent years, studies have shown that psychological problems such as postoperative anxiety and depression can significantly affect the patient's recovery process. Therefore, it has become a trend to incorporate subjective psychological assessment into the comprehensive assessment system.

[0005] However, existing technologies still face challenges in how to effectively combine subjective and objective data. Traditional models often have difficulty dealing with uncertainty and ambiguity in data, limiting the accuracy and applicability of evaluation results.

[0006] Therefore, there is an urgent need for a new assessment method that can systematically integrate multi-dimensional data and use advanced algorithms for analysis to provide a more comprehensive and reliable assessment of the patient's status after anesthesia. Summary of the invention

[0007] A method for evaluating the status of a patient after anesthesia surgery comprises the following steps:

[0008] Step S1: Obtaining post-anesthesia patient-related data, including objective data and subjective data. The objective data includes real-time patient body data, anesthesia surgery data and disease history data, wherein the anesthesia surgery data and disease history data are a posteriori data, and the real-time patient body data is a priori data; the subjective data includes patient description data and expert description data;

[0009] Step S2: Preprocess the data related to the patients after anesthesia, perform data cleaning, feature division, data conversion, outlier detection and null value filling on the objective data, obtain the feature data of each subtype of the objective data, and use them as objective parameters; send the subjective data to the keyword extraction model for calculation, obtain the patient description keywords and expert description keywords of the subjective data, and use them as subjective parameters; construct the objective parameters and subjective parameters into patient evaluation samples;

[0010] Step S3: Send the patient evaluation sample into the comprehensive evaluation model for calculation to obtain multiple patient evaluation results. The comprehensive evaluation model is constructed by combining three sub-models, namely, an objective fuzzy evaluation model, a subjective weight model, and an objective optimization evaluation model; the objective fuzzy evaluation model calculates the objective parameters corresponding to the prior data of the patient evaluation sample to obtain an objective first evaluation result, and the objective first evaluation result is an evaluation of the patient's physical state; the subjective weight model calculates the subjective parameters and the objective first evaluation result to obtain the psychological weights of the subjective evaluation result and the objective parameters corresponding to the prior data, and the subjective evaluation result is an evaluation of the patient's psychological state; the objective parameters corresponding to the prior data are weighted using the psychological weights, and the weighted results and the objective parameters corresponding to the posterior data are sent to the objective optimization evaluation model for calculation to obtain an objective second evaluation result, and the objective second evaluation result is an evaluation of the patient's physical state and neural state; the objective second evaluation result and the subjective evaluation result are sent to the comprehensive evaluation layer of the comprehensive evaluation model for matching, and multiple patient evaluation results are output, and the patient evaluation results are evaluations of the patient's physical, psychological, and neural directions;

[0011] Step S4: obtaining the state pattern credibility of the patient evaluation result through the matching degree of the patient's physical state, psychological state and neural state, obtaining the deviation credibility based on the accuracy of the patient evaluation result, performing weighted calculation on the state pattern credibility and the deviation credibility to obtain the evaluation credibility, and outputting the final patient state evaluation result based on the evaluation credibility ranking of the patient evaluation result;

[0012] Step S5: Record the outputs of the patient evaluation samples in the comprehensive evaluation model and each sub-model, collect patient feedback data, mark the results that deviate from the patient feedback data and the final patient evaluation results, and optimize the parameters of the evaluation credibility based on the marked records.

[0013] As a preferred technical solution of the present invention, the patient-related data is specifically:

[0014] Real-time patient data, including vital signs such as heart rate, blood pressure, oxygen saturation, and body temperature, are collected in real time through physiological monitoring equipment;

[0015] Anesthesia and surgery data, including anesthetic drug use, anesthesia type, surgery duration, and intraoperative complications, were obtained through anesthesia management logs;

[0016] Medical history data, including patients’ past medical history, comorbidities, and preoperative evaluation results, were obtained through patients’ electronic medical records;

[0017] Patient descriptive data, including postoperative pain perception, anxiety level, and satisfaction, were collected through questionnaires and interviews; expert descriptive data, including descriptions of abnormal manifestations of physical, psychological, and neurological states, were obtained through descriptions of medical expert patients;

[0018] All the above data are stored in the database, each patient's data is marked with the same ID, and data association is established.

[0019] As a preferred technical solution of the present invention, data preprocessing includes:

[0020] Preprocessing of objective data:

[0021] Data cleaning: clean the acquired objective data, identify and remove duplicate records, incorrect inputs and irrelevant data;

[0022] Feature segmentation: feature segmentation of objective data according to data characteristics. For numerical data, feature labels are divided according to numerical intervals. For text data, descriptive labels are set as feature labels for feature segmentation, and a feature label-data comparison table is constructed.

[0023] Data conversion: convert all objective data into the format of feature labels. Feature labels are objective parameters. In the comprehensive evaluation model, objective parameters are restored through the feature label-data comparison table to ensure data availability.

[0024] Outlier detection uses the interquartile range (IQR) to identify and mark outliers of objective parameters, confirms outliers, and corrects or removes outliers based on the confirmation results. The confirmation process includes classifying outliers according to the divided interquartile range intervals, and the classification results are absolute anomalies and relative anomalies. If the classification result is relative anomalies, the outliers are corrected by taking the average value of the data obtained at the previous moment and the next moment, otherwise, the outliers are removed.

[0025] Null value filling: For missing values ​​in objective parameters, the mean filling method is used to fill them;

[0026] The collected subjective data are preprocessed using a keyword extraction model, including:

[0027] Text preprocessing: cleaning the patient description data and expert description data, removing stop words and punctuation marks, and performing stem extraction;

[0028] Feature representation, converting the processed text into a computer-understandable format and applying word embedding methods to generate feature vectors;

[0029] Keyword scoring, using the TextRank algorithm to evaluate the importance of each word and generate a corresponding keyword list;

[0030] Select keywords, and extract the top N keywords from the patient and expert descriptions based on the scoring results to form patient description keywords and expert description keywords;

[0031] Construct subjective parameters and use the extracted keywords as subjective parameters to support the calculation of the subsequent comprehensive evaluation model.

[0032] As a preferred technical solution of the present invention, the comprehensive evaluation model includes:

[0033] The objective fuzzy evaluation layer is composed of a complete objective fuzzy evaluation model. It calculates the preliminary evaluation results of the patient's physical condition by performing fuzzy logic processing on the objective parameters corresponding to the prior data in the patient evaluation sample. The model considers the fuzziness and uncertainty of multiple physiological indicators and obtains the objective first evaluation as the prior indicator of the comprehensive evaluation model in the entire evaluation dimension.

[0034] The subjective evaluation layer and the subjective weight layer are composed of a complete subjective weight model. The subjective evaluation layer calculates the preliminary evaluation results of the patient's psychological state through subjective parameters and the objective first evaluation results output by the objective fuzzy evaluation model, where the subjective parameters include patient description keywords and expert description keywords; the subjective weight layer matches the psychological weights of the objective parameters corresponding to the prior data in the psychological state influence table according to the subjective evaluation results;

[0035] The objective optimization evaluation layer is composed of a complete objective optimization evaluation model. It uses the psychological weights generated by the subjective weight layer to perform weighted optimization on the objective parameters corresponding to the prior data, and combines the objective parameters corresponding to the posterior data to calculate the objective second evaluation results, thereby comprehensively evaluating the patient's physical and neurological status.

[0036] The comprehensive assessment layer is used to integrate the objective second assessment results and the subjective assessment results, comprehensively consider the physical, psychological and neurological assessment results, and output the final patient assessment results;

[0037] The data recording layer is used to record the output results of the comprehensive evaluation model and its sub-models, collect patient feedback data, mark deviation results, realize dynamic optimization and parameter adjustment of the model, and improve evaluation accuracy.

[0038] As a preferred technical solution of the present invention, the construction of the comprehensive evaluation model includes:

[0039] Collect a number of patient evaluation samples, add labels to the patient evaluation samples, and the content of the labels includes three aspects: physical state, psychological state and neural state; the patient evaluation samples are combined into a patient evaluation data set, and the patient evaluation data set is divided into a training set, a test set and a validation set;

[0040] Construct an initial objective fuzzy evaluation model, an initial subjective evaluation model, and an initial objective optimization evaluation model. The initial objective fuzzy evaluation model is based on SVM. The initial subjective evaluation model is based on BERT, and a subjective weight layer for matching psychological weights is added on the basis of BERT. The initial objective optimization evaluation model is based on RCNN, and an input layer is added before the input layer on the basis of RCNN to use psychological weights to weight the objective parameters corresponding to the input prior data.

[0041] For each initial sub-model, in each iterative training process, randomly select a set proportion of samples from the training set and the validation set in the patient evaluation data set to form the data set for this training, use the parameters corresponding to the sub-model to train the initial sub-model, and take the label corresponding to the sub-model as the target to train the optimized sub-model; send the patient evaluation test set to the optimized sub-model for accuracy evaluation, if the accuracy is within the set range, then the optimized sub-model output is used as the final sub-model, otherwise, use Bayesian optimization to optimize the hyperparameters of the sub-model until the accuracy of the sub-model is within the set range;

[0042] The three trained sub-models are integrated into the hidden layer of the initial comprehensive evaluation model, and a data comprehensive evaluation layer and a data recording layer are added to the hidden layer of the initial comprehensive evaluation model. The patient state matching table is stored in the comprehensive evaluation layer. The patient state matching table is obtained by statistically analyzing the state combinations in the three directions of physical, psychological and neural aspects of a large amount of historical patient state data. In the comprehensive evaluation layer, the state combinations in the three directions of physical, psychological and neural aspects of the patient are matched through the patient state matching table, and all state combinations whose matching rates exceed the set pre-signal are output as the patient state evaluation results.

[0043] As a preferred technical solution of the present invention, using Bayesian optimization to optimize the hyperparameters of the sub-model includes:

[0044] The Gaussian process is used as the prior probability to model the potential properties of the target sub-model to obtain the uncertainty measurement value of the target sub-model under different hyperparameter combinations. The Gaussian process formula is expressed as: , where GP is represented by Gaussian process, m(x) is the mean function, is the covariance function;

[0045] In the iterative process of hyperparameter optimization, the next parameter combination is selected by improving the expected amount EI to maximize the uncertainty of the objective function. The calculation formula of EI is: , where f(x) is the prior probability value corresponding to the hyperparameter combination of the target model, is the minimum value of the objective function;

[0046] After selecting the hyperparameter combination of the target model, choose whether to update the model hyperparameters based on the accuracy in the patient evaluation test set. Repeat the above steps until the set number of iterations is reached or the results converge to obtain the optimal hyperparameter combination of the model.

[0047] As a preferred technical solution of the present invention, the calculation of evaluating credibility includes:

[0048] The calculation formula for evaluating the credibility P is: , where M1, M2 and M3 are the proportions of the patient's physical-psychological, psychological-neurological and physical-neurological state combinations in the patient state matching table, , , To comprehensively evaluate the accuracy of the outputs of the three sub-models of the model, the data recording layer of the comprehensive evaluation model is used to obtain the outputs. , , They are the adjustment weights of the three state combinations respectively, and the sum of the three adjustment weights is 1.

[0049] As a preferred technical solution of the present invention, optimizing the parameters for evaluating credibility according to the marked records includes:

[0050] Get all the marked records, and whenever it is detected that there are more than a set number of records with only one state mismatch in the marked records, increase the two adjustment weights corresponding to the other state in the pairwise state combination.

[0051] A post-anesthesia patient status assessment system, comprising the following modules:

[0052] The data acquisition module is used to obtain relevant data of patients after anesthesia, including objective data and subjective data. The objective data includes real-time patient body data, anesthesia operation data and disease history data, among which the anesthesia operation data and disease history data are a posteriori data, and the real-time patient body data is a priori data; the subjective data includes patient description data and expert description data;

[0053] The data preprocessing module includes an objective processing unit and a subjective processing unit. The objective processing unit is used to preprocess the data related to the patients after anesthesia, and to perform data cleaning, feature division, data conversion, abnormal value detection and null value filling on the objective data, so as to obtain the feature data of each sub-type data in the objective data and use them as objective parameters; the subjective processing unit is used to send the subjective data to the keyword extraction model for calculation, obtain the patient description keywords and expert description keywords of the subjective data, and use them as subjective parameters; and is also used to construct the objective parameters and subjective parameters into patient evaluation samples;

[0054] The comprehensive evaluation module includes an objective fuzzy evaluation unit, a subjective weight unit, an objective optimization evaluation unit and a comprehensive evaluation unit; the objective fuzzy evaluation unit is used to calculate the objective parameters corresponding to the prior data of the patient evaluation sample to obtain an objective first evaluation result; the subjective weight unit is used to calculate the subjective parameters and the objective first evaluation result to obtain the psychological weights of the subjective evaluation result and the objective parameters corresponding to the prior data; the objective optimization evaluation unit is used to weight the objective parameters corresponding to the prior data using the psychological weights, and send the weighted results and the objective parameters corresponding to the posterior data into the objective optimization evaluation model for calculation to obtain the objective second evaluation result; the comprehensive evaluation unit is used to match the objective second evaluation result with the subjective evaluation result, and output multiple patient evaluation results;

[0055] An evaluation verification module is used to obtain the state pattern credibility of the patient evaluation result through the matching degree of the patient's physical state, psychological state and neural state, obtain the deviation credibility based on the accuracy of the patient evaluation result, perform weighted calculation on the state pattern credibility and the deviation credibility to obtain the evaluation credibility, and output the final patient state evaluation result based on the evaluation credibility ranking of the patient evaluation result;

[0056] The record optimization module is used to record the outputs of patient evaluation samples in the comprehensive evaluation model and each sub-model, collect patient feedback data, mark the results that deviate from the patient feedback data and the final patient evaluation results, and optimize the parameters of the evaluation credibility based on the marked records.

[0057] The present invention has the following advantages:

[0058] The present invention provides a comprehensive assessment of the patient's physical, psychological and neurological status by combining objective data (vital signs, anesthesia surgery data and medical history) with subjective data (descriptions from patients and experts), thereby ensuring the comprehensiveness and accuracy of the assessment results; through the combination of multiple models (objective fuzzy assessment model and subjective weight model), the uncertainty and ambiguity of the data are effectively handled, thereby improving the accuracy and reliability of the patient status assessment.

[0059] The present invention records the deviation between patient feedback and evaluation results, continuously optimizes evaluation parameters, and forms a virtuous feedback loop, thereby continuously improving the performance of the evaluation model; through detailed status assessment, potential complications and psychological problems are discovered in a timely manner, so that corresponding intervention measures can be taken to improve patient postoperative safety and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The present invention is a schematic diagram of the structure of a system for evaluating the status of a patient after anesthesia. DETAILED DESCRIPTION

[0061] In order to enable persons 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.

[0062] Embodiment 1, a method for evaluating the status of a patient after anesthesia, comprising the following steps:

[0063] Step S1: Obtaining post-anesthesia patient-related data, including objective data and subjective data. The objective data includes real-time patient body data, anesthesia surgery data and disease history data, wherein the anesthesia surgery data and disease history data are a posteriori data, and the real-time patient body data is a priori data; the subjective data includes patient description data and expert description data;

[0064] The patient-related data are as follows:

[0065] Real-time patient data, including vital signs such as heart rate, blood pressure, oxygen saturation, and body temperature, are collected in real time through physiological monitoring equipment;

[0066] Anesthesia and surgery data, including anesthetic drug use, anesthesia type, surgery duration, and intraoperative complications, were obtained through anesthesia management logs;

[0067] Medical history data, including patients’ past medical history, comorbidities, and preoperative evaluation results, were obtained through patients’ electronic medical records;

[0068] Patient descriptive data, including postoperative pain perception, anxiety level, and satisfaction, were collected through questionnaires and interviews; expert descriptive data, including descriptions of abnormal manifestations of physical, psychological, and neurological states, were obtained through descriptions of medical expert patients;

[0069] All the above data are stored in the database, each patient's data is marked with the same ID, and data association is established.

[0070] Step S2: Preprocess the data related to the patients after anesthesia, perform data cleaning, feature division, data conversion, outlier detection and null value filling on the objective data, obtain the feature data of each subtype of the objective data, and use them as objective parameters; send the subjective data to the keyword extraction model for calculation, obtain the patient description keywords and expert description keywords of the subjective data, and use them as subjective parameters; construct the objective parameters and subjective parameters into patient evaluation samples;

[0071] Data preprocessing includes:

[0072] Preprocessing of objective data:

[0073] Data cleaning: clean the acquired objective data, identify and remove duplicate records, incorrect inputs and irrelevant data;

[0074] Feature segmentation: feature segmentation of objective data according to data characteristics. For numerical data, feature labels are divided according to numerical intervals. For text data, descriptive labels are set as feature labels for feature segmentation, and a feature label-data comparison table is constructed.

[0075] Data conversion: convert all objective data into the format of feature labels. Feature labels are objective parameters. In the comprehensive evaluation model, objective parameters are restored through the feature label-data comparison table to ensure data availability.

[0076] Outlier detection uses the interquartile range (IQR) to identify and mark outliers of objective parameters, confirms outliers, and corrects or removes outliers based on the confirmation results. The confirmation process includes classifying outliers according to the divided interquartile range intervals, and the classification results are absolute anomalies and relative anomalies. If the classification result is relative anomalies, the outliers are corrected by taking the average value of the data obtained at the previous moment and the next moment, otherwise, the outliers are removed.

[0077] Null value filling: For missing values ​​in objective parameters, the mean filling method is used to fill them;

[0078] The collected subjective data are preprocessed using a keyword extraction model, including:

[0079] Text preprocessing: cleaning the patient description data and expert description data, removing stop words and punctuation marks, and performing stem extraction;

[0080] Feature representation, converting the processed text into a computer-understandable format and applying word embedding methods to generate feature vectors;

[0081] Keyword scoring, using the TextRank algorithm to evaluate the importance of each word and generate a corresponding keyword list;

[0082] Select keywords, and extract the top N keywords from the patient and expert descriptions based on the scoring results to form patient description keywords and expert description keywords;

[0083] Construct subjective parameters and use the extracted keywords as subjective parameters to support the calculation of the subsequent comprehensive evaluation model.

[0084] Step S3: Send the patient evaluation sample into the comprehensive evaluation model for calculation to obtain multiple patient evaluation results. The comprehensive evaluation model is constructed by combining three sub-models, namely, an objective fuzzy evaluation model, a subjective weight model, and an objective optimization evaluation model; the objective fuzzy evaluation model calculates the objective parameters corresponding to the prior data of the patient evaluation sample to obtain an objective first evaluation result, and the objective first evaluation result is an evaluation of the patient's physical state; the subjective weight model calculates the subjective parameters and the objective first evaluation result to obtain the psychological weights of the subjective evaluation result and the objective parameters corresponding to the prior data, and the subjective evaluation result is an evaluation of the patient's psychological state; the objective parameters corresponding to the prior data are weighted using the psychological weights, and the weighted results and the objective parameters corresponding to the posterior data are sent to the objective optimization evaluation model for calculation to obtain an objective second evaluation result, and the objective second evaluation result is an evaluation of the patient's physical state and neural state; the objective second evaluation result and the subjective evaluation result are sent to the comprehensive evaluation layer of the comprehensive evaluation model for matching, and multiple patient evaluation results are output, and the patient evaluation results are evaluations of the patient's physical, psychological, and neural directions;

[0085] The comprehensive assessment model includes:

[0086] The objective fuzzy evaluation layer is composed of a complete objective fuzzy evaluation model. It calculates the preliminary evaluation results of the patient's physical condition by performing fuzzy logic processing on the objective parameters corresponding to the prior data in the patient evaluation sample. The model considers the fuzziness and uncertainty of multiple physiological indicators and obtains the objective first evaluation as the prior indicator of the comprehensive evaluation model in the entire evaluation dimension.

[0087] The subjective evaluation layer and the subjective weight layer are composed of a complete subjective weight model. The subjective evaluation layer calculates the preliminary evaluation results of the patient's psychological state through subjective parameters and the objective first evaluation results output by the objective fuzzy evaluation model, where the subjective parameters include patient description keywords and expert description keywords; the subjective weight layer matches the psychological weights of the objective parameters corresponding to the prior data in the psychological state influence table according to the subjective evaluation results;

[0088] The objective optimization evaluation layer is composed of a complete objective optimization evaluation model. It uses the psychological weights generated by the subjective weight layer to perform weighted optimization on the objective parameters corresponding to the prior data, and combines the objective parameters corresponding to the posterior data to calculate the objective second evaluation results, thereby comprehensively evaluating the patient's physical and neurological status.

[0089] The comprehensive assessment layer is used to integrate the objective second assessment results and the subjective assessment results, comprehensively consider the physical, psychological and neurological assessment results, and output the final patient assessment results;

[0090] The data recording layer is used to record the output results of the comprehensive evaluation model and its sub-models, collect patient feedback data, mark deviation results, realize dynamic optimization and parameter adjustment of the model, and improve evaluation accuracy.

[0091] The construction of the comprehensive evaluation model includes:

[0092] Collect a number of patient evaluation samples, add labels to the patient evaluation samples, and the content of the labels includes three aspects: physical state, psychological state, and neural state; the patient evaluation samples are combined into a patient evaluation data set, and the patient evaluation data set is divided into a training set, a test set, and a validation set;

[0093] Construct an initial objective fuzzy evaluation model, an initial subjective evaluation model, and an initial objective optimization evaluation model. The initial objective fuzzy evaluation model is based on SVM. The initial subjective evaluation model is based on BERT, and a subjective weight layer for matching psychological weights is added on the basis of BERT. The initial objective optimization evaluation model is based on RCNN, and an input layer is added before the input layer on the basis of RCNN to use psychological weights to weight the objective parameters corresponding to the input prior data.

[0094] For each initial sub-model, in each iterative training process, randomly select a set proportion of samples from the training set and the validation set in the patient evaluation data set to form the data set for this training, use the parameters corresponding to the sub-model to train the initial sub-model, and take the label corresponding to the sub-model as the target to train the optimized sub-model; send the patient evaluation test set to the optimized sub-model for accuracy evaluation, if the accuracy is within the set range, then the optimized sub-model output is used as the final sub-model, otherwise, use Bayesian optimization to optimize the hyperparameters of the sub-model until the accuracy of the sub-model is within the set range;

[0095] The three trained sub-models are integrated into the hidden layer of the initial comprehensive evaluation model, and a data comprehensive evaluation layer and a data recording layer are added to the hidden layer of the initial comprehensive evaluation model. The patient state matching table is stored in the comprehensive evaluation layer. The patient state matching table is obtained by statistically analyzing the state combinations in the three directions of physical, psychological and neural aspects of a large amount of historical patient state data. In the comprehensive evaluation layer, the state combinations in the three directions of physical, psychological and neural aspects of the patient are matched through the patient state matching table, and all state combinations whose matching rates exceed the set pre-signal are output as the patient state evaluation results.

[0096] The optimization of hyperparameters of sub-models using Bayesian optimization includes:

[0097] The Gaussian process is used as the prior probability to model the potential properties of the target sub-model to obtain the uncertainty measurement value of the target sub-model under different hyperparameter combinations. The Gaussian process formula is expressed as: , where GP is represented by Gaussian process, m(x) is the mean function, is the covariance function;

[0098] In the iterative process of hyperparameter optimization, the next parameter combination is selected by improving the expected amount EI to maximize the uncertainty of the objective function. The calculation formula of EI is: , where f(x) is the prior probability value corresponding to the hyperparameter combination of the target model, is the minimum value of the objective function;

[0099] After selecting the hyperparameter combination of the target model, choose whether to update the model hyperparameters based on the accuracy in the patient evaluation test set. Repeat the above steps until the set number of iterations is reached or the results converge to obtain the optimal hyperparameter combination of the model.

[0100] Step S4: obtaining the state pattern credibility of the patient evaluation result through the matching degree of the patient's physical state, psychological state and neural state, obtaining the deviation credibility based on the accuracy of the patient evaluation result, performing weighted calculation on the state pattern credibility and the deviation credibility to obtain the evaluation credibility, and outputting the final patient state evaluation result based on the evaluation credibility ranking of the patient evaluation result;

[0101] Calculations to assess credibility include:

[0102] The calculation formula for evaluating the credibility P is: , where M1, M2 and M3 are the proportions of the patient's physical-psychological, psychological-neurological and physical-neurological state combinations in the patient state matching table, , , To comprehensively evaluate the accuracy of the outputs of the three sub-models of the model, the data recording layer of the comprehensive evaluation model is used to obtain the outputs. , , They are the adjustment weights of the three state combinations respectively, and the sum of the three adjustment weights is 1.

[0103] Step S5: Record the outputs of the patient evaluation samples in the comprehensive evaluation model and each sub-model, collect patient feedback data, mark the results that deviate from the patient feedback data and the final patient evaluation results, and optimize the parameters of the evaluation credibility based on the marked records.

[0104] As a preferred technical solution of the present invention, optimizing the parameters for evaluating credibility according to the marked records includes:

[0105] Get all the marked records, and whenever it is detected that there are more than a set number of records with only one state mismatch in the marked records, increase the two adjustment weights corresponding to the other state in the pairwise state combination.

[0106] Embodiment 2, a system for evaluating the status of a patient after anesthesia, using the method for evaluating the status of a patient after anesthesia as described above, see Figure 1 As shown, it includes the following modules:

[0107] The data acquisition module is used to obtain relevant data of patients after anesthesia, including objective data and subjective data. The objective data includes real-time patient body data, anesthesia operation data and disease history data, among which the anesthesia operation data and disease history data are a posteriori data, and the real-time patient body data is a priori data; the subjective data includes patient description data and expert description data;

[0108] The data preprocessing module includes an objective processing unit and a subjective processing unit. The objective processing unit is used to preprocess the data related to the patients after anesthesia, and to perform data cleaning, feature division, data conversion, abnormal value detection and null value filling on the objective data, so as to obtain the feature data of each sub-type data in the objective data and use them as objective parameters; the subjective processing unit is used to send the subjective data to the keyword extraction model for calculation, obtain the patient description keywords and expert description keywords of the subjective data, and use them as subjective parameters; and is also used to construct the objective parameters and subjective parameters into patient evaluation samples;

[0109] The comprehensive evaluation module includes an objective fuzzy evaluation unit, a subjective weight unit, an objective optimization evaluation unit and a comprehensive evaluation unit; the objective fuzzy evaluation unit is used to calculate the objective parameters corresponding to the prior data of the patient evaluation sample to obtain an objective first evaluation result; the subjective weight unit is used to calculate the subjective parameters and the objective first evaluation result to obtain the psychological weights of the subjective evaluation result and the objective parameters corresponding to the prior data; the objective optimization evaluation unit is used to weight the objective parameters corresponding to the prior data using the psychological weights, and send the weighted results and the objective parameters corresponding to the posterior data into the objective optimization evaluation model for calculation to obtain the objective second evaluation result; the comprehensive evaluation unit is used to match the objective second evaluation result with the subjective evaluation result, and output multiple patient evaluation results;

[0110] An evaluation verification module is used to obtain the state pattern credibility of the patient evaluation result through the matching degree of the patient's physical state, psychological state and neural state, obtain the deviation credibility based on the accuracy of the patient evaluation result, perform weighted calculation on the state pattern credibility and the deviation credibility to obtain the evaluation credibility, and output the final patient state evaluation result based on the evaluation credibility ranking of the patient evaluation result;

[0111] The record optimization module is used to record the outputs of patient evaluation samples in the comprehensive evaluation model and each sub-model, collect patient feedback data, mark the results that deviate from the patient feedback data and the final patient evaluation results, and optimize the parameters of the evaluation credibility based on the marked records.

[0112] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for evaluating the status of a patient after anesthesia, characterized in that: The following steps are involved: Step S1: Obtaining post-anesthesia patient-related data, including objective data and subjective data. The objective data includes real-time patient body data, anesthesia surgery data and disease history data, wherein the anesthesia surgery data and disease history data are a posteriori data, and the real-time patient body data is a priori data; the subjective data includes patient description data and expert description data; Step S2: preprocess the patient-related data after anesthesia, perform data cleaning, feature segmentation, data conversion, outlier detection and null value filling on the objective data, obtain the feature data of each subtype of the objective data, and use them as objective parameters; send the subjective data to the keyword extraction model for calculation, obtain the patient description keywords and expert description keywords of the subjective data, and use them as subjective parameters; construct the objective parameters and subjective parameters into patient evaluation samples; Step S3: Send the patient evaluation samples to the comprehensive evaluation model for calculation to obtain multiple patient evaluation results. The comprehensive evaluation model is constructed by combining three sub-models, namely, an objective fuzzy evaluation model, a subjective weight model, and an objective optimization evaluation model; The objective fuzzy evaluation model calculates the objective parameters corresponding to the prior data of the patient evaluation sample to obtain the objective first evaluation result, which is the evaluation of the patient's physical condition; The subjective weight model calculates the subjective parameters and the objective first evaluation results to obtain the psychological weights of the subjective evaluation results and the objective parameters corresponding to the prior data. The subjective evaluation results are the evaluation of the patient's psychological state. The objective parameters corresponding to the prior data are weighted using psychological weights, and the weighted results and the objective parameters corresponding to the posterior data are sent to the objective optimization evaluation model for calculation to obtain an objective second evaluation result, which is an evaluation of the patient's physical state and neurological state; The objective second assessment result and the subjective assessment result are sent to the comprehensive assessment layer of the comprehensive assessment model for matching, and multiple patient assessment results are output. The patient assessment results are assessments of the patient's physical, psychological and neurological aspects; Step S4: obtaining the state pattern credibility of the patient evaluation result through the matching degree of the patient's physical state, psychological state and neural state, obtaining the deviation credibility based on the accuracy of the patient evaluation result, performing weighted calculation on the state pattern credibility and the deviation credibility to obtain the evaluation credibility, and outputting the final patient state evaluation result based on the evaluation credibility ranking of the patient evaluation result; Step S5: Record the outputs of the patient evaluation samples in the comprehensive evaluation model and each sub-model, collect patient feedback data, mark the results that deviate from the patient feedback data and the final patient evaluation results, and optimize the parameters of the evaluation credibility based on the marked records.

2. A method for evaluating the status of a patient after anesthesia according to claim 1, characterized in that: The patient-related data are as follows: Real-time patient data, including vital signs such as heart rate, blood pressure, oxygen saturation, and body temperature, are collected in real time through physiological monitoring equipment; Anesthesia and surgery data, including anesthetic drug use, anesthesia type, surgery duration, and intraoperative complications, were obtained through anesthesia management logs; Medical history data, including patients’ past medical history, comorbidities, and preoperative evaluation results, were obtained through patients’ electronic medical records; Patient descriptive data, including postoperative pain perception, anxiety level, and satisfaction, were collected through questionnaires and interviews; expert descriptive data, including descriptions of abnormal manifestations of physical, psychological, and neurological states, were obtained through descriptions of medical expert patients; All the above data are stored in the database, each patient's data is marked with the same ID, and data association is established.

3. A method for evaluating the status of a patient after anesthesia according to claim 1, characterized in that: Data preprocessing includes: Preprocessing of objective data: Data cleaning: clean the acquired objective data, identify and remove duplicate records, incorrect inputs and irrelevant data; Feature segmentation: feature segmentation of objective data according to data characteristics. For numerical data, feature labels are divided according to numerical intervals. For text data, descriptive labels are set as feature labels for feature segmentation, and a feature label-data comparison table is constructed. Data conversion: convert all objective data into the format of feature labels. Feature labels are objective parameters. In the comprehensive evaluation model, objective parameters are restored through the feature label-data comparison table to ensure data availability. Outlier detection uses the interquartile range (IQR) to identify and mark outliers of objective parameters, confirms outliers, and corrects or removes outliers based on the confirmation results. The confirmation process includes classifying outliers according to the divided interquartile range intervals, and the classification results are absolute anomalies and relative anomalies. If the classification result is relative anomalies, the outliers are corrected by taking the average value of the data obtained at the previous moment and the next moment, otherwise, the outliers are removed. Null value filling: For missing values ​​in objective parameters, the mean filling method is used to fill them; The collected subjective data are preprocessed using a keyword extraction model, including: Text preprocessing: cleaning the patient description data and expert description data, removing stop words and punctuation marks, and performing stem extraction; Feature representation, converting the processed text into a computer-understandable format and applying word embedding methods to generate feature vectors; Keyword scoring, using the TextRank algorithm to evaluate the importance of each word and generate a corresponding keyword list; Select keywords, and extract the top N keywords from the patient and expert descriptions based on the scoring results to form patient description keywords and expert description keywords; Construct subjective parameters and use the extracted keywords as subjective parameters to support the calculation of the subsequent comprehensive evaluation model.

4. A method for evaluating the status of a patient after anesthesia according to claim 1, characterized in that: The comprehensive assessment model includes: The objective fuzzy evaluation layer is composed of a complete objective fuzzy evaluation model. It calculates the preliminary evaluation results of the patient's physical condition by performing fuzzy logic processing on the objective parameters corresponding to the prior data in the patient evaluation sample. The model considers the fuzziness and uncertainty of multiple physiological indicators and obtains the objective first evaluation as the prior indicator of the comprehensive evaluation model in the entire evaluation dimension. The subjective evaluation layer and the subjective weight layer are composed of a complete subjective weight model. The subjective evaluation layer calculates the preliminary evaluation results of the patient's psychological state through subjective parameters and the objective first evaluation results output by the objective fuzzy evaluation model, where the subjective parameters include patient description keywords and expert description keywords; the subjective weight layer matches the psychological weights of the objective parameters corresponding to the prior data in the psychological state influence table according to the subjective evaluation results; The objective optimization evaluation layer is composed of a complete objective optimization evaluation model. It uses the psychological weights generated by the subjective weight layer to perform weighted optimization on the objective parameters corresponding to the prior data, and combines the objective parameters corresponding to the posterior data to calculate the objective second evaluation results, thereby comprehensively evaluating the patient's physical and neurological status. The comprehensive assessment layer is used to integrate the objective second assessment results and the subjective assessment results, comprehensively consider the physical, psychological and neurological assessment results, and output the final patient assessment results; The data recording layer is used to record the output results of the comprehensive evaluation model and its sub-models, collect patient feedback data, mark deviation results, realize dynamic optimization and parameter adjustment of the model, and improve evaluation accuracy.

5. A method for evaluating the status of a patient after anesthesia according to claim 4, characterized in that: The construction of the comprehensive evaluation model includes: Collect a number of patient evaluation samples, add labels to the patient evaluation samples, and the content of the labels includes three aspects: physical state, psychological state, and neural state; the patient evaluation samples are combined into a patient evaluation data set, and the patient evaluation data set is divided into a training set, a test set, and a validation set; Construct an initial objective fuzzy evaluation model, an initial subjective evaluation model, and an initial objective optimization evaluation model. The initial objective fuzzy evaluation model is based on SVM. The initial subjective evaluation model is based on BERT, and a subjective weight layer for matching psychological weights is added on the basis of BERT. The initial objective optimization evaluation model is based on RCNN, and an input layer is added before the input layer on the basis of RCNN to use psychological weights to weight the objective parameters corresponding to the input prior data. For each initial sub-model, in each iterative training process, randomly select a set proportion of samples from the training set and the validation set in the patient evaluation data set to form the data set for this training, use the parameters corresponding to the sub-model to train the initial sub-model, and take the label corresponding to the sub-model as the target to train the optimized sub-model; send the patient evaluation test set to the optimized sub-model for accuracy evaluation, if the accuracy is within the set range, then the optimized sub-model output is used as the final sub-model, otherwise, use Bayesian optimization to optimize the hyperparameters of the sub-model until the accuracy of the sub-model is within the set range; The three trained sub-models are integrated into the hidden layer of the initial comprehensive evaluation model, and a data comprehensive evaluation layer and a data recording layer are added to the hidden layer of the initial comprehensive evaluation model. The patient state matching table is stored in the comprehensive evaluation layer. The patient state matching table is obtained by statistically analyzing the state combinations in the three directions of physical, psychological and neural aspects of a large amount of historical patient state data. In the comprehensive evaluation layer, the state combinations in the three directions of physical, psychological and neural aspects of the patient are matched through the patient state matching table, and all state combinations whose matching rates exceed the set pre-signal are output as the patient state evaluation results.

6. A method for evaluating the status of a patient after anesthesia according to claim 5, characterized in that: The optimization of hyperparameters of sub-models using Bayesian optimization includes: The Gaussian process is used as the prior probability to model the potential properties of the target sub-model to obtain the uncertainty measurement value of the target sub-model under different hyperparameter combinations. The Gaussian process formula is expressed as: , where GP is represented by Gaussian process, m(x) is the mean function, is the covariance function; In the iterative process of hyperparameter optimization, the next parameter combination is selected by improving the expected amount EI to maximize the uncertainty of the objective function. The calculation formula of EI is: , where f(x) is the prior probability value corresponding to the hyperparameter combination of the target model, is the minimum value of the objective function; After selecting the hyperparameter combination of the target model, choose whether to update the model hyperparameters based on the accuracy in the patient evaluation test set. Repeat the above steps until the set number of iterations is reached or the results converge to obtain the optimal hyperparameter combination of the model.

7. A method for evaluating the status of a patient after anesthesia according to claim 1, characterized in that: Calculations to assess credibility include: The calculation formula for evaluating the credibility P is: , where M1, M2 and M3 are the proportions of the patient's physical-psychological, psychological-neurological and physical-neurological state combinations in the patient state matching table, , , To comprehensively evaluate the accuracy of the outputs of the three sub-models of the model, the data recording layer of the comprehensive evaluation model is used to obtain the outputs. , , They are the adjustment weights of the three state combinations respectively, and the sum of the three adjustment weights is 1.

8. A method for evaluating the status of a patient after anesthesia according to claim 1, characterized in that: Optimization of parameters for evaluating credibility based on marked records includes: Get all the marked records, and whenever it is detected that there are more than a set number of records with only one state mismatch in the marked records, increase the two adjustment weights corresponding to the other state in the pairwise state combination.

9. A system for evaluating the status of a patient after anesthesia, characterized in that: The system is used to implement a method for evaluating the state of a patient after anesthesia according to any one of claims 1 to 8 above, and the system for evaluating the state of a patient after anesthesia comprises the following modules: The data acquisition module is used to obtain relevant data of patients after anesthesia, including objective data and subjective data. The objective data includes real-time patient body data, anesthesia operation data and disease history data, among which the anesthesia operation data and disease history data are a posteriori data, and the real-time patient body data is a priori data; the subjective data includes patient description data and expert description data; The data preprocessing module includes an objective processing unit and a subjective processing unit. The objective processing unit is used to preprocess the data related to the patients after anesthesia, and to perform data cleaning, feature division, data conversion, abnormal value detection and null value filling on the objective data, so as to obtain the feature data of each sub-type data in the objective data and use them as objective parameters; the subjective processing unit is used to send the subjective data to the keyword extraction model for calculation, obtain the patient description keywords and expert description keywords of the subjective data, and use them as subjective parameters; and is also used to construct the objective parameters and subjective parameters into patient evaluation samples; The comprehensive evaluation module includes an objective fuzzy evaluation unit, a subjective weight unit, an objective optimization evaluation unit and a comprehensive evaluation unit; the objective fuzzy evaluation unit is used to calculate the objective parameters corresponding to the prior data of the patient evaluation sample to obtain the objective first evaluation result; the subjective weight unit is used to calculate the subjective parameters and the objective first evaluation result to obtain the psychological weight of the subjective evaluation result and the objective parameters corresponding to the prior data; The objective optimization evaluation unit is used to weight the objective parameters corresponding to the prior data using the psychological weights, and send the weighted results and the objective parameters corresponding to the posterior data into the objective optimization evaluation model for calculation to obtain the objective second evaluation results; the comprehensive evaluation unit is used to match the objective second evaluation results with the subjective evaluation results, and output multiple patient evaluation results; An evaluation verification module is used to obtain the state pattern credibility of the patient evaluation result through the matching degree of the patient's physical state, psychological state and neural state, obtain the deviation credibility based on the accuracy of the patient evaluation result, perform weighted calculation on the state pattern credibility and the deviation credibility to obtain the evaluation credibility, and output the final patient state evaluation result based on the evaluation credibility ranking of the patient evaluation result; The record optimization module is used to record the outputs of patient evaluation samples in the comprehensive evaluation model and each sub-model, collect patient feedback data, mark the results that deviate from the patient feedback data and the final patient evaluation results, and optimize the parameters of the evaluation credibility based on the marked records.

Images