Anesthesia recovery state intelligent assessment method and system

Abstract

The present invention discloses an intelligent evaluation method and system for anesthesia recovery status, which relates to the field of anesthesia evaluation technology. An intelligent evaluation system for anesthesia recovery status includes: a data acquisition module, a status evaluation module, a recovery sample construction module, a sample evaluation model module, an adjustment strategy generation module, a data statistics and management module, and a result output module. The present invention uses intelligent devices to collect multi-dimensional physiological data of patients in real time, and combines the sample evaluation model of segment matching and long segment matching to achieve accurate real-time evaluation of the patient's recovery status; through dynamically optimized sample areas, combined with the patient's basic information, the samples are grouped and clustered for analysis, and the individual characteristics of the patient are dynamically loaded and matched to ensure that the evaluation model can adapt to the individual differences of different patients, significantly improving the adaptability and robustness of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of anesthesia assessment, and in particular to an intelligent assessment method and system for anesthesia recovery status. Background Art

[0002] With the advancement of modern medical technology, anesthesia technology has become an indispensable part of the surgical process. The anesthesia recovery status of patients after surgery directly affects their life safety and rehabilitation effect. Therefore, the scientific evaluation of the patient's anesthesia recovery status is of great significance. At present, the evaluation of anesthesia recovery status mainly relies on the clinical experience of medical personnel and simple physiological indicator monitoring. However, this traditional method has the following shortcomings: single data reliance, poor real-time performance, lack of personalization, subjectivity of evaluation results, and lack of dynamic adjustment strategies;

[0003] How to use modern intelligent technologies to achieve accurate assessment, dynamic prediction, and personalized adjustment of anesthesia recovery status has become an important research direction. In recent years, with the development of big data, machine learning, and smart devices, data-driven intelligent assessment methods have provided the possibility of solving the above problems. Summary of the Invention

[0004] In response to the above problems, the present invention proposes an intelligent assessment method and system for anesthesia recovery status, which addresses the shortcomings of traditional assessment methods through an innovative multi-level sample assessment model, a dynamic category area construction mechanism, and a personalized adjustment strategy generation method.

[0005] An intelligent evaluation method for anesthesia recovery status comprises the following steps:

[0006] Step S1: Data collection: The patient's physiological data is collected in real time through smart devices, and the collected data is de-noised, standardized, and time-aligned; the physiological data includes but is not limited to heart rate, blood pressure, body temperature, respiratory rate, and blood oxygen saturation;

[0007] Step S2: Single state assessment, based on the real-time collected physiological data and the patient, extracting the patient's physiological characteristics, and evaluating and calculating the patient's current anesthesia recovery state according to the preset assessment criteria based on the patient's basic information to obtain a real-time recovery value;

[0008] Step S3: Constructing a recovery sample: Matching the corresponding recovery stage according to the evaluation results of a single state, and constructing a recovery sample based on the results of all single state evaluations; calculating the recovery comparison value by comparing the physiological data of the current recovery stage with the corresponding standard physiological data; the final recovery sample includes the patient's basic information, each recovery stage and the corresponding multiple sets of physiological data, real-time recovery value and recovery comparison value;

[0009] Step S4: Input the restored sample into the sample evaluation model for evaluation calculation to obtain the sample evaluation result and the restoration prediction result; after the restored sample is input into the sample evaluation model, the category area is selected and the segment matching calculation and the long segment matching calculation are performed, and the two types of calculation results are weighted to obtain the final sample evaluation result and the restoration prediction result;

[0010] Step S5: Recommendation of adjustment strategy: generating a personalized recovery adjustment strategy based on the sample evaluation results and recovery prediction results;

[0011] Step S6: Data statistics and management, statistics, storage and management of all patients' recovery samples, sample evaluation results, recovery prediction results and recovery adjustment strategies, support data backup, query and long-term tracking analysis, while ensuring data security and privacy protection.

[0012] As a preferred technical solution of the present invention, the step of calculating the real-time recovery value includes:

[0013] Extract the patient’s physiological data collected in real time, recorded as ,in represents the patient’s i-th physiological data;

[0014] Compare the real-time collected physiological data with the preset healthy physiological data to obtain the difference value of each physiological data item , the calculation formula is: ;in, is the healthy physiological data corresponding to the i-th physiological data item;

[0015] The difference value for each physiological data item A nonlinear function is used for conversion, such as the following formula:

[0016] ;

[0017] The nonlinear function is selected according to the type of physiological data and the size of the difference value. The selection conditions of the four functions in the formula are: For continuous changes, For the regression trend, Less than the change threshold and greater than the change threshold;

[0018] Perform weighted summation on the difference values ​​of all converted physiological data items to obtain the initial real-time recovery value , the calculation formula is: ,in is the weighting coefficient of the i-th physiological data item;

[0019] Initial real-time recovery value Perform regression analysis and correction to obtain the final real-time recovery value. The calculation formula is:

[0020] ,in is the preset regression coefficient used to correct the recovery value.

[0021] As a preferred technical solution of the present invention, the construction of the recovery sample includes the following steps:

[0022] Matching of single state assessment results: Based on the patient's real-time recovery value of a single state, the patient's current physiological data is matched with the standard physiological data of the corresponding recovery stage to determine the patient's recovery stage;

[0023] Calculation of the recovery comparison value: extract the physiological data of all single states in the current recovery stage, calculate the overall average physiological data value and the standard average physiological data value of the standard recovery stage; compare the physiological data standardization of the current recovery stage with that of the standard recovery stage, and calculate the recovery comparison value of the current recovery stage , the calculation formula is: ,in represents the average value of the i-th physiological data in the current recovery stage, represents the average value of the i-th physiological data in the standard recovery phase, n represents the total number of physiological data items, and A represents the adjustment coefficient;

[0024] Based on the results of a single status assessment and the recovery comparison value of the recovery stage, a recovery sample of the patient is generated; and the recovery sample is stored in the form of structured data.

[0025] As a preferred technical solution of the present invention, the structure of the sample evaluation model includes:

[0026] The input layer is used to receive the patient's physiological data and basic information, and convert it into the feature vector form X={x1,x2,…,x n}, where n represents the feature dimension;

[0027] The pairing layer is used to select a category area based on the patient's basic information, classify and map the input feature vector X through the mapping function M() of the patient's basic information, and determine the category area that matches the input patient information; the basic information includes anesthesia type, disease type, age range, gender, and chronic disease history;

[0028] Category area units include:

[0029] The segmented matching layer uses a local matching network based on the attention mechanism to extract representative local features in the feature vector, and performs local feature matching on the key physiological indicators of the patient's recovery data to obtain the local matching result R s;

[0030] The long segment matching layer, based on the recurrent neural network, performs global modeling on the time series features of the input data and obtains the matching results R of the input features in the time dimension. l ;

[0031] The result fitting layer is used to fit and integrate the matching results of the segment matching layer and the long segment matching layer, through the weighting function: F(R s ,R l )=w s *R s +w l *R l , calculate the intermediate evaluation value R m ; where w s and w l is the weight parameter of the matching result, and satisfies w s +w l =1;

[0032] Weighted output layer, used to evaluate the intermediate value R m Combined with the specific weight W of the patient's current recovery stage p , generate the final sample evaluation result R f and predicted recovery results P r .

[0033] As a preferred technical solution of the present invention, obtaining the category area includes:

[0034] By preprocessing historical recovery samples, the collected original sample data is cleaned, denoised, standardized, and time-aligned to generate a preprocessed sample set;

[0035] Based on a combination of one or more attributes in the patient's basic information, the preprocessed sample set is grouped and divided into multiple sample subsets, wherein the number of samples in each of the grouped sample subsets is greater than a preset threshold;

[0036] The density clustering algorithm is used to cluster the sample subsets and the number of samples in each cluster category is calculated;

[0037] Filter the clustering results and take the sample set whose valid category number exceeds the set value as the valid sample set; the valid category number is the number of samples of the corresponding cluster category exceeding the set ratio;

[0038] Model training is performed separately for each valid sample set, and a category area model is constructed based on the support vector machine. The corresponding training process includes:

[0039] Using the physiological data and basic information of the valid sample set as input, the model parameters are optimized to obtain the feature mapping function of each category area model;

[0040] Use the validation set to evaluate the category area model, optimize the model's hyperparameters, and ensure the model's accuracy and robustness;

[0041] Finally, all category area models are combined to form a category area, which is used to select the corresponding category area for calculation in the sample evaluation model according to the basic information of the input patient.

[0042] As a preferred technical solution of the present invention, obtaining the weight parameter of the matching result includes:

[0043] Obtain segment matching results and long segment matching results, use the segment matching layer to extract local features of the input feature X, and generate local matching results R s , the calculation formula is: R s =f s (X,W s ), where f s () is the piecewise matching function, W s is the weight parameter of the segment matching layer;

[0044] The long segment matching layer is used to globally model the overall time series features of the input feature X and generate the global matching result R l , the calculation formula is: R l =f l (X,W l ), where f l () is the long segment matching function, W l is the weight parameter of the long segment matching layer;

[0045] Calculate the similarity and calculate the segment matching results R respectively s And the long segment matching result R l Similarity with the input feature X; use the cosine similarity function as the similarity measurement function to obtain the similarity value S s and S l ;

[0046] For the similarity value S s and S l Normalized calculation is performed to obtain the weight parameter w of the segment matching results and the long segment matching results s and w l .

[0047] As a preferred technical solution of the present invention, the training of the sample evaluation model includes the following steps:

[0048] A1. Initialize the basic model: Construct a basic sample evaluation model consisting of an input layer, a pairing layer, a classification unit, and a weighted output layer. The hidden layer parameters of the classification unit are replaced with the training results of the sample subset. The parameters of the remaining layers in the basic model are initialized with standard parameters.

[0049] A2. Independent training of sample areas: Divide and preprocess historical recovery samples to generate sample subsets. For each sample subset, train the hidden layer parameters of the class area units separately. Specifically,

[0050] Based on the physiological data and basic information of the sample subset, the parameters of the segment matching layer and the long segment matching layer are trained in the category area unit to generate multiple sets of parameters;

[0051] The trained parameter set is stored according to the category area mapping rules and dynamically loaded into the category area unit as the hidden layer parameters during the evaluation process;

[0052] A3. Overall model training: Based on the hidden layer parameters obtained from training the sample subset using the category unit replacement, the remaining parameters of the sample evaluation model are trained, including:

[0053] Training of the pairing layer optimizes the category area selection logic by training the patient's basic information and the category area mapping function;

[0054] Training of the weighted output layer, adjusting the weight coefficients and bias parameters by minimizing the objective function;

[0055] A4. Joint optimization of model parameters: By inputting standardized training set samples, the entire model is jointly optimized. The hidden layer parameters in the class unit are fixed, and only the parameters of the pairing layer, result fitting layer, and weighted output layer are optimized to ultimately obtain the global optimal parameters of the sample evaluation model.

[0056] A5. Verification and Evaluation: Evaluate the performance of the sample evaluation model using the validation set to verify the model's matching effect in different category areas, and fine-tune the remaining parameters based on the verification results.

[0057] As a preferred technical solution of the present invention, the recommendation of the adjustment strategy includes:

[0058] Based on the sample assessment results and recovery prediction results, combined with the patient's basic information and real-time collected physiological data, the patient's current key health characteristics are extracted;

[0059] Calculate the target recovery value R based on the patient's recovery stage and assessment criteria t , the calculation formula is: R t =f g (S p ), where S pis the evaluation standard parameter of the current recovery stage, f g () is the target recovery value calculation function;

[0060] Compare the patient's current recovery value with the target recovery value and calculate the recovery difference D r , determining the patient's recovery status as normal, delayed, or advanced according to the interval of the value of the recovery difference;

[0061] Based on the patient's recovery status and current recovery stage, combined with the preset recovery adjustment strategy library, a personalized adjustment strategy is generated through the rule engine;

[0062] Conduct simulation verification of the recommended adjustment strategy and dynamically update the adjustment strategy based on subsequent changes in the patient's physiological data;

[0063] Adjustment strategies are presented in a structured manner and delivered to medical staff or patients through smart devices, providing visual reports including adjustment suggestions, recovery status assessments, and predicted trends.

[0064] As a preferred technical solution of the present invention, the data statistics and management include:

[0065] The patient's recovery samples, sample evaluation results, recovery prediction results and adjustment strategies are stored in layers, specifically including the basic data layer, stage data layer and evaluation result layer; the basic data layer stores the patient's basic information and original physiological data; the stage data layer stores the standardized recovery samples and corresponding recovery stage data; the evaluation result layer stores the output results of the sample evaluation model and the recommended adjustment strategy;

[0066] Based on the patient's real-time recovery data and historical data, the sample area division logic and category area construction are dynamically updated through an incremental learning algorithm. For newly collected patient data, its similarity with existing category area samples is calculated in real time, and the category area is dynamically adjusted based on the similarity results. For category areas with uneven data distribution in the sample area, the category division is optimized through merging or splitting strategies to ensure that the number of samples in each category area meets the set threshold.

[0067] Provides multi-condition combination query function, searching by basic information, recovery stage, and assessment results; supports visual display of query results, including distribution charts and trend charts of key statistical indicators and corresponding analysis reports;

[0068] Conduct multi-dimensional statistical analysis based on recovery data, including time series statistics, stage recovery statistics, and analysis of recovery differences among patient groups;

[0069] The scope of data access is limited through a hierarchical permission control mechanism, so that different users can only access data corresponding to their permission level;

[0070] Carry out long-term tracking and management of patient recovery data, and analyze the long-term effects of recovery stages and adjustment strategies through data mining algorithms.

[0071] An intelligent assessment system for anesthesia recovery status includes the following modules:

[0072] Data acquisition module: used to collect patients' physiological data in real time through smart devices, and perform denoising, standardization and time alignment on the collected data;

[0073] State assessment module: used to extract the patient's physiological characteristics based on the real-time collected physiological data, calculate the patient's current anesthesia recovery state according to the preset assessment criteria, and generate a real-time recovery value;

[0074] Recovery sample construction module: used to match the recovery phase according to the single state evaluation results and construct the recovery sample;

[0075] Sample evaluation model module: used to input the restored samples into the sample evaluation model, select the category area, perform calculations through segment matching and long segment matching, and integrate the matching results to generate sample evaluation results and restoration prediction results;

[0076] Adjustment strategy generation module: used to generate personalized recovery adjustment strategies based on sample evaluation results and recovery prediction results;

[0077] Data statistics and management module: used to store, statistically analyze and manage patients' recovery samples, assessment results, prediction results and adjustment strategies;

[0078] Result output module: used to present the recovery assessment results, prediction results and adjustment strategies in a visual way, and provide them to medical staff or patients in an interactive form.

[0079] The present invention has the following advantages:

[0080] The present invention uses intelligent devices to collect patients' multi-dimensional physiological data in real time and combines it with a sample evaluation model of segmented matching and long-segment matching to achieve accurate real-time evaluation of the patient's recovery status; through dynamically optimized sample areas, samples are grouped and clustered in combination with the patient's basic information, and the individual characteristics of the patient are dynamically loaded and matched to ensure that the evaluation model can adapt to the individual differences of different patients, significantly improving the adaptability and robustness of the model.

[0081] The present invention uses a preset strategy library and rule engine, combined with the patient's real-time physiological data, to generate personalized recovery adjustment strategies, including physiological indicator optimization suggestions, medication adjustment plans and life management plans; strategy recommendations also support dynamic feedback and optimization, and adjust the patient's recovery path in real time, overcoming the lack of dynamism in traditional experience-based guidance strategies.

[0082] The present invention uses a multi-level data statistics and management module to store patient data in layers according to the basic data layer, stage data layer and evaluation result layer, and combines the incremental learning algorithm to optimize the category area construction, thereby realizing efficient data management and long-term tracking; the use of data encryption and hierarchical authority control technology effectively protects the privacy and security of patient data, meeting the high standards of medical data management.

[0083] Through the time series prediction model and incremental learning algorithm, the present invention can dynamically predict the patient's recovery status, and analyze the patient's recovery trend in combination with long-term tracking data, providing a reliable basis for medical intervention and solving the problem of the lack of predictive ability of traditional evaluation methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only schematic diagrams of the present invention. Those skilled in the art can also derive other drawings based on the provided drawings without inventive effort.

[0085] Figure 1 This is a schematic diagram of the structure of an intelligent evaluation system for anesthesia recovery status adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0086] To make the objectives, technical solutions, and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0087] Example 1, an intelligent assessment method for anesthesia recovery status, comprising the following steps:

[0088] Step S1: Data collection: The patient's physiological data is collected in real time through smart devices, and the collected data is de-noised, standardized, and time-aligned; the physiological data includes but is not limited to heart rate, blood pressure, body temperature, respiratory rate, and blood oxygen saturation;

[0089] Step S2: Single state assessment, based on the real-time collected physiological data and the patient, extracting the patient's physiological characteristics, and evaluating and calculating the patient's current anesthesia recovery state according to the preset assessment criteria based on the patient's basic information to obtain a real-time recovery value;

[0090] The step of calculating the real-time recovery value includes:

[0091] Extract the patient’s physiological data collected in real time, recorded as ,in represents the patient’s i-th physiological data;

[0092] Compare the real-time collected physiological data with the preset healthy physiological data to obtain the difference value of each physiological data item , the calculation formula is: ;in, is the healthy physiological data corresponding to the i-th physiological data item;

[0093] The difference value for each physiological data item A nonlinear function is used for conversion, such as the following formula:

[0094] ;

[0095] The nonlinear function is selected according to the type of physiological data and the size of the difference value. The selection conditions of the four functions in the formula are: For continuous changes, For the regression trend, Less than the change threshold and greater than the change threshold;

[0096] Perform weighted summation on the difference values ​​of all converted physiological data items to obtain the initial real-time recovery value , the calculation formula is: ,in is the weighting coefficient of the i-th physiological data item;

[0097] Initial real-time recovery value Perform regression analysis and correction to obtain the final real-time recovery value. The calculation formula is:

[0098] ,in is the preset regression coefficient used to correct the recovery value.

[0099] Step S3: Constructing a recovery sample: Matching the corresponding recovery stage according to the evaluation results of a single state, and constructing a recovery sample based on the results of all single state evaluations; calculating the recovery comparison value by comparing the physiological data of the current recovery stage with the corresponding standard physiological data; the final recovery sample includes the patient's basic information, each recovery stage and the corresponding multiple sets of physiological data, real-time recovery value and recovery comparison value;

[0100] The construction of the recovery sample includes the following steps:

[0101] Matching of single state assessment results: Based on the patient's real-time recovery value of a single state, the patient's current physiological data is matched with the standard physiological data of the corresponding recovery stage to determine the patient's recovery stage;

[0102] Calculation of the recovery comparison value: extract the physiological data of all single states in the current recovery stage, calculate the overall average physiological data value and the standard average physiological data value of the standard recovery stage; compare the physiological data standardization of the current recovery stage with that of the standard recovery stage, and calculate the recovery comparison value of the current recovery stage , the calculation formula is: ,in represents the average value of the i-th physiological data in the current recovery stage, represents the average value of the i-th physiological data in the standard recovery phase, n represents the total number of physiological data items, and A represents the adjustment coefficient;

[0103] Based on the results of a single status assessment and the recovery comparison value of the recovery stage, a recovery sample of the patient is generated; and the recovery sample is stored in the form of structured data.

[0104] Step S4: Input the restored sample into the sample evaluation model for evaluation calculation to obtain the sample evaluation result and the restoration prediction result; after the restored sample is input into the sample evaluation model, the category area is selected and the segment matching calculation and the long segment matching calculation are performed, and the two types of calculation results are weighted to obtain the final sample evaluation result and the restoration prediction result;

[0105] The structure of the sample evaluation model includes:

[0106] The input layer is used to receive the patient's physiological data and basic information, and convert it into the feature vector form X={x1,x2,…,x n}, where n represents the feature dimension;

[0107] The pairing layer is used to select a category area based on the patient's basic information, classify and map the input feature vector X through the mapping function M() of the patient's basic information, and determine the category area that matches the input patient information; the basic information includes anesthesia type, disease type, age range, gender, and chronic disease history;

[0108] Category area units include:

[0109] The segmented matching layer uses a local matching network based on the attention mechanism to extract representative local features in the feature vector, and performs local feature matching on the key physiological indicators of the patient's recovery data to obtain the local matching result R s ;

[0110] The long segment matching layer, based on the recurrent neural network, performs global modeling on the time series features of the input data and obtains the matching results R of the input features in the time dimension. l ;

[0111] The result fitting layer is used to fit and integrate the matching results of the segment matching layer and the long segment matching layer, through the weighting function: F(R s ,R l )=w s *R s +w l *R l , calculate the intermediate evaluation value R m ; where w s and w l is the weight parameter of the matching result, and satisfies w s +w l =1;

[0112] Weighted output layer, used to evaluate the intermediate value R m Combined with the specific weight W of the patient's current recovery stage p , generate the final sample evaluation result R f and predicted recovery results P r .

[0113] The acquisition of the category area includes:

[0114] By preprocessing historical recovery samples, the collected original sample data is cleaned, denoised, standardized, and time-aligned to generate a preprocessed sample set;

[0115] Based on a combination of one or more attributes in the patient's basic information, the preprocessed sample set is grouped and divided into multiple sample subsets, wherein the number of samples in each of the grouped sample subsets is greater than a preset threshold;

[0116] The density clustering algorithm is used to cluster the sample subsets and the number of samples in each cluster category is calculated;

[0117] Filter the clustering results and take the sample set whose valid category number exceeds the set value as the valid sample set; the valid category number is the number of samples of the corresponding cluster category exceeding the set ratio;

[0118] Model training is performed separately for each valid sample set, and a category area model is constructed based on the support vector machine. The corresponding training process includes:

[0119] Using the physiological data and basic information of the valid sample set as input, the model parameters are optimized to obtain the feature mapping function of each category area model;

[0120] Use the validation set to evaluate the category area model, optimize the model's hyperparameters, and ensure the model's accuracy and robustness;

[0121] Finally, all category area models are combined to form a category area, which is used to select the corresponding category area for calculation in the sample evaluation model according to the basic information of the input patient.

[0122] The acquisition of the weight parameter of the matching result includes:

[0123] Obtain segment matching results and long segment matching results, use the segment matching layer to extract local features of the input feature X, and generate local matching results R s , the calculation formula is: R s =f s (X,W s ), where f s () is the piecewise matching function, W s is the weight parameter of the segment matching layer;

[0124] The long segment matching layer is used to globally model the overall time series features of the input feature X and generate the global matching result R l , the calculation formula is: R l =f l (X,W l ), where f l () is the long segment matching function, W l is the weight parameter of the long segment matching layer;

[0125] Calculate the similarity and calculate the segment matching results R respectively s And the long segment matching result R l Similarity with the input feature X; use the cosine similarity function as the similarity measurement function to obtain the similarity value S s and S l ;

[0126] For the similarity value S s and S l Normalized calculation is performed to obtain the weight parameter w of the segment matching results and the long segment matching results s and w l .

[0127] The training of the sample evaluation model includes the following steps:

[0128] A1. Initialize the basic model: Construct a basic sample evaluation model consisting of an input layer, a pairing layer, a classification unit, and a weighted output layer. The hidden layer parameters of the classification unit are replaced with the training results of the sample subset. The parameters of the remaining layers in the basic model are initialized with standard parameters.

[0129] A2. Independent training of sample areas: Divide and preprocess historical recovery samples to generate sample subsets. For each sample subset, train the hidden layer parameters of the class area units separately. Specifically,

[0130] Based on the physiological data and basic information of the sample subset, the parameters of the segment matching layer and the long segment matching layer are trained in the category area unit to generate multiple sets of parameters;

[0131] The trained parameter set is stored according to the category area mapping rules and dynamically loaded into the category area unit as the hidden layer parameters during the evaluation process;

[0132] A3. Overall model training: Based on the hidden layer parameters obtained from training the sample subset using the category unit replacement, the remaining parameters of the sample evaluation model are trained, including:

[0133] Training of the pairing layer optimizes the category area selection logic by training the patient's basic information and the category area mapping function;

[0134] Training of the weighted output layer, adjusting the weight coefficients and bias parameters by minimizing the objective function;

[0135] A4. Joint optimization of model parameters: By inputting standardized training set samples, the entire model is jointly optimized. The hidden layer parameters in the class unit are fixed, and only the parameters of the pairing layer, result fitting layer, and weighted output layer are optimized to ultimately obtain the global optimal parameters of the sample evaluation model.

[0136] A5. Verification and Evaluation: Evaluate the performance of the sample evaluation model using the validation set to verify the model's matching effect in different category areas, and fine-tune the remaining parameters based on the verification results.

[0137] Step S5: Recommendation of adjustment strategy: generating a personalized recovery adjustment strategy based on the sample evaluation results and recovery prediction results;

[0138] Recommendations for adjustment strategies include:

[0139] Based on the sample assessment results and recovery prediction results, combined with the patient's basic information and real-time collected physiological data, the patient's current key health characteristics are extracted;

[0140] Calculate the target recovery value R based on the patient's recovery stage and assessment criteria t , the calculation formula is: R t =f g (S p ), where S p is the evaluation standard parameter of the current recovery stage, f g () is the target recovery value calculation function;

[0141] Compare the patient's current recovery value with the target recovery value and calculate the recovery difference D r , determining the patient's recovery status as normal, delayed, or advanced according to the interval of the value of the recovery difference;

[0142] Based on the patient's recovery status and current stage, combined with a pre-set recovery adjustment strategy library, a personalized adjustment strategy is generated through a rule engine. Adjustments include, but are not limited to: physiological indicator adjustment suggestions, such as target ranges for respiratory rate, heart rate, or blood pressure; medication use suggestions, including dosage adjustments or medication timing; and lifestyle management suggestions, including personalized plans for diet, exercise, and rest.

[0143] The recommended adjustment strategy is simulated and verified, and dynamically updated based on subsequent changes in the patient's physiological data. This includes: real-time correction of recovery differences and strategy recommendations based on subsequently collected physiological data; and recalculation of the adjustment strategy if the patient's recovery value does not approach the target value for multiple consecutive periods.

[0144] Adjustment strategies are presented in a structured manner and delivered to medical staff or patients through smart devices, providing visual reports including adjustment suggestions, recovery status assessments, and predicted trends.

[0145] Step S6: Data statistics and management, statistics, storage and management of all patients' recovery samples, sample evaluation results, recovery prediction results and recovery adjustment strategies, support data backup, query and long-term tracking analysis, while ensuring data security and privacy protection.

[0146] The data statistics and management include:

[0147] B1. Store the patient's recovery samples, sample assessment results, recovery prediction results, and adjustment strategies in layers, specifically including a basic data layer, a stage data layer, and an assessment result layer;

[0148] The basic data layer stores the patient's basic information and original physiological data; the stage data layer stores the standardized recovery samples and corresponding recovery stage data; the evaluation result layer stores the output results of the sample evaluation model and the recommended adjustment strategies;

[0149] All data is stored and transmitted through encryption algorithms, and distributed database technology is used to achieve high reliability and disaster recovery backup functions;

[0150] B2. Dynamic optimization construction of sample area:

[0151] Based on the patient's real-time recovery data and historical data, the sample area division logic and category area construction are dynamically updated through incremental learning algorithms;

[0152] For newly collected patient data, the similarity between it and existing category area samples is calculated in real time, and the category area is dynamically adjusted based on the similarity results;

[0153] For category areas with uneven data distribution in the sample area, optimize the category division through merging or splitting strategies to ensure that the number of samples in each category area reaches the set threshold;

[0154] B3.Data Query and Retrieval:

[0155] Provides multi-condition combination query function, supporting retrieval by patient attributes (such as gender, age, disease type), recovery stage, assessment results, etc.

[0156] Supports visual display of query results, including distribution charts and trend charts of key statistical indicators and corresponding analysis reports;

[0157] B4. Statistics and Analysis:

[0158] Conduct multi-dimensional statistical analysis based on recovery data, including time series statistics, stage recovery statistics, and analysis of recovery differences among patient groups;

[0159] Provide dynamic analysis reports on data distribution within the sample area, and count the number of samples in each category area, the mean value and fluctuation range of key physiological indicators;

[0160] B5. Data Privacy and Rights Management:

[0161] A hierarchical permission control mechanism limits data access, ensuring that different users can only access data corresponding to their permission level. A log is kept for every access, modification, and download of data to ensure traceability of data operations. Anonymization and differential privacy technologies are used to protect patient data privacy and prevent the leakage of sensitive information.

[0162] B6. Long-term tracking and forecast optimization:

[0163] Conduct long-term tracking and management of patient recovery data, and analyze the long-term effects of recovery stages and adjustment strategies through data mining algorithms;

[0164] Use time series prediction models to predict patient recovery trends and dynamically adjust prediction model parameters to adapt to new data;

[0165] Provides time span analysis of patient data in the sample area to ensure that the construction of category areas is consistent with the long-term data distribution.

[0166] Example 2, an intelligent assessment system for anesthesia recovery status, see Figure 1 As shown, it includes the following modules:

[0167] Data acquisition module: used to collect patients' physiological data in real time through smart devices, and perform denoising, standardization and time alignment on the collected data;

[0168] State assessment module: used to extract the patient's physiological characteristics based on the real-time collected physiological data, calculate the patient's current anesthesia recovery state according to the preset assessment criteria, and generate a real-time recovery value;

[0169] Recovery sample construction module: used to match the recovery phase according to the single state evaluation results and construct the recovery sample;

[0170] Sample evaluation model module: used to input the restored samples into the sample evaluation model, select the category area, perform calculations through segment matching and long segment matching, and integrate the matching results to generate sample evaluation results and restoration prediction results;

[0171] Adjustment strategy generation module: used to generate personalized recovery adjustment strategies based on sample evaluation results and recovery prediction results;

[0172] Data statistics and management module: used to store, statistically analyze and manage patients' recovery samples, assessment results, prediction results and adjustment strategies;

[0173] Result output module: used to present the recovery assessment results, prediction results and adjustment strategies in a visual way, and provide them to medical staff or patients in an interactive form.

[0174] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent assessment method for anesthesia recovery status, characterized in that: The following steps are involved: Step S1: Data collection: The patient's physiological data is collected in real time through smart devices, and the collected data is de-noised, standardized, and time-aligned; the physiological data includes but is not limited to heart rate, blood pressure, body temperature, respiratory rate, and blood oxygen saturation; Step S2: Single state assessment, based on the real-time collected physiological data and the patient, extracting the patient's physiological characteristics, and evaluating and calculating the patient's current anesthesia recovery state according to the preset assessment criteria based on the patient's basic information to obtain a real-time recovery value; Step S3: Construction of recovery samples, matching the corresponding recovery stages according to the evaluation results of the single state, and constructing the recovery samples based on the results of all single state evaluations; The recovery comparison value is calculated by comparing the physiological data of the current recovery stage with the corresponding standard physiological data. The final recovery sample includes the patient's basic information, each recovery stage and the corresponding multiple sets of physiological data, real-time recovery value and recovery comparison value. Step S4: Input the restored sample into the sample evaluation model for evaluation calculation to obtain the sample evaluation result and the restoration prediction result; After the restored sample is input into the sample evaluation model, the category area is selected and segment matching calculation and long segment matching calculation are performed, and the two types of calculation results are weighted to obtain the final sample evaluation result and restoration prediction result; The structure of the sample evaluation model includes: The input layer is used to receive the patient's physiological data and basic information, and convert it into the feature vector form X={x1,x2,…,x n }, where n represents the feature dimension; The pairing layer is used to select a category area based on the patient's basic information, classify and map the input feature vector X through the mapping function M() of the patient's basic information, and determine the category area that matches the input patient information; the basic information includes anesthesia type, disease type, age range, gender, and chronic disease history; Category area units include: The segmented matching layer uses a local matching network based on the attention mechanism to extract representative local features in the feature vector, and performs local feature matching on the key physiological indicators of the patient's recovery data to obtain the local matching result R s ; The long segment matching layer, based on the recurrent neural network, performs global modeling on the time series features of the input data and obtains the matching results R of the input features in the time dimension. l ; The result fitting layer is used to fit and integrate the matching results of the segment matching layer and the long segment matching layer, through the weighting function: F(R s ,R l )=w s *R s +w l *R l , calculate the intermediate evaluation value R m ; where w s and w l is the weight parameter of the matching result, and satisfies w s +w l =1; Weighted output layer, used to evaluate the intermediate value R m Combined with the specific weight W of the patient's current recovery stage p , generate the final sample evaluation result R f and predicted recovery results P r ; The acquisition of the weight parameter of the matching result includes: Obtain segment matching results and long segment matching results, use the segment matching layer to extract local features of the input feature X, and generate local matching results R s , the calculation formula is: R s =f s (X,W s ), where f s () is the piecewise matching function, W s is the weight parameter of the segment matching layer; The long segment matching layer is used to globally model the overall time series features of the input feature X and generate the global matching result R l , the calculation formula is: R l =f l (X,W l ), where f l () is the long segment matching function, W l is the weight parameter of the long segment matching layer; Calculate the similarity and calculate the segment matching results R respectively s And the long segment matching result R l Similarity with the input feature X; use the cosine similarity function as the similarity measurement function to obtain the similarity value S s and S l ; For the similarity value S s and S l Normalized calculation is performed to obtain the weight parameter w of the segment matching results and the long segment matching results s and w l ; Step S5: Recommendation of adjustment strategy: generating a personalized recovery adjustment strategy based on the sample evaluation results and recovery prediction results; Step S6: Data statistics and management, statistics, storage and management of all patients' recovery samples, sample evaluation results, recovery prediction results and recovery adjustment strategies, support data backup, query and long-term tracking analysis, while ensuring data security and privacy protection.

2. The method for intelligently assessing anesthesia recovery status according to claim 1, wherein: The step of calculating the real-time recovery value includes: Extract the patient’s physiological data collected in real time, recorded as ,in represents the patient’s i-th physiological data; Compare the real-time collected physiological data with the preset healthy physiological data to obtain the difference value of each physiological data item , the calculation formula is: ;in, is the healthy physiological data corresponding to the i-th physiological data item; The difference value for each physiological data item A nonlinear function is used for conversion, such as the following formula: ； The nonlinear function is selected according to the type of physiological data and the size of the difference value. The selection conditions of the four functions in the formula are: For continuous changes, For the regression trend, Less than the change threshold and greater than the change threshold; Perform weighted summation on the difference values ​​of all converted physiological data items to obtain the initial real-time recovery value , the calculation formula is: ,in is the weighting coefficient of the i-th physiological data item; Initial real-time recovery value Perform regression analysis and correction to obtain the final real-time recovery value. The calculation formula is: ,in is the preset regression coefficient used to correct the recovery value.

3. The method for intelligently assessing anesthesia recovery status according to claim 1, wherein: The construction of the recovery sample includes the following steps: Matching of single state assessment results: Based on the patient's real-time recovery value of a single state, the patient's current physiological data is matched with the standard physiological data of the corresponding recovery stage to determine the patient's recovery stage; Calculation of the recovery comparison value: extract the physiological data of all single states in the current recovery stage, calculate the overall average physiological data value and the standard average physiological data value of the standard recovery stage; compare the physiological data standardization of the current recovery stage with that of the standard recovery stage, and calculate the recovery comparison value of the current recovery stage , the calculation formula is: ,in represents the average value of the i-th physiological data in the current recovery stage, represents the average value of the i-th physiological data in the standard recovery phase, n represents the total number of physiological data items, and A represents the adjustment coefficient; Based on the results of a single status assessment and the recovery comparison value of the recovery stage, a recovery sample of the patient is generated; and the recovery sample is stored in the form of structured data.

4. The method for intelligently assessing anesthesia recovery status according to claim 1, wherein: The acquisition of the category area includes: By preprocessing historical recovery samples, the collected original sample data is cleaned, denoised, standardized, and time-aligned to generate a preprocessed sample set; Based on a combination of one or more attributes in the patient's basic information, the preprocessed sample set is grouped and divided into multiple sample subsets, wherein the number of samples in each of the grouped sample subsets is greater than a preset threshold; The density clustering algorithm is used to cluster the sample subsets and the number of samples in each cluster category is calculated; Filter the clustering results and take the sample set whose valid category number exceeds the set value as the valid sample set; the valid category number is the number of samples of the corresponding cluster category exceeding the set ratio; Model training is performed separately for each valid sample set, and a category area model is constructed based on the support vector machine. The corresponding training process includes: Using the physiological data and basic information of the valid sample set as input, the model parameters are optimized to obtain the feature mapping function of each category area model; Use the validation set to evaluate the category area model, optimize the model's hyperparameters, and ensure the model's accuracy and robustness; Finally, all category area models are combined to form a category area, which is used to select the corresponding category area for calculation in the sample evaluation model according to the basic information of the input patient.

5. The method for intelligently assessing anesthesia recovery status according to claim 1, wherein: The training of the sample evaluation model includes the following steps: A1. Initialize the basic model: Construct a basic sample evaluation model consisting of an input layer, a pairing layer, a classification unit, and a weighted output layer. The hidden layer parameters of the classification unit are replaced with the training results of the sample subset. The parameters of the remaining layers in the basic model are initialized with standard parameters. A2. Independent training of sample areas: Divide and preprocess historical recovery samples to generate sample subsets. For each sample subset, train the hidden layer parameters of the class area units separately. Specifically, Based on the physiological data and basic information of the sample subset, the parameters of the segment matching layer and the long segment matching layer are trained in the category area unit to generate multiple sets of parameters; The trained parameter set is stored according to the category area mapping rules and dynamically loaded into the category area unit as the hidden layer parameters during the evaluation process; A3. Overall model training: Based on the hidden layer parameters obtained from training the sample subset using the category unit replacement, the remaining parameters of the sample evaluation model are trained, including: Training of the pairing layer optimizes the category area selection logic by training the patient's basic information and the category area mapping function; Training of the weighted output layer, adjusting the weight coefficients and bias parameters by minimizing the objective function; A4. Joint optimization of model parameters: By inputting standardized training set samples, the entire model is jointly optimized. The hidden layer parameters in the class unit are fixed, and only the parameters of the pairing layer, result fitting layer, and weighted output layer are optimized to ultimately obtain the global optimal parameters of the sample evaluation model. A5. Verification and Evaluation: Evaluate the performance of the sample evaluation model using the validation set to verify the model's matching effect in different category areas, and fine-tune the remaining parameters based on the verification results.

6. The method for intelligently assessing anesthesia recovery status according to claim 1, wherein: Recommendations for adjustment strategies include: Based on the sample assessment results and recovery prediction results, combined with the patient's basic information and real-time collected physiological data, the patient's current key health characteristics are extracted; Calculate the target recovery value R based on the patient's recovery stage and assessment criteria t , the calculation formula is: R t =f g (S p ), where S p is the evaluation standard parameter of the current recovery stage, f g () is the target recovery value calculation function; Compare the patient's current recovery value with the target recovery value and calculate the recovery difference D r , determining the patient's recovery status as normal, delayed, or advanced according to the interval of the value of the recovery difference; Based on the patient's recovery status and current recovery stage, combined with the preset recovery adjustment strategy library, a personalized adjustment strategy is generated through the rule engine; Conduct simulation verification of the recommended adjustment strategy and dynamically update the adjustment strategy based on subsequent changes in the patient's physiological data; Adjustment strategies are presented in a structured manner and delivered to medical staff or patients through smart devices, providing visual reports including adjustment suggestions, recovery status assessments, and predicted trends.

7. The method for intelligently assessing anesthesia recovery status according to claim 1, wherein: The data statistics and management include: The patient's recovery samples, sample evaluation results, recovery prediction results and adjustment strategies are stored in layers, specifically including the basic data layer, stage data layer and evaluation result layer; the basic data layer stores the patient's basic information and original physiological data; the stage data layer stores the standardized recovery samples and corresponding recovery stage data; the evaluation result layer stores the output results of the sample evaluation model and the recommended adjustment strategy; Based on the patient's real-time recovery data and historical data, the sample area division logic and category area construction are dynamically updated through an incremental learning algorithm. For newly collected patient data, its similarity with existing category area samples is calculated in real time, and the category area is dynamically adjusted based on the similarity results. For category areas with uneven data distribution in the sample area, the category division is optimized through merging or splitting strategies to ensure that the number of samples in each category area meets the set threshold. Provides multi-condition combination query function, searching by basic information, recovery stage, and assessment results; supports visual display of query results, including distribution charts and trend charts of key statistical indicators and corresponding analysis reports; Conduct multi-dimensional statistical analysis based on recovery data, including time series statistics, stage recovery statistics, and analysis of recovery differences among patient groups; The scope of data access is limited through a hierarchical permission control mechanism, so that different users can only access data corresponding to their permission level; Carry out long-term tracking and management of patient recovery data, and analyze the long-term effects of recovery stages and adjustment strategies through data mining algorithms.

8. An intelligent assessment system for anesthesia recovery status, characterized in that: The system applies an intelligent assessment method for anesthesia recovery status according to any one of claims 1 to 7, and includes the following modules: Data acquisition module: used to collect patients' physiological data in real time through smart devices, and perform denoising, standardization and time alignment on the collected data; State assessment module: used to extract the patient's physiological characteristics based on the real-time collected physiological data, calculate the patient's current anesthesia recovery state according to the preset assessment criteria, and generate a real-time recovery value; Recovery sample construction module: used to match the recovery phase according to the single state evaluation results and construct the recovery sample; Sample evaluation model module: used to input the restored samples into the sample evaluation model, select the category area, perform calculations through segment matching and long segment matching, and integrate the matching results to generate sample evaluation results and restoration prediction results; Adjustment strategy generation module: used to generate personalized recovery adjustment strategies based on sample evaluation results and recovery prediction results; Data statistics and management module: used to store, statistically analyze and manage patients' recovery samples, assessment results, prediction results and adjustment strategies; Result output module: used to present the recovery assessment results, prediction results and adjustment strategies in a visual way, and provide them to medical staff or patients in an interactive form.

Images