Cerebrovascular postoperative neurological function recovery monitoring system based on big data

Abstract

The invention discloses a cerebrovascular postoperative neural function recovery monitoring system based on big data. The cerebrovascular postoperative neural function recovery monitoring system comprises a data management module, a neural function analysis module, a recovery process analysis module and a postoperative neural function recovery monitoring module. The invention belongs to the technical field of neurological function monitoring, and particularly relates to a cerebrovascular postoperative neurological function recovery monitoring system based on big data, which performs staged neurological function recovery monitoring by combining staged neurological function analysis and recovery process analysis, and predicts the basic condition of current neurological recovery through neurological function analysis. The trending recovery effect is analyzed through the recovery process, the data utilization depth in the cerebral vascular postoperative neural function recovery monitoring process is improved, and the overall accuracy of the system is also improved; a time sequence regression integrated model of the recovery state of the patient is improved through combined modeling, and the modeling effect of the recovery state of the patient is improved through combined modeling of a dynamic Bayesian network and a hidden Markov model.

Description

Technical Field

[0001] The present invention belongs to the technical field of neurological function monitoring, and specifically relates to a monitoring system for postoperative neurological function recovery of cerebral blood vessels based on big data. Background Art

[0002] A monitoring system for postoperative neurological function recovery of cerebral blood vessels based on big data is an intelligent system that comprehensively utilizes multi-dimensional information such as postoperative neurological function data, vital signs, and imaging data of patients, and realizes real-time monitoring and evaluation of the neurological function recovery process of patients undergoing cerebral blood vessel surgery through deep learning and big data analysis techniques. This system can accurately track the recovery progress of patients, identify potential complication risks, predict the recovery time, and provide data support for personalized treatment plans, ultimately helping doctors optimize treatment decisions, improve the rehabilitation effect of patients, and reduce medical costs.

[0003] However, in the existing monitoring systems for postoperative neurological function recovery of cerebral blood vessels, there are technical problems such as the existing neurological function recovery monitoring mainly using single-stage real-time prediction to achieve recovery monitoring through real-time big data analysis. In fact, the neurological function recovery, especially the postoperative recovery, is a recovery process with obvious stage characteristics. The practice of single-stage real-time prediction causes a large amount of data redundancy and sacrifices prediction accuracy for implementation efficiency. In the existing postoperative neurological function analysis methods, there are technical problems such as postoperative neurological function analysis involving complex indicators, and the analysis for a single task cannot meet the objective needs. In the existing postoperative recovery process analysis methods, there are technical problems such as the insufficient representation ability of the existing methods for the patient's recovery state, which further leads to the machine's inability to fully understand the current stage and state of the patient's recovery. Even with sufficient data, it is impossible to effectively analyze the process information of postoperative recovery. Summary of the Invention

[0004] In the existing monitoring system for postoperative neurological function recovery of cerebrovascular diseases, the existing monitoring of neurological function recovery mainly adopts single-stage real-time prediction. Through real-time big data analysis, the recovery monitoring is realized. However, in fact, the recovery of neurological function, especially postoperative recovery, is a recovery process with obvious stage characteristics. The practice of single-stage real-time prediction causes a large amount of data redundancy and sacrifices the prediction accuracy for implementation efficiency. This solution creatively combines phased neurological function analysis and recovery process analysis to conduct phased monitoring of neurological function recovery. By analyzing neurological function, the basic situation of current nerve recovery is predicted, and the trend of recovery effect is analyzed through the recovery process analysis, which improves the depth of data utilization in the process of monitoring postoperative neurological function recovery of cerebrovascular diseases and also improves the overall accuracy of the system. In the existing methods for postoperative neurological function analysis, there are complex indicators involved in postoperative neurological function analysis, and the analysis of single tasks cannot meet the objective needs. This solution creatively adopts a multi-task learning method combining deep convolutional long short-term network. By dividing subtasks for neurological function analysis and conducting multi-task learning, the analysis dimension and result depth of postoperative neurological function analysis are improved. In the existing methods for postoperative recovery process analysis, the existing methods have insufficient ability to represent the patient's recovery state, resulting in the machine's inability to fully understand the current stage and state of the patient's recovery. Even with sufficient data, the process information of postoperative recovery cannot be effectively analyzed. This solution creatively adopts a joint modeling to improve the time series regression integration model of the patient's recovery state for recovery process analysis. Through the joint modeling of dynamic Bayesian network and hidden Markov model, the modeling effect of the patient's recovery state is improved, and through the time series regression integration model, the recovery process analysis under complex time series data is realized, which improves the overall usability of the system.

[0005] The technical solution adopted by the present invention is as follows: The monitoring system for postoperative neurological function recovery of cerebrovascular diseases based on big data provided by the present invention includes a data management module, a neurological function analysis module, a recovery process analysis module, and a postoperative neurological function recovery monitoring module;

[0006] The data management module is used for data collection, storage, and management. Through data management, an optimized dataset for neurological function recovery monitoring is obtained, and the optimized dataset for neurological function recovery monitoring is sent to the neurological function analysis module and the recovery process analysis module;

[0007] The neurological function analysis module is used for analyzing the phased recovery of neurological function after cerebrovascular surgery. Through neurological function analysis, evaluation data on neurological function recovery is obtained, and the evaluation data on neurological function recovery is sent to the recovery process analysis module and the postoperative neurological function recovery monitoring module;

[0008] The recovery process analysis module is used to analyze the patient's nerve function recovery process. Through the recovery process analysis, recovery process analysis data is obtained, and the recovery process analysis data is sent to the postoperative nerve function recovery monitoring module;

[0009] The postoperative nerve function recovery monitoring module is used to conduct phased monitoring of the nerve function recovery after cerebrovascular surgery. Through postoperative nerve function recovery monitoring, postoperative nerve function recovery health analysis reference data is obtained.

[0010] Furthermore, the data management specifically includes data acquisition and processing and data storage management. Through data acquisition and processing, the original data set required for postoperative nerve function recovery monitoring is collected, and optimized and enhanced processing is performed to obtain an optimized data set for nerve function recovery monitoring. Through data storage management, the optimized data set for nerve function recovery monitoring is uniformly stored and managed for data security and traceability maintenance;

[0011] The data acquisition and processing specifically perform multi-dimensional data acquisition and data optimization processing;

[0012] The multi-dimensional data acquisition specifically includes the following steps: physiological data acquisition, imaging data acquisition, clinical evaluation data acquisition, and behavioral state data acquisition;

[0013] The physiological data acquisition specifically acquires heart rate data, blood pressure data, blood oxygen saturation data, body temperature data, respiratory rate data, and sleep quality data;

[0014] The imaging data acquisition specifically acquires brain CT data, ultrasound examination data, and electroencephalogram data;

[0015] The clinical evaluation data acquisition specifically acquires nerve consciousness evaluation data, activities of daily living ability data, motor recovery situation data, and doctor's nerve function evaluation data;

[0016] The behavioral state data acquisition specifically acquires gait ability evaluation data, hand movement evaluation data, cognitive function data, and mental health degree evaluation data;

[0017] The data optimization processing specifically includes the following steps: data cleaning, data standardization, data integration, feature engineering, and data optimization and enhancement;

[0018] The data storage management specifically stores and manages the optimized data set for nerve function recovery monitoring through data integration and encrypted storage.

[0019] Furthermore, the neural function analysis specifically uses a multi-task learning method integrating a deep convolutional long short-term network based on the optimized dataset for monitoring neural function recovery to perform neural function analysis and obtain evaluation data on neural function recovery, which specifically includes the following steps: constructing a deep convolutional long short-term integrated network, multi-task learning modeling, model loss optimization, neural function analysis model training, and neural function analysis;

[0020] The construction of the deep convolutional long short-term integrated network specifically involves successively constructing a standard deep neural network, a convolutional neural network, and a long short-term memory neural network, and constructing the deep convolutional long short-term integrated network through subnet integration;

[0021] The multi-task learning modeling specifically involves setting up a shared layer and initializing a multi-task learning model to perform multi-task learning modeling, sharing learning parameters through the shared layer, extracting multi-task features and performing feature fusion to obtain multi-task feature data;

[0022] The initialization of the multi-task learning model specifically refers to defining four sub-tasks, specifically including the task of the neural function recovery stage, the task of predicting the recovery speed, the task of predicting the complication risk, and the task of predicting the cognitive function score;

[0023] The model loss optimization specifically involves constructing an integrated loss function to optimize model training, and the calculation formula is:

[0024] ;

[0025] In the formula, L total is the integrated loss function, is the weight of neural function recovery, and L stage is the loss function of neural function recovery, specifically using the cross-entropy loss function, is the weight of predicting the recovery speed, and L speed is the loss function of predicting the recovery speed, specifically using the mean squared error loss function, is the weight of predicting the complication risk, and L complication is the loss function of predicting the complication risk, specifically using the mean squared error loss function, is the weight of the cognitive function score, and L cognitive is the loss function of the cognitive function score, specifically using the mean squared error loss function;

[0026] The training of the neural function analysis model specifically involves training the neural function analysis model through the construction of the deep convolutional long short-term integrated network, the multi-task learning modeling, and the model loss optimization to obtain the neural function analysis model;

[0027] The above-mentioned neurological function analysis specifically involves performing neurological function analysis based on the optimized dataset for monitoring neurological function recovery, using the neurological function analysis model, and obtaining the evaluation data on the recovery of neurological function.

[0028] The evaluation data on the recovery of neurological function specifically includes the quantified data of the neurological function recovery score, the predicted data of the recovery speed, the predicted data of the complication risk, and the cognitive function score data.

[0029] Furthermore, the above-mentioned analysis of the recovery process specifically involves performing an analysis of the recovery process based on the optimized dataset for monitoring neurological function recovery and the evaluation data on the recovery of neurological function, using a time series regression ensemble model that jointly models and improves the patient's recovery status, and obtaining the analysis data on the recovery process. The specific steps are as follows: Joint modeling and improvement of the patient's recovery status, construction of a time series regression ensemble model, construction of a time series regression classification output, training of the recovery process analysis model, and analysis of the recovery process.

[0030] The joint modeling and improvement of the patient's recovery status specifically involves constructing a joint modeling model of a dynamic Bayesian function and a hidden Markov model, and optimizing the parameters of the joint modeling model by maximizing the likelihood function of the observed data, thereby obtaining the joint modeling of the patient's recovery status and the characteristics of the patient's recovery status.

[0031] The above-mentioned dynamic Bayesian network specifically involves setting the representation of the recovery status node and calculating the state transition probability to obtain the output of the state transition prediction probability. The calculation formula is:

[0032] ;

[0033] In the formula, is the output of the state transition prediction probability, which is used to represent the conditional probability that the patient recovers from the previous state data x t-1 to the current state data x t . Among them, x t is the state data, specifically the input data of the dynamic Bayesian network, which is used to represent the integration of the optimized dataset for monitoring neurological function recovery and the evaluation data on the recovery of neurological function. I is the total number of recovery status nodes, i is the index of the recovery status node, and the specific value range includes the acute stage, the recovery stage, and the stable stage. t is the time index. is the previous state data x of the patient t-1 is the output of the state transition prediction probability when the recovery status node is i, is the probability that the patient's recovery status is the recovery status node i;

[0034] The Hidden Markov Model specifically predicts the current probability output of the recovery state by constructing an observation matrix and calculating the likelihood function of the observed data, and optimizes the parameters of the joint modeling model by maximizing the likelihood function of the observed data to obtain the joint modeling of the patient's recovery state and the characteristics of the patient's recovery state. The calculation formula for predicting the current probability output of the recovery state is as follows:

[0035] ;

[0036] In the formula, is the state transition probability observation matrix representation function, which is used to represent the current probability output of the recovery state. is the element representation of the state transition probability observation matrix, where x t is the state data, j is the second index of the recovery state node, and i is the index of the recovery state node. is the likelihood function representation of the observed data, which is used to represent that at time t, the current state data x t specifically takes the observed value O when the second index j of the recovery state node is taken. t The corresponding observed probability data, where O t is the observed value, and b j (·) is the observation matrix representation function;

[0037] The construction of the time series regression ensemble model specifically constructs a standard long short-term memory model and a gated recurrent unit to extract time series relationship features, and performs time series regression prediction based on the patient's recovery state features to obtain time series ensemble feature data;

[0038] The construction of the time series regression classification output specifically constructs a random forest model, receives the time series ensemble feature data for regression prediction, and obtains the time series prediction output;

[0039] The training of the recovery process analysis model specifically trains the recovery process analysis model through the improvement of the patient's recovery state joint modeling, the construction of the time series regression ensemble model, and the construction of the time series regression classification output to obtain the recovery process analysis model;

[0040] The recovery process analysis specifically uses the recovery process analysis model based on the optimized dataset for monitoring nerve function recovery and the evaluation data of nerve function recovery to perform recovery process analysis and obtain recovery process analysis data;

[0041] The recovery process analysis data specifically includes recovery speed prediction data, recovery time prediction data, recovery trend score data, and recovery process stage prediction data.

[0042] Furthermore, the postoperative nerve function recovery monitoring specifically involves combining the evaluation data on the nerve function recovery situation to predict and comprehensively evaluate the nerve function recovery situation at different stages, and comprehensively analyzing the recovery process situation by combining the recovery process analysis data to obtain the reference data for the analysis of postoperative nerve function recovery health;

[0043] The reference data for the analysis of postoperative nerve function recovery health specifically includes the overall recovery health score data, the recovery priority score data, and the data on the prediction of the rehabilitation effect.

[0044] The beneficial effects achieved by the present invention using the above solution are as follows:

[0045] (1) Aiming at the technical problem that in the existing postoperative nerve function recovery monitoring system for cerebrovascular diseases, the existing nerve function recovery monitoring mainly uses single-stage real-time prediction to achieve recovery monitoring through real-time big data analysis. However, in fact, the nerve function recovery, especially the postoperative recovery, is a recovery process with obvious stage characteristics. The practice of single-stage real-time prediction causes a large amount of data redundancy and sacrifices the prediction accuracy for the implementation efficiency. This solution creatively combines the staged nerve function analysis and the recovery process analysis to conduct staged nerve function recovery monitoring. It predicts the basic condition of the current nerve recovery through nerve function analysis and the trend recovery effect through recovery process analysis, improving the depth of data utilization in the postoperative nerve function recovery monitoring process for cerebrovascular diseases and also improving the overall accuracy of the system;

[0046] (2) Aiming at the technical problem that in the existing postoperative nerve function analysis methods, the postoperative nerve function analysis involves complex indicators, and the analysis for a single task cannot meet the objective needs. This solution creatively adopts a multi-task learning method combining deep convolutional long short-term networks. By dividing sub-tasks for nerve function analysis and conducting multi-task learning, it improves the analysis dimension and result depth of postoperative nerve function analysis;

[0047] (3) Aiming at the technical problem that in the existing postoperative recovery process analysis methods, the existing methods have insufficient ability to represent the patient's recovery state, resulting in the machine's inability to fully understand the current stage and state of the patient's recovery, and even if sufficient data is provided, it cannot effectively analyze the process information of postoperative recovery. This solution creatively adopts a joint modeling to improve the time series regression integration model of the patient's recovery state for recovery process analysis. Through the joint modeling of a dynamic Bayesian network and a hidden Markov model, it improves the modeling effect of the patient's recovery state, and through the time series regression integration model, it realizes the recovery process analysis under a large amount of complex time series data, improving the overall usability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1Schematic structural diagram of a monitoring system for postoperative neurological function recovery based on big data;

[0049] Figure 2 Schematic flowchart of the steps performed by a monitoring system for postoperative neurological function recovery based on big data;

[0050] Figure 3 Schematic flowchart of the steps performed by the neurological function analysis module;

[0051] Figure 4 Schematic flowchart of the steps performed by the recovery process analysis module.

[0052] The accompanying drawings are used to provide a further understanding of the present invention and form a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. Detailed implementation manners

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0054] Embodiment 1. Refer to Figure 1 A monitoring system for postoperative neurological function recovery based on big data provided by the present invention includes a data management module, a neurological function analysis module, a recovery process analysis module, and a postoperative neurological function recovery monitoring module;

[0055] The data management module is used for data collection, storage, and management. Through data management, an optimized dataset for postoperative neurological function recovery monitoring is obtained, and the optimized dataset for postoperative neurological function recovery monitoring is sent to the neurological function analysis module and the recovery process analysis module;

[0056] The neurological function analysis module is used for analyzing the phased recovery of neurological function after cerebrovascular surgery. Through neurological function analysis, evaluation data on the recovery of neurological function is obtained, and the evaluation data on the recovery of neurological function is sent to the recovery process analysis module and the postoperative neurological function recovery monitoring module;

[0057] The recovery process analysis module is used for analyzing the recovery process of the patient's neurological function. Through recovery process analysis, analysis data on the recovery process is obtained, and the analysis data on the recovery process is sent to the postoperative neurological function recovery monitoring module;

[0058] The postoperative nerve function recovery monitoring module is used to conduct phased monitoring of the postoperative nerve function recovery of cerebrovascular diseases. Through the postoperative nerve function recovery monitoring, reference data for the analysis of the health of postoperative nerve function recovery can be obtained.

[0059] By performing the above operations, in the existing postoperative nerve function recovery monitoring system for cerebrovascular diseases, the existing nerve function recovery monitoring mainly adopts single-stage real-time prediction, and the recovery monitoring is realized through real-time big data analysis. However, in fact, the nerve function recovery, especially the postoperative recovery, is a recovery process with obvious phased characteristics. The practice of single-stage real-time prediction has caused a large amount of data redundancy, and at the same time, the prediction accuracy has been sacrificed for the implementation efficiency. The present solution creatively combines phased nerve function analysis and recovery process analysis to conduct phased nerve function recovery monitoring. The basic condition of the current nerve recovery is predicted through nerve function analysis, and the trend of the recovery effect is analyzed through the recovery process analysis, which improves the depth of data utilization in the process of postoperative nerve function recovery monitoring of cerebrovascular diseases and also improves the overall accuracy of the system.

[0060] Embodiment 2, this embodiment is based on the above embodiment, referring to Figure 1 and Figure 2 , the data management specifically includes data acquisition and processing and data storage management. Through data acquisition and processing, the original data set required for the postoperative nerve function recovery monitoring of cerebrovascular diseases is collected, and optimized and enhanced processing is carried out to obtain an optimized data set for nerve function recovery monitoring. Through data storage management, the optimized data set for nerve function recovery monitoring is uniformly stored and managed for data security and traceability maintenance;

[0061] The data acquisition and processing is specifically to perform multi-dimensional data acquisition and data optimization processing;

[0062] The multi-dimensional data acquisition specifically includes the following steps: physiological data acquisition, imaging data acquisition, clinical evaluation data acquisition, and behavioral state data acquisition;

[0063] The physiological data acquisition specifically collects heart rate data, blood pressure data, blood oxygen saturation data, body temperature data, respiratory rate data, and sleep quality data;

[0064] The imaging data acquisition specifically collects brain CT data, ultrasound examination data, and electroencephalogram data;

[0065] The clinical evaluation data acquisition specifically collects nerve consciousness evaluation data, activities of daily living ability data, motor recovery situation data, and doctor's nerve function evaluation data;

[0066] The collection of the behavioral state data specifically collects gait ability assessment data, hand movement assessment data, cognitive function data, and mental health degree assessment data;

[0067] The data optimization and processing specifically include the following steps: data cleaning, data standardization, data integration, feature engineering, and data optimization and enhancement;

[0068] The data cleaning specifically includes noise removal, missing value processing, and duplicate value processing;

[0069] The data standardization specifically refers to Z-score standardization processing;

[0070] The data integration specifically refers to fusing and sorting the physiological data, imaging data, clinical assessment data, and behavioral state data in terms of data and time series;

[0071] The feature engineering specifically extracts a feature data set from the integrated data set through signal processing, principal component analysis, and artificial feature calculation extraction methods;

[0072] The data optimization and enhancement specifically refer to performing data optimization and enhancement through the data cleaning, the data standardization, the data integration, and the feature engineering to obtain an optimized data set for monitoring nerve function recovery;

[0073] The optimized data set for monitoring nerve function recovery specifically includes integrated nerve function recovery monitoring data and nerve function recovery monitoring feature data;

[0074] The data storage and management specifically store and manage the optimized data set for monitoring nerve function recovery through data integration and encrypted storage.

[0075] Example 3, this example is based on the above example, referring to Figure 1 、 Figure 2 and Figure 3 The nerve function analysis specifically uses a multi-task learning method integrating deep convolutional long short-term networks to perform nerve function analysis based on the optimized data set for monitoring nerve function recovery to obtain nerve function recovery situation assessment data, which specifically includes the following steps: constructing a deep convolutional long short-term integrated network, multi-task learning modeling, model loss optimization, nerve function analysis model training, and nerve function analysis;

[0076] The construction of the deep convolutional long short-term integrated network specifically constructs a standard deep neural network, a convolutional neural network, and a long short-term memory neural network in sequence, and constructs the deep convolutional long short-term integrated network through subnet integration;

[0077] The multi-task learning modeling specifically involves initializing a model by setting up a shared layer and multi-task learning, performing multi-task learning modeling, sharing learning parameters through the shared layer, extracting multi-task features and performing feature fusion to obtain multi-task feature data;

[0078] The initialization of the multi-task learning model specifically refers to defining four sub-tasks, including the prediction task of the neural function recovery stage, the prediction task of the recovery speed, the prediction task of the complication risk, and the prediction of the cognitive function score;

[0079] The prediction task of the neural function recovery stage is used to predict the specific recovery stage that the patient is in. The recovery stage specifically includes the acute phase, the recovery phase, and the stable phase;

[0080] The prediction task of the recovery speed is used to predict the time required for the patient to fully recover;

[0081] The prediction task of the complication risk is used to predict the occurrence probability of postoperative complications. The postoperative complications specifically include infection complications and thrombus complications;

[0082] The prediction of the cognitive function score is used to evaluate the recovery of the patient's cognitive function;

[0083] The optimization of the model loss specifically involves constructing an integrated loss function and performing model training optimization. The calculation formula is:

[0084] ;

[0085] In the formula, L total is the integrated loss function, is the weight of the neural function recovery, L stage is the neural function recovery loss function, specifically using the cross-entropy loss function, is the weight of the recovery speed prediction loss, L speed is the recovery speed prediction loss function, specifically using the mean squared error loss function, is the weight of the complication risk prediction, L complication is the complication risk prediction loss function, specifically using the mean squared error loss function, is the weight of the cognitive function score, L cognitive is the cognitive function score loss function, specifically using the mean squared error loss function;

[0086] The training of the neural function analysis model specifically involves training the neural function analysis model through the construction of the deep convolutional long short-term integration network, the multi-task learning modeling, and the model loss optimization to obtain the neural function analysis model;

[0087] The nerve function analysis specifically involves using the nerve function analysis model based on the optimized dataset for monitoring nerve function recovery to perform nerve function analysis and obtain the evaluation data on the nerve function recovery status.

[0088] The evaluation data on the nerve function recovery status specifically includes the quantified data of the nerve function recovery score, the predicted data of the recovery speed, the predicted data of the complication risk, and the cognitive function score data.

[0089] By performing the above operations, in view of the technical problem in the existing postoperative nerve function analysis methods that the postoperative nerve function analysis involves complex indicators and the analysis for a single task cannot meet the objective needs, this solution creatively adopts a multi-task learning method combining deep convolutional long short-term networks. By dividing sub-tasks for nerve function analysis and performing multi-task learning, the analysis dimension and the depth of the results of postoperative nerve function analysis are improved.

[0090] Example 4, this example is based on the above example, referring to Figure 2 、 Figure 3 and Figure 4 The analysis of the recovery process specifically involves using a time series regression ensemble model that jointly models and improves the patient's recovery status based on the optimized dataset for monitoring nerve function recovery and the evaluation data on the nerve function recovery status to perform the analysis of the recovery process and obtain the analysis data of the recovery process. The specific steps are as follows: joint modeling and improvement of the patient's recovery status, construction of a time series regression ensemble model, construction of a time series regression classification output, training of the recovery process analysis model, and analysis of the recovery process.

[0091] The joint modeling and improvement of the patient's recovery status specifically involves constructing a joint modeling model of a dynamic Bayesian function and a hidden Markov model, and optimizing the parameters of the joint modeling model by maximizing the likelihood function of the observed data to obtain the joint modeling of the patient's recovery status and the characteristics of the patient's recovery status.

[0092] The dynamic Bayesian network specifically involves setting the representation of the recovery status node and calculating the state transition probability to obtain the output of the state transition prediction probability. The calculation formula is:

[0093] ;

[0094] In the formula, is the output of the state transition prediction probability, which is used to represent the conditional probability that the patient recovers from the previous state data x t-1 to the current state data x t . Among them, x tis the status data, specifically the input data of the dynamic Bayesian network, which is used to represent the integration of the optimized dataset for monitoring nerve function recovery and the evaluation data of nerve function recovery. I is the total number of recovery status nodes, and i is the index of the recovery status node. The specific value range includes the acute stage, the recovery stage, and the stable stage. t is the time index. is the previous status data x of the patient t-1 is the output of the state transition prediction probability when the recovery status is node i is the probability that the patient's recovery status is recovery status node i;

[0095] The hidden Markov model specifically constructs an observation matrix, calculates the likelihood function of the observation data, predicts the current recovery status probability output, and optimizes the parameters of the joint modeling model by maximizing the likelihood function of the observation data to obtain the joint modeling of the patient's recovery status and the characteristics of the patient's recovery status. The calculation formula for predicting the current recovery status probability output is:

[0096] ;

[0097] In the formula, is the state transition probability observation matrix representation function, which is used to represent the current recovery status probability output. is the element representation of the state transition probability observation matrix, where x t is the status data, j is the second index of the recovery status node, and i is the index of the recovery status node. is the likelihood function representation of the observation data, which is used to represent the current status data x at time t. t The specific value is the observation value O when the second index j of the recovery status node is reached. t The corresponding observation probability data, where O t is the observation value, and b j (·) is the observation matrix representation function;

[0098] The construction of the time series regression integration model specifically constructs a standard long short-term memory model and a gated recurrent unit, extracts time series relationship features, and performs time series regression prediction based on the patient's recovery status features to obtain time series integration feature data;

[0099] The calculation formula for performing time series regression prediction is:

[0100] ;

[0101] In the formula, is the output hidden state of the standard long short-term memory model, and LSTM(·) is the standard long short-term memory model representation function, X tis the original input data, which is used to represent the characteristics of the patient's recovery status, and t is the time index. is the output hidden state of the gated recurrent unit, GRU(·) is the representation function of the gated recurrent unit, and h t is the temporal regression ensemble hidden state, a is the temporal ensemble weight, and the specific value is 0.6, and y t is the temporal ensemble prediction output, which is used to represent the temporal ensemble feature data, and W r is the temporal ensemble weight, and b r is the temporal ensemble bias term;

[0102] The construction of the temporal regression classification output is specifically to construct a random forest model, receive the temporal ensemble feature data for regression prediction, and obtain the temporal prediction output;

[0103] The training of the recovery process analysis model is specifically to train the recovery process analysis model through the improvement of the joint modeling of the patient's recovery status, the construction of the temporal regression ensemble model, and the construction of the temporal regression classification output, and obtain the recovery process analysis model;

[0104] The recovery process analysis is specifically to perform the recovery process analysis based on the neurofunctional recovery monitoring optimization dataset and the neurofunctional recovery evaluation data, using the recovery process analysis model, and obtain the recovery process analysis data;

[0105] The recovery process analysis data specifically includes recovery speed prediction data, recovery time prediction data, recovery trend score data, and recovery process stage prediction data.

[0106] By performing the above operations, in the existing postoperative recovery process analysis methods, there are technical problems that the existing methods have insufficient representation ability for the patient's recovery status, resulting in the machine's inability to fully understand the current stage and status of the patient's recovery, and even if sufficient data is provided, it is impossible to effectively analyze the postoperative recovery process information. This solution creatively adopts a temporal regression ensemble model that jointly models and improves the patient's recovery status for recovery process analysis. Through the joint modeling of the dynamic Bayesian network and the hidden Markov model, the modeling effect of the patient's recovery status is improved, and through the temporal regression ensemble model, the recovery process analysis under complex temporal data is realized, and the overall usability of the system is improved.

[0107] Example Five, this example is based on the above example, refer to Figure 1 The postoperative neurofunctional recovery monitoring is specifically to combine the neurofunctional recovery evaluation data to perform phased neurofunctional recovery prediction and comprehensive evaluation, and through combining the recovery process analysis data, comprehensively analyze the recovery process situation to obtain the postoperative neurofunctional recovery health analysis reference data;

[0108] The postoperative nerve function recovery and health analysis reference data specifically includes overall recovery and health score data, recovery priority score data, and rehabilitation effect prediction situation data;

[0109] The overall recovery and health score data is used to represent the comprehensive quantitative score of the patient's overall health status after surgery, and the value range is [0, 100];

[0110] The recovery priority score data is used to represent the judgment score of whether the patient's current recovery status is a recovery priority, and the value range is {0, 1};

[0111] The rehabilitation effect prediction situation data is used to represent the prediction types of future recovery situations, including good, medium, and poor.

[0112] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process and method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such a process and method.

[0113] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

[0114] The above describes the present invention and its implementation manners. Such a description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. All in all, if those of ordinary skill in the art are inspired by it and without departing from the purpose of the present invention, and design similar structural manners and embodiments to this technical solution without creative efforts, they should all fall within the protection scope of the present invention.

Claims

1. A big data-based neurological function recovery monitoring system after cerebrovascular surgery, characterized by: It includes data management module, neurological function analysis module, recovery process analysis module and postoperative neurological function recovery monitoring module; The data management module is used for data collection, storage and management. Through data management, a neural function recovery monitoring optimization data set is obtained, and the neural function recovery monitoring optimization data set is sent to the neural function analysis module and the recovery process analysis module; The neurological function analysis module is used to analyze the staged recovery of neurological function after cerebrovascular surgery, obtain neurological function recovery assessment data through neurological function analysis, and send the neurological function recovery assessment data to the recovery process analysis module and the postoperative neurological function recovery monitoring module; The neural function analysis adopts a multi-task learning method integrating deep convolutional long-term and short-term networks to perform neural function analysis and obtain neural function recovery assessment data, which specifically includes the following steps: constructing a deep convolutional long-term and short-term integrated network, multi-task learning modeling, model loss optimization, neural function analysis model training and neural function analysis; The recovery process analysis module is used to analyze the patient's neurological function recovery process, obtain recovery process analysis data through recovery process analysis, and send the recovery process analysis data to the postoperative neurological function recovery monitoring module; The recovery process analysis adopts joint modeling to improve the time series regression integrated model of the patient's recovery status, performs recovery process analysis, and obtains recovery process analysis data, which specifically includes the following steps: joint modeling improvement of the patient's recovery status, construction of the time series regression integrated model, construction of the time series regression classification output, recovery process analysis model training and recovery process analysis; The postoperative neurological function recovery monitoring module is used to perform phased monitoring of neurological function recovery after cerebrovascular surgery, and obtain postoperative neurological function recovery health analysis reference data through postoperative neurological function recovery monitoring.

2. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 1 is characterized by: The data management specifically includes data acquisition processing and data storage management. Through data acquisition processing, the original data set required for monitoring neurological function recovery after cerebrovascular surgery is collected, and optimization and enhancement processing are performed to obtain an optimized data set for monitoring neurological function recovery. Through data storage management, the optimized data set for monitoring neurological function recovery is uniformly stored and managed for data security and traceability maintenance.

3. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 2 is characterized by: The data collection and processing specifically includes multi-dimensional data collection and data optimization processing; The multi-dimensional data collection specifically includes the following steps: physiological data collection, imaging data collection, clinical assessment data collection and behavioral status data collection; The physiological data collection specifically collects heart rate data, blood pressure data, blood oxygen saturation data, body temperature data, respiratory rate data and sleep quality data; The image data acquisition specifically includes acquiring brain CT data, ultrasound examination data and electroencephalogram data; The clinical assessment data collection specifically collects neurological awareness assessment data, daily living activities ability data, motor recovery data and physician neurological function assessment data; The behavioral status data collection specifically collects gait ability assessment data, hand movement assessment data, cognitive function data and mental health assessment data; The data optimization process specifically includes the following steps: data cleaning, data standardization, data integration, feature engineering and data optimization and enhancement; The data storage management is specifically to store and manage the neurological function recovery monitoring optimization data set through data integration and encrypted storage.

4. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 3 is characterized by: The neural function analysis is specifically based on the neural function recovery monitoring optimization data set, using a multi-task learning method integrating a deep convolutional long-term and short-term network to perform neural function analysis to obtain neural function recovery assessment data, specifically including the following steps: constructing a deep convolutional long-term and short-term integrated network, multi-task learning modeling, model loss optimization, neural function analysis model training and neural function analysis; The constructing of the deep convolutional long-short term integrated network specifically comprises sequentially constructing a standard deep neural network, a convolutional neural network and a long short-term memory neural network, and constructing the deep convolutional long-short term integrated network through subnet integration; The multi-task learning modeling is specifically to perform multi-task learning modeling by setting a shared layer and a multi-task learning initialization model, and to share learning parameters through the shared layer, extract multi-task features and perform feature fusion to obtain multi-task feature data; The multi-task learning initialization model specifically refers to the definition of four sub-items of tasks, including a neurological function recovery stage prediction task, a recovery speed prediction task, a complication risk prediction task, and a cognitive function score prediction task; The neurological function recovery stage prediction task is used to predict the specific recovery stage of the patient, and the recovery stage specifically includes the acute stage, the recovery stage and the stable stage; The recovery speed prediction task is used to predict the time required for the patient to fully recover; The complication risk prediction task is used to predict the probability of occurrence of postoperative complications in patients, and the postoperative complications specifically include infectious complications and thrombotic complications; The cognitive function score prediction is used to evaluate the patient's cognitive function recovery; The model loss optimization specifically involves constructing an integrated loss function to perform model training optimization; The neural function analysis model training is specifically to train the neural function analysis model by constructing a deep convolutional long-term integrated network, the multi-task learning modeling and the model loss optimization to obtain a neural function analysis model; The neurological function analysis is specifically to perform neurological function analysis based on the neurological function recovery monitoring optimization data set and using the neurological function analysis model to obtain neurological function recovery assessment data; The neurological function recovery assessment data specifically include neurological function recovery score quantification data, recovery speed prediction data, complication risk prediction data and cognitive function score data.

5. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 4 is characterized by: The integrated loss function is constructed and the calculation formula is: ； Where, L total is the integrated loss function, is the neurological function recovery weight, L stage is the neural function recovery loss function, specifically the cross entropy loss function, is the recovery speed prediction loss weight, L speed is the recovery speed prediction loss function, specifically using the mean square error loss function, is the complication risk prediction weight, L complication is the complication risk prediction loss function, specifically the mean square error loss function, is the cognitive function score weight, L cognitive It is the cognitive function score loss function, specifically the mean square error loss function is adopted.

6. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 5 is characterized by: The recovery process analysis is specifically to improve the time series regression integrated model of the patient's recovery status by joint modeling based on the neurological function recovery monitoring optimization data set and the neurological function recovery assessment data, to perform recovery process analysis and obtain recovery process analysis data, and specifically includes the following steps: joint modeling improvement of the patient's recovery status, construction of the time series regression integrated model, construction of the time series regression classification output, recovery process analysis model training and recovery process analysis; The improvement of the joint modeling of the patient's recovery status is specifically to construct a joint modeling model of a dynamic Bayesian function and a hidden Markov model, and optimize the parameters of the joint modeling model by maximizing the likelihood function of the observed data, thereby obtaining the joint modeling of the patient's recovery status and obtaining the patient's recovery status characteristics; The dynamic Bayesian network specifically sets the recovery state node representation and performs state transition probability calculation to obtain the state transition prediction probability output. The calculation formula is: ； In the formula, It is the state transition prediction probability output, which is used to represent the patient's transition from the previous state data x t-1 Restore to current state data x t The conditional probability of t is state data, specifically dynamic Bayesian network input data, used to represent the integration of the neural function recovery monitoring optimization data set and the neural function recovery evaluation data, I is the total number of recovery state nodes, i is the recovery state node index, and the specific value range includes acute phase, recovery phase and stable phase, t is the time index, is the patient's previous state data x t-1 is the state transition prediction probability output when restoring state node i, is the probability that the patient’s recovery state is the recovery state node i; The hidden Markov model specifically constructs an observation matrix and calculates the likelihood function of the observation data to predict the current recovery state probability output, and optimizes the parameters of the joint modeling model by maximizing the likelihood function of the observation data to obtain the joint modeling of the patient's recovery state and obtain the patient's recovery state characteristics. The calculation formula for predicting the current recovery state probability output is: ； In the formula, is the state transition probability observation matrix representation function, which is used to represent the current recovery state probability output. is the element representation of the state transition probability observation matrix, where x t is the state data, j is the second index of the restored state node, i is the restored state node index, It is the likelihood function representation of the observed data, which is used to represent the current state data x at time t. t The specific value is the observation value O when the second index j of the state node is restored. t The corresponding observation probability data, where O t is the observed value, b j (·) is the observation matrix representation function; The constructing of the time series regression integrated model specifically involves constructing a standard long short-term memory model and a gated recurrent unit to extract time series relationship features, and performing time series regression prediction based on the patient's recovery status features to obtain time series integrated feature data; The constructing of the time series regression classification output specifically involves constructing a random forest model, receiving the time series integrated feature data for regression prediction, and obtaining a time series prediction output; The recovery process analysis model training is specifically to train the recovery process analysis model through the joint modeling improvement of the patient recovery status, the construction of the time series regression integrated model and the construction of the time series regression classification output, so as to obtain the recovery process analysis model; The recovery process analysis is specifically to perform recovery process analysis based on the neural function recovery monitoring optimization data set and the neural function recovery situation evaluation data using the recovery process analysis model to obtain recovery process analysis data; The recovery process analysis data specifically includes recovery speed prediction data, recovery time prediction data, recovery trend scoring data and recovery process stage prediction data.

7. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 6 is characterized by: The postoperative neurological function recovery monitoring specifically combines the neurological function recovery assessment data to conduct a phased neurological function recovery prediction and comprehensive assessment, and combines the recovery progress analysis data to conduct a comprehensive analysis of the recovery progress to obtain postoperative neurological function recovery health analysis reference data.

8. The big data-based post-cerebrovascular surgery neurological function recovery monitoring system according to claim 7 is characterized by: The postoperative neurological function recovery health analysis reference data specifically includes overall recovery health score data, recovery priority score data and rehabilitation effect prediction data.

Images