CN118609743B

CN118609743B - Medical data management method, system and storage medium based on artificial intelligence

Info

Publication number: CN118609743B
Application number: CN202411077418.2A
Authority: CN
Inventors: 何美玲; 霍卫红; 肖忠闪; 李淑敏; 华金丹
Original assignee: Puyang Medical College; Shenzhen Minghao Biotechnology Co ltd
Current assignee: Puyang Medical College; Shenzhen Minghao Biotechnology Co ltd
Priority date: 2024-08-07
Filing date: 2024-08-07
Publication date: 2024-11-26
Anticipated expiration: 2044-08-07
Also published as: CN118609743A

Abstract

The application relates to the technical field of data processing, and discloses a medical data management method, a medical data management system and a medical data storage medium based on artificial intelligence. The method comprises the following steps: preprocessing the multi-source medical data to obtain a standardized medical data set, and performing intelligent classification and labeling to obtain structured medical knowledge data; constructing a digital twin body to obtain a virtual medical twin body, and performing data mining and feature extraction to obtain medical analysis data; comprehensively analyzing individual data of patients and operation data of hospitals to obtain personalized medical schemes and resource optimization strategies; and carrying out data optimization on the personalized medical scheme and the resource optimization strategy through the patient flow data to obtain the target management strategy. According to the application, heterogeneous data is integrated by preprocessing the multi-source medical data, and intelligent classification and labeling are performed to obtain the structured medical knowledge data, so that the efficiency and accuracy of medical data management based on artificial intelligence are improved on the basis of improving the availability of the data.

Description

Medical data management method, system and storage medium based on artificial intelligence

Technical Field

The application relates to the technical field of data processing, in particular to a medical data management method, system and storage medium based on artificial intelligence.

Background

The existing medical data management system mainly comprises an electronic medical record system, a hospital information system, a laboratory information system, an image archiving and communication system and the like. These systems enable electronic recording of patient information, basic medical data analysis, and simple hospital resource management. In recent years, with the development of artificial intelligence and internet of things technologies, some advanced medical institutions have begun to try to apply these new technologies to the medical field. For example, medical image aided diagnosis is performed using deep learning algorithms, electronic medical records are analyzed using natural language processing techniques, and patient physiological indicators are monitored in real time through wearable devices and smart sensors. In addition, some medical institutions have begun to explore big data analysis techniques, which provide support for clinical decisions through mining of massive medical data.

However, there are a number of disadvantages to the prior art. Most systems are still relatively independent, lack of effective data integration and sharing mechanisms, and result in serious information islanding problems, and difficulty in fully utilizing multi-source heterogeneous medical data. Secondly, the existing system generally lacks dynamic adaptability, and resource allocation and service flow are difficult to quickly adjust according to real-time conditions. Third, in personalized medicine, existing systems often only provide statistical-based generic advice, making it difficult to achieve truly accurate medicine. Fourth, hospital resource optimization management still relies mainly on human experience, lacking intelligent and automated decision support tools. Finally, existing systems also face challenges in terms of data security and privacy protection, especially in the context of data sharing and cross-institution collaboration. These deficiencies severely restrict the improvement of medical service quality and the efficient utilization of medical resources.

Disclosure of Invention

The application provides a medical data management method, a system and a storage medium based on artificial intelligence, which are used for improving the efficiency and the accuracy of medical data management based on artificial intelligence.

In a first aspect, the present application provides an artificial intelligence based medical data management method, comprising: preprocessing multi-source medical data acquired from an electronic medical record system, medical equipment and wearable equipment to obtain a standardized medical data set; performing intelligent classification and labeling on the standardized medical data set to obtain structured medical knowledge data; constructing a digital twin body based on the structured medical knowledge data to obtain a virtual medical twin body; performing data mining and feature extraction on the virtual medical twin body to obtain medical analysis data; based on the medical analysis data, comprehensively analyzing the patient individual data and the hospital operation data acquired in real time to obtain a personalized medical scheme and a resource optimization strategy; performing data optimization on the personalized medical scheme and the resource optimization strategy through pre-acquired patient flow data to obtain a target management strategy, wherein the target management strategy comprises the following steps: resource allocation sub-policies, patient management sub-policies, and service optimization sub-policies.

With reference to the first aspect, in a first implementation manner of the first aspect of the present application, preprocessing multi-source medical data collected from an electronic medical record system, a medical device, and a wearable device to obtain a standardized medical data set includes: -performing data cleansing on the multi-source medical data to obtain initial cleansing data, and performing format unification processing on the initial cleansing data to obtain unified format data; performing data desensitization processing on the unified format data to obtain privacy protection data, and performing time sequence alignment on the privacy protection data to obtain time-consistent data; performing interpolation processing on the missing values in the time-consistent data to obtain a complete data set, and performing outlier detection and processing on the complete data set to obtain outlier correction data; carrying out data normalization processing on the abnormal correction data to obtain normalized data, and carrying out dimension reduction processing on the normalized data through a principal component analysis algorithm to obtain dimension reduction data; carrying out data segmentation and coding on the reduced data to obtain coded data; and performing metadata annotation and index creation on the encoded data to obtain the standardized medical data set.

With reference to the first aspect, in a second implementation manner of the first aspect of the present application, the performing intelligent classification and labeling on the standardized medical data set to obtain structural medical knowledge data includes: performing multidimensional feature extraction on the standardized medical data set to obtain a medical feature vector set, and performing unsupervised cluster analysis on the medical feature vector set to obtain an initial medical data category; semantic mapping is carried out on the initial medical data category based on a preset medical ontology to obtain semantic enhanced medical data, and multi-modal relation extraction is carried out on the semantic enhanced medical data to obtain a medical entity association map; medical term standardization processing is carried out on the medical entity association map to obtain a standardized medical vocabulary network, and a medical event time sequence chain is constructed based on the standardized medical vocabulary network to obtain medical time sequence association data; performing pattern recognition and mining on the medical time sequence associated data to obtain a medical management pattern library; performing simulation evaluation on the medical decision flow based on the medical management mode library to obtain a decision influence index set; carrying out data fusion on the decision-making influence index set and the medical time sequence associated data to obtain a fused medical knowledge graph; and carrying out structural processing and index establishment on the fusion medical knowledge graph to obtain the structural medical knowledge data.

With reference to the first aspect, in a third implementation manner of the first aspect of the present application, the constructing a digital twin based on the structured medical knowledge data to obtain a virtual medical twin includes: carrying out semantic analysis on the structured medical knowledge data to obtain a medical entity relation network, and constructing a three-dimensional space model based on the medical entity relation network to obtain an initial virtual medical environment; performing dynamic parameter mapping on the initial virtual medical environment to obtain an interactable medical scene; performing real-time data stream access on the interactive medical scene to obtain a data-driven virtual medical model; establishing a multi-dimensional evaluation index based on the data driving type virtual medical model to obtain a medical twin evaluation index set, and performing simulation verification on the medical twin evaluation index set to obtain a twin body operation rule set; performing medical procedure digital mapping based on the twin body operation rule set to obtain virtual medical operation data; and performing multi-scene test and optimization on the virtual medical operation data to obtain the virtual medical twin.

With reference to the first aspect, in a fourth implementation manner of the first aspect of the present application, the performing data mining and feature extraction on the virtual medical twin body to obtain medical analysis data includes: performing space-time data decomposition on the virtual medical twin body to obtain a multi-dimensional medical data stream, and performing data cleaning and standardization processing on the multi-dimensional medical data stream to obtain a standardized medical data set; performing time sequence correlation analysis on the normalized medical data set to obtain a medical event correlation chain; constructing a medical decision tree based on the medical event association chain to obtain an initial medical decision rule set; cross-verifying and pruning optimization are carried out on the initial medical decision rule set to obtain a simplified medical decision rule; mining a medical resource utilization mode based on the simplified medical decision rule to obtain a resource allocation feature map; performing cluster analysis on the resource allocation feature map to obtain a medical resource utilization efficiency evaluation index; constructing a multi-objective optimization function based on the medical resource utilization efficiency evaluation index to obtain a resource allocation strategy set; performing sensitivity analysis on the resource allocation strategy set to obtain a key influence factor set; and performing parameter tuning on the virtual medical twin body based on the key influence factor set to obtain the medical analysis data.

With reference to the first aspect, in a fifth implementation manner of the first aspect of the present application, the comprehensively analyzing, based on the medical analysis data, the patient individual data and the hospital operation data collected in real time to obtain a personalized medical solution and a resource optimization policy includes: performing multidimensional decomposition on the medical analysis data to obtain a patient characteristic vector and a hospital resource state matrix; performing cluster analysis on the patient feature vectors to obtain patient grouping data, and constructing personalized risk assessment indexes based on the patient grouping data to obtain a patient risk distribution map; performing dynamic load analysis on the hospital resource state matrix to obtain a resource utilization rate curve, and performing peak prediction based on the resource utilization rate curve to obtain a resource demand prediction table; cross matching is carried out on the patient risk distribution diagram and the resource demand prediction table to obtain an initial resource allocation scheme, and a multi-constraint optimization rule set is constructed based on the initial resource allocation scheme to obtain a resource scheduling strategy; performing discrete event simulation on the resource scheduling strategy to obtain a strategy evaluation result, and performing iterative optimization on the resource scheduling strategy based on the strategy evaluation result to obtain an optimized resource allocation scheme; and carrying out association analysis on the optimized resource allocation scheme and the patient individual data to obtain the personalized medical scheme and the resource optimization strategy.

With reference to the first aspect, in a sixth implementation manner of the first aspect of the present application, the data optimization is performed on the personalized medical solution and the resource optimization policy by using pre-acquired patient flow data to obtain a target management policy, where the target management policy includes: a resource allocation sub-policy, a patient management sub-policy, and a service optimization sub-policy, comprising: performing wavelet transformation on pre-acquired patient flow data to obtain a multi-scale time sequence decomposition result; performing autoregressive analysis on the multi-scale time series decomposition result to obtain a short-term predicted value of the patient flow; calculating the resource utilization rate of each department based on the short-term predicted value of the patient flow to obtain a dynamic resource load distribution diagram, and performing hot spot analysis on the dynamic resource load distribution diagram to obtain a resource bottleneck identification report; generating a resource dynamic allocation instruction set based on the resource bottleneck identification report to obtain a resource allocation sub-strategy; performing queue theoretical analysis on pre-acquired patient flow data to obtain a patient waiting time distribution function, and constructing diagnosis priority rules based on the patient waiting time distribution function to obtain an intelligent diagnosis scheme; combining the intelligent diagnosis scheme with individual characteristic data of the patient to obtain a personalized diagnosis path; constructing a patient guiding mechanism based on the personalized diagnosis path to obtain a patient management sub-strategy; performing collaborative optimization on the resource allocation sub-strategy and the patient management sub-strategy to obtain a service flow reconstruction scheme, and constructing a service quality monitoring index based on the service flow reconstruction scheme to obtain a service optimization sub-strategy; and integrating the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy to obtain the target management strategy.

In a second aspect, the present application provides an artificial intelligence based medical data management system comprising:

The processing module is used for preprocessing multi-source medical data acquired from the electronic medical record system, the medical equipment and the wearable equipment to obtain a standardized medical data set;

the classification module is used for intelligently classifying and marking the standardized medical data set to obtain structured medical knowledge data;

the construction module is used for constructing a digital twin body based on the structured medical knowledge data to obtain a virtual medical twin body;

The extraction module is used for carrying out data mining and feature extraction on the virtual medical twin body to obtain medical analysis data;

The analysis module is used for comprehensively analyzing the patient individual data and the hospital operation data acquired in real time based on the medical analysis data to obtain a personalized medical scheme and a resource optimization strategy;

The optimizing module is used for carrying out data optimization on the personalized medical scheme and the resource optimizing strategy through the pre-acquired patient flow data to obtain a target management strategy, wherein the target management strategy comprises the following steps: resource allocation sub-policies, patient management sub-policies, and service optimization sub-policies.

A third aspect of the present application provides a computer readable storage medium having instructions stored therein which, when run on a computer, cause the computer to perform the artificial intelligence based medical data management method described above.

According to the technical scheme provided by the application, the effective integration of heterogeneous data is realized through the pretreatment and standardization of the multi-source medical data, the problem of information island in the traditional system is solved, and a solid data foundation is laid for the subsequent intelligent analysis. Secondly, by intelligently classifying and labeling the standardized medical data set, the structured medical knowledge data is obtained, which not only improves the understandability and usability of the data, but also provides more comprehensive and accurate information support for medical decision. The virtual medical twin constructed based on the structured medical knowledge data provides a comprehensive and dynamic digital mapping for medical management, so that medical institutions can perform various simulations and predictions in a virtual environment, and decision risks are greatly reduced. The virtual medical twin body is subjected to data mining and feature extraction, and the obtained medical analysis data provides scientific basis for the establishment of personalized medical schemes and the generation of resource optimization strategies. The decision support based on the depth data analysis remarkably improves the accuracy of medical services and the efficiency of resource utilization. By comprehensively analyzing the patient individual data and the hospital operation data acquired in real time, the method can generate a truly personalized medical scheme, meets the requirement of accurate medical treatment, can dynamically adjust resource allocation according to real-time conditions, and improves the operation efficiency of hospitals. In particular, the personalized medical scheme and the resource optimization strategy are further optimized through the pre-collected patient flow data, the obtained target management strategy covers multiple aspects of resource allocation, patient management, service optimization and the like, and a comprehensive intelligent solution is provided for hospital management. The dynamic and intelligent management mode can effectively cope with the complexity and variability of the medical environment, and the medical service quality and the patient satisfaction are obviously improved. In addition, the application of the method can also reduce the medical cost, reduce the waste of medical resources and create greater economic benefits for medical institutions. The method relates to the processing of a large amount of sensitive medical data, fully considers the problems of data safety and privacy protection in the construction and implementation processes, adopts multiple safety measures such as data desensitization, access control and the like, can effectively protect the privacy of patients, and improves the efficiency and the accuracy of medical data management based on artificial intelligence.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of one embodiment of an artificial intelligence-based medical data management method in an embodiment of the present application;

FIG. 2 is a schematic diagram of one embodiment of an artificial intelligence based medical data management system in an embodiment of the application.

Detailed Description

The embodiment of the application provides a medical data management method, a medical data management system and a medical data storage medium based on artificial intelligence. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

For ease of understanding, a specific flow of an embodiment of the present application is described below with reference to fig. 1, where an embodiment of an artificial intelligence-based medical data management method according to an embodiment of the present application includes:

Step S101, preprocessing multi-source medical data acquired from an electronic medical record system, medical equipment and wearable equipment to obtain a standardized medical data set;

It will be appreciated that the implementation subject of the present application may be an artificial intelligence based medical data management system, or may be a terminal or a server, and is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.

In particular, various types of medical data are obtained from different sources. The electronic medical record system provides basic information, diagnosis records, treatment schemes and other text data of the patient; medical devices such as CT, MRI, etc. provide high resolution medical image data; the wearable device provides real-time physiological indexes of the patient, such as continuous data of heart rate, blood pressure, blood sugar and the like. The variety of these data types and formats presents challenges for subsequent processing.

The first step in data preprocessing is data cleansing, which aims to remove noise and outliers. For example, for heart rate data collected by a wearable device, abnormally high or low values due to loosening of the device may occur, which data needs to be identified and processed. Data cleaning also includes processing missing values, which can be resolved by interpolation or mean filling. The unified processing of the data formats follows, converting the data from different sources into a consistent format. For example, the data formats output by medical devices of different vendors are unified into a standard DICOM format. Data desensitization is a key step in protecting patient privacy and involves encryption or replacement of sensitive information such as name, identification card number, etc. Time series alignment is an important step in processing multi-source data, ensuring that data from different devices remains consistent in the time dimension. For example, heart rate data recorded per minute by the wearable device is aligned with body temperature data recorded per day in the electronic medical record on a time axis. The aligned data may still have missing values and require interpolation. Common methods include linear interpolation, spline interpolation, and the like. Outlier detection and handling is an important step to ensure data quality, and outliers can be identified and handled using statistical methods or machine learning algorithms. The data normalization process is to convert data of different scales to the same scale, which is helpful for subsequent data analysis and model training. Principal Component Analysis (PCA) is a commonly used dimension reduction method that can reduce the data dimension while preserving the main information of the data. For high-dimensional medical data, such as gene expression data, PCA can effectively extract key features. The data segmentation and coding is to better organize and manage the data, and can segment according to time, disease type and other factors, and assign unique coding to each data segment.

Finally, metadata tagging and index creation are to facilitate subsequent data retrieval and analysis. Metadata includes information such as the source of the data, the time of acquisition, the type of data, etc., while indexing can speed up the query speed of the data. The original multi-source medical data is converted into a standardized medical data set, which lays a foundation for subsequent intelligent classification, labeling and analysis. This standardized dataset is not only consistent in format, but also aligned in the time dimension, while guaranteeing the quality and privacy security of the data.

For example: a hospital collects data for a week for a patient, including daily body temperature data recorded in electronic medical records, hourly blood pressure data recorded by medical devices, and hourly heart rate data recorded by wearable devices. First, these data are cleaned, removing outliers due to equipment failure, such as obvious error values of 0 or 300, which occur in heart rate data. The time stamps of all data are then uniformed to the hour scale, and the average is taken for finer granularity data (e.g., heart rate). Then, the personal information of the patient is desensitized, and the unique ID is used for replacing the name and the ID card number. After the time series are aligned, the lack of the blood pressure data is found, and the lack values are filled by a linear interpolation method. Several abnormal body temperature data points are detected by a box graph method, and are adjusted to a normal range after being confirmed by a doctor. Finally, the body temperature, blood pressure and heart rate data are normalized to be in the range of 0-1 respectively, and the original multidimensional data (including systolic pressure, diastolic pressure and the like) are reduced in dimension to main dimensions through a PCA algorithm. Finally, the processed data are given a uniform coding format, and corresponding metadata labels are added, so that a standardized medical data set is formed. The standardized data set not only ensures the quality and consistency of the data, but also greatly improves the efficiency and accuracy of subsequent analysis.

Step S102, intelligent classification and labeling are carried out on the standardized medical data set to obtain structured medical knowledge data;

Specifically, a multi-dimensional feature extraction is performed on a standardized medical dataset, which analyzes the features of the data from different angles using various feature extraction algorithms. For example, for time series data, statistical features (such as mean and variance), frequency domain features (such as fourier transform coefficients), time domain features (such as trend and periodicity) and the like can be extracted. For text data, natural language processing techniques may be used to extract features such as keywords, syntactic structures, etc. Unsupervised clustering analysis is performed on the medical feature vector set, and common algorithms include K-means, hierarchical clustering and the like. The purpose of this step is to automatically categorize similar data to form an initial medical data category. For example, patients with similar symptom patterns may be clustered together, or similar diagnostic reports may be grouped into a class. An advantage of unsupervised learning is that potential patterns in the data can be found without the need for predefining categories.

In order to further improve the accuracy and semantic meaning of classification, the initial medical data category is subjected to semantic mapping based on a preset medical ontology. A medical ontology is a formalized knowledge representation that contains concepts, relationships, and rules of the medical field. By mapping the initial category to concepts in the medical ontology, semantically enhanced medical data can be obtained that have not only statistically similar but also well-defined medical semantics. Multimodal extraction of semantically enhanced medical data is the next key step. Multimodal herein refers to different types of medical data, such as text, images, numerical values, etc. The purpose of the relationship extraction is to identify relationships between different entities, such as relationships between symptoms and diseases, relationships between drugs and therapeutic effects, etc. The result of this step is a medical entity association graph, which is a complex network structure that demonstrates various associations between medical concepts.

In order to ensure standardization and consistency of data, medical term standardization processing is required to be performed on medical entity association maps. This step unifies medical terms of different origin and different expressions into a standardized medical vocabulary system, such as the International Classification of Diseases (ICD) or the systematic medical terms (SNOMED CT). The result of the normalization process is a normalized medical vocabulary network that contains not only standardized medical terms but also retains their relationships. Based on the normalized medical vocabulary network, constructing a medical event timing chain is the next important step. This step aims to capture the chronological and causal relationship of medical events. For example, a complete timing chain of patients from initial visits, diagnoses, treatments to rehabilitation, or a timing pattern of the progression of a disease from early symptoms to complications may be constructed. These timing chains form medical timing-related data, providing an important basis for subsequent pattern recognition and predictive analysis.

Pattern recognition and mining of medical time series associated data is an in-depth analysis process aimed at finding rules and patterns hidden in the data. This may include common disease progression patterns, therapeutic response patterns, drug interaction patterns, and the like. Through this step, a medical management pattern library is obtained, which is a knowledge base containing various medical patterns. The simulation evaluation of the medical decision flow based on the medical management mode library is an important link for applying the extracted knowledge to the actual decision. This step can simulate different medical decision scenarios, evaluating potential impact and effects of decisions. Through such simulation evaluation, a set of decision impact indicators is obtained that quantifies the various impacts that different decisions may have.

The decision-making influence index set and the medical time sequence associated data are subjected to data fusion, so that a static decision-making evaluation result is combined with dynamic time sequence data to form a more comprehensive knowledge system. The fusion process obtains a fused medical knowledge graph, which is a complex knowledge network integrating time sequence relations, decision influence and medical knowledge. And carrying out structural processing and index establishment on the fusion medical knowledge graph to obtain final structural medical knowledge data. This step converts the knowledge graph into a structured data format that is easy to store, retrieve and apply, and builds a corresponding index to improve the data access efficiency.

For example: a hospital collected 5-year follow-up data for 1000 diabetics, including regular blood glucose test results, medication records, complications, etc. First, the data are subjected to multidimensional feature extraction, such as calculating average blood glucose level, blood glucose fluctuation amplitude, administration frequency, etc., of each patient, to form a feature vector of each patient. These feature vectors were then clustered using the K-means algorithm, setting k=5, classifying the patients into 5 initial categories. Then, based on a preset diabetes medical ontology, semantic mapping is performed on the 5 categories, so that categories with medical significance such as 'good blood sugar control', 'high risk of complications', and the like are obtained. Further, by analyzing the medication records and blood sugar changes of the patient, the medicine-effect relationship is extracted, and a medical entity association map is constructed. In the term normalization process, different expressions (such as "diabetic nephropathy" and "diabetic nephropathy") used by different doctors are unified into standard terms. Based on the standardized data, a disease development time sequence chain of each patient is constructed, wherein the time sequence chain comprises key time points of first diagnosis, complication occurrence, treatment adjustment and the like. By pattern mining these time-series chains, several typical disease progression patterns are found, such as "rapid complication development", "stable control", etc. Based on these patterns, simulation evaluations of different treatment regimens are performed to derive an indication of the expected outcome of each regimen for different types of patients. Finally, the modes, the evaluation results and the original time sequence data are fused to form a comprehensive diabetes knowledge graph, and the comprehensive diabetes knowledge graph is converted into a structured data format, so that subsequent inquiry and analysis are facilitated. This structured knowledge data not only contains the original medical data, but also incorporates pattern, relationship and decision support information extracted from the data.

Step S103, constructing a digital twin body based on the structured medical knowledge data to obtain a virtual medical twin body;

Specifically, semantic analysis is performed on the structured medical knowledge data, and concepts, entities and relations in the data are deeply understood by using natural language processing and knowledge graph technology. Through this analysis, a network of medical entity relationships is obtained that clearly demonstrates the complex associations between the various medical concepts. Based on the medical entity relationship network, constructing a three-dimensional space model is the next key step. The abstract medical knowledge is transformed into a visualized three-dimensional structure including the physical layout of the hospital, the device distribution, the patient flow path, etc. This process involves computer graphics and spatial modeling techniques, resulting in an initial virtual medical environment, which is a static digitized representation of the medical system.

Dynamic parameter mapping of the initial virtual medical environment is critical to "live" the digital twin. The real-time data, the historical data and the predicted data are mapped into the virtual environment so that they can reflect the dynamic changes of the medical system. For example, patient admission, transfer, discharge, etc. information is updated in real time into the virtual environment, or data such as device usage status, healthcare worker workload, etc. is dynamically displayed. This process results in an interactable medical scenario that not only demonstrates the current state, but also simulates the system response under different conditions.

Real-time data stream access to interactive medical scenarios is critical to keep digital twins synchronized with the actual medical system. A data bridge between the actual medical system and the virtual environment is established, and the virtual environment is ensured to reflect the change of the actual situation in real time. This includes accessing data streams for various medical devices, real-time data for patient monitoring systems, updates for hospital information systems, and the like. By the real-time data access, a data-driven virtual medical model is obtained, which can accurately reflect the real-time state and dynamic change of a medical system. The multi-dimensional evaluation index is established based on the data-driven virtual medical model to comprehensively evaluate the performance and efficiency of the medical system. Such metrics include, but are not limited to, patient waiting time, bed turnover rate, medical device utilization, healthcare worker workload, and the like. Through the indexes, a medical twin evaluation index set is obtained, and a quantitative basis is provided for the management and optimization of a medical system.

Simulation verification of the medical twin evaluation index set is an important step of ensuring that the digital twin can accurately reflect and predict the actual medical system behavior. This process constantly adjusts and optimizes the parameters and algorithms of the digital twin by comparing the predicted results in the virtual environment with the performance of the actual medical system. Through repeated verification and adjustment, a twin body operation rule set is obtained, and the rule set defines how the virtual medical twin body operates and responds to various conditions. The medical procedure digital mapping based on the twin body operation rule set is to convert the abstract rule into a specific virtual operation procedure. Various procedures in the medical system, such as a patient visit procedure, a medical resource scheduling procedure, an emergency response procedure, and the like, are simulated. Through this mapping virtual medical operation data is obtained, which data reflects the operation state and performance of the medical system under different conditions. Finally, performing multi-scenario testing and optimization on the virtual medical operation data is a key step for perfecting the virtual medical twin. This process simulates various possible medical scenarios such as daily operation, peak management, emergency handling, etc., and continuously optimizes twin parameters and algorithms based on test results. Through such repeated testing and optimization, a highly accurate and reliable virtual medical twin is finally obtained.

For example: a trimethyl hospital decides to construct a digital twin to optimize the operation of its emergency department. First, semantic analysis is performed on structured medical knowledge data of emergency departments, and key entities such as a patient, a doctor, a nurse, a consulting room, an examination device and the like, and relationships between the key entities such as a patient-waiting-consulting room, a doctor-using-examination device and the like are identified to form a medical entity relationship network. Based on the network, a three-dimensional space model of emergency departments is constructed, including physical layouts of waiting areas, consulting rooms, rescue rooms and the like, and the distribution of medical staff and equipment. Dynamic parameters such as patient latency, physician workload, etc. are then mapped into this model, creating an interactable virtual emergency scenario. The virtual scene can reflect the actual situation in real time by accessing a real-time data stream, such as the arrival rate of the patient updated every 15 minutes, the use frequency of the examination equipment 10 times per hour on average, and the like. Based on this model, an evaluation index set including average waiting time, doctor utilization, bed turnover rate, and the like is established. The twin prediction algorithm is continuously optimized by comparing the predicted results (e.g., predicted peak average latency of 2 hours) with the actual data (actual peak average latency of 1.8 hours) in the virtual environment. Based on the optimized rules, the whole emergency treatment process is digitally mapped, including the whole process from arrival, triage, waiting, treatment to departure of the patient. Finally, the virtual twin is continuously adjusted and optimized by simulating the operation conditions under different conditions (such as common workdays, holiday peaks, sudden public health events, and the like).

Step S104, data mining and feature extraction are carried out on the virtual medical twin body, and medical analysis data are obtained;

The method comprises the steps of carrying out space-time data decomposition on a virtual medical twin body, and splitting complex twin body data according to time and space dimensions to obtain a multi-dimensional medical data stream. The time dimension may include different granularity of hours, days, weeks, months, etc., and the space dimension may include different departments, wards, devices, etc. This decomposition enables subsequent analysis to be performed on different spatio-temporal scales, improving the flexibility and accuracy of the analysis. The multi-dimensional medical data stream is subjected to data cleaning and standardization processing, so that the data quality is improved, and the accuracy of subsequent analysis is ensured. Data cleaning comprises abnormal value removal, missing data processing and the like, and data in different dimensions are unified on the same scale by standardized processing, so that comparison and analysis are facilitated. Through this process, a normalized medical dataset is obtained, which is the basis for subsequent analysis.

Time-series correlation analysis of normalized medical datasets is a key step in mining temporal patterns. Internal links of data in the time dimension are explored using time series analysis techniques such as autocorrelation analysis, cross correlation analysis, and the like. For example, analyzing the relationship between patient flow and medical resource utilization at different points in time, or exploring the effect of time delay between therapeutic intervention and patient recovery rate. From this analysis, a chain of medical event associations is obtained, which shows the mutual influence and causal relationship of medical events over time. Constructing a medical decision tree based on a medical event association chain is an important step in converting the results of the time series analysis into operational decisions. Decision trees are intuitive decision support tools that guide the decision maker in selecting the best action path according to different conditions and situations. In this process, decision rules are extracted from the medical event association chain using a machine learning algorithm, such as an ID3 or C4.5 algorithm, forming a preliminary medical decision rule set. This rule set contains a series of "if-then" decision logic that provides a preliminary guidance framework for medical decisions.

Cross-validation and pruning optimization of a preliminary medical decision rule set are key to improving decision tree accuracy and generalization capability. Cross-validation evaluates the stability and accuracy of the model by dividing the data set into multiple subsets, and repeatedly training and testing the model. And in pruning optimization, unnecessary branches in the decision tree are removed, so that the risk of overfitting is reduced, and the performance of the model on new data is improved. Through the process, simplified medical decision rules are obtained, and the rules are simpler and more efficient and have stronger generalization capability. The mining of medical resource utilization modes based on the simplified medical decision rules is an important link for optimizing resource allocation. Potential patterns of resource utilization are discovered from a large amount of medical data using data mining techniques, such as association rule mining, sequential pattern mining, and the like. For example, a resource waste or bottleneck in certain treatment procedures is found, or an efficient resource allocation scheme is identified. By this mining, a resource allocation feature map is obtained that intuitively demonstrates the various features and modes of medical resource utilization.

The clustering analysis of the resource allocation feature map is a step of further generalizing and summarizing the resource utilization patterns. The clustering analysis uses algorithms such as K-means or hierarchical clustering, and similar resource allocation modes are classified into one type, so that the medical resource utilization efficiency evaluation index is obtained. These metrics may include resource utilization, turnover efficiency, cost-effectiveness ratio, etc., providing a quantified basis for resource optimization. The multi-objective optimization function is constructed based on the medical resource utilization efficiency evaluation index in order to find a balance among the plurality of objectives. This function takes into account a number of objectives such as resource utilization efficiency, patient satisfaction, medical quality, etc., and a set of resource allocation policies is obtained through mathematical modeling and optimization algorithms, such as genetic algorithms or particle swarm optimization algorithms. This policy set contains the optimal resource allocation schemes under different conditions.

The sensitivity analysis of the resource allocation policy set is an important step in assessing policy stability and adaptability. Sensitivity analysis the effect of these changes on the effects of the strategy is observed by changing the values of the different parameters, thus yielding a set of key influencing factors. These factors are variables that affect the resource allocation effect most significantly, and need to be focused and controlled in actual operation. Based on the key influence factor set, parameter tuning is carried out on the virtual medical twin, all the analysis results are applied to the twin, and the behavior of an actual medical system can be reflected and predicted more accurately by adjusting various parameters of the twin. Through this process, medical analysis data is obtained which contains not only the in-depth analysis results of the current medical system, but also predictions and optimization suggestions for future conditions.

For example: the emergency department virtual medical twins of a complex hospital contain detailed operational data of the past year. First, these data are spatially and spatially resolved, and the data are split in the dimensions of hours, days, zhou Dengshi, and in the dimensions of space such as different clinics and examination rooms. After data cleansing and normalization, a normalized data set is obtained that contains 10 key metrics (e.g., patient latency, doctor workload, device usage, etc.). Through the time series correlation analysis, it is found that there is a hysteresis effect of 2 hours between the patient arrival rate and the average waiting time, i.e. the arrival rate at the present moment affects the waiting time after 2 hours. Based on this association, a decision tree is constructed for predicting peak resource demand. Through cross-validation and pruning, a reduced decision rule containing 15 key decision nodes is obtained. Through analysis of these rules, it was found that adding a triage nurse between 9 and 11 monday morning reduced the average waiting time from 40 minutes to 25 minutes. Clustering the resource allocation patterns by using a K-means algorithm to obtain 3 typical resource utilization patterns: high efficiency, balanced and overloaded. Based on these patterns, a multi-objective optimization function is constructed that takes into account resource utilization, patient latency, and medical quality. Solving through genetic algorithm to obtain a strategy set containing 50 optimal resource allocation schemes under different conditions. Sensitivity analysis shows that the number of triage nurses and the length of time that a doctor works are the two most critical factors affecting system performance. Finally, based on the analysis results, parameters of the virtual medical twin are optimized, so that the system performance under different conditions can be accurately predicted, and real-time suggestions are provided for resource allocation.

Step 105, comprehensively analyzing patient individual data and hospital operation data acquired in real time based on medical analysis data to obtain a personalized medical scheme and a resource optimization strategy;

Specifically, the medical analysis data is subjected to multidimensional decomposition, and the complex medical data is split according to different dimensions to obtain a patient characteristic vector and a hospital resource state matrix. The patient feature vector contains information such as various physiological indexes, medical history, treatment response and the like of the patient, and the hospital resource state matrix reflects the real-time use condition and availability of various resources of the hospital. And performing cluster analysis on the characteristic vectors of the patients, and grouping the patients with similar characteristics into a group by using a machine learning algorithm such as K-means or hierarchical clustering to obtain patient grouping data. This grouping helps to identify different types of patient populations, providing a basis for subsequent personalized treatments. Based on patient grouping data, constructing personalized risk assessment indexes, wherein the process involves multi-factor analysis and risk modeling, and finally, a patient risk distribution map is obtained. This profile intuitively demonstrates the health risk levels of different patient populations, providing an important reference for the formulation of personalized medical regimens. And (3) carrying out dynamic load analysis on the hospital resource state matrix, wherein a time sequence analysis technology such as a moving average method, an exponential smoothing method and the like is used in the step, and the dynamic change trend of resource use is analyzed to obtain a resource utilization rate curve. This curve reflects the usage of various resources of the hospital in different time periods. And carrying out peak prediction based on the resource utilization rate curve, and predicting the resource demand in a future period of time by using a time sequence prediction model, such as ARIMA or Prophet, so as to obtain a resource demand prediction table. This predictive table provides prospective guidance for the rational allocation of resources.

The cross matching of the patient risk distribution diagram and the resource demand prediction table is a key step, and the process uses a heuristic algorithm or an optimization algorithm to optimally match the patient demand with the resource supply of the hospital, so as to obtain an initial resource allocation scheme. This approach is a preliminary allocation strategy formulated taking into account patient risk and resource availability. A multi-constraint optimization rule set is constructed based on an initial resource allocation scheme, a plurality of constraint conditions such as medical quality, cost effectiveness, patient satisfaction and the like are considered, and a mathematical programming method such as linear programming or integer programming is used for obtaining a resource scheduling strategy. This strategy not only meets the basic requirements of resource allocation, but also seeks a balance among multiple objectives.

Performing discrete event simulation on the resource scheduling policy is an important step in verifying the effectiveness of the policy. The process uses discrete event simulation software to simulate the execution effect of strategies under different conditions to obtain strategy evaluation results. And (3) carrying out iterative optimization on the resource scheduling strategy based on a strategy evaluation result, wherein a heuristic algorithm, such as a genetic algorithm or a simulated annealing algorithm, is used for continuously adjusting and improving the strategy, and finally, an optimized resource allocation scheme is obtained. The scheme is optimized and verified for multiple rounds, and can achieve a good effect in practical application. And performing association analysis on the optimized resource allocation scheme and the individual data of the patient, wherein the process uses a data mining technology, such as association rule mining or decision tree analysis, and combines the resource allocation scheme with the requirements of the specific patient to obtain a personalized medical scheme and a resource optimization strategy. This final output not only allows for optimal configuration of the overall resources, but also provides personalized treatment advice for each patient's specific situation.

For example: the cardiovascular department of a trimethyl hospital is using this system to optimize its medical services. First, the system performs a multidimensional decomposition on medical analysis data of the past year, and obtains feature vectors (each vector contains 50 features such as age, blood pressure, blood fat and the like) of 10000 patients and a state matrix reflecting the use condition of 30 key resources of cardiovascular department. Patients were divided into 5 major groups, such as hypertension groups, coronary heart disease groups, etc., by the K-means clustering algorithm. Based on these groupings, the system constructs a risk assessment model that yields an average risk score (1-10 cents) for each group of patients. Meanwhile, the resource state matrix is subjected to time sequence analysis, the ARIMA model is used for predicting the resource demand of the future week, and the fact that the use peak of the inspection equipment occurs at 9-11 am on monday is found.

The system matches the patient risk distribution with the resource demand forecast, and initially establishes a resource allocation scheme, such as suggesting that high risk patients (risk score > 8) be preferentially scheduled for treatment during periods of low resource utilization (e.g., 2-4 pm on tuesday). Based on the initial scheme, the system builds a multi-objective optimization model considering medical quality, waiting time and cost, and obtains a better resource scheduling strategy through integer programming solution. This strategy was validated by discrete event simulation, and simulation results showed that the new strategy could reduce the average patient latency from 45 minutes to 30 minutes while reducing the average workload of the physician from 85% to 75%.

After multiple rounds of iterative optimization, the system finally generates an optimized resource allocation scheme. This approach not only optimally allocates the overall resources, but also generates personalized visit advice for each patient. For example, for a 65 year old hypertensive patient, the system recommends a routine examination at 10 am on Wednesday and schedules a 15 minute consultation with specialists at 2 pm on the day, while recommending a personalized medication regimen. The comprehensive scheme not only improves the utilization efficiency of medical resources, but also ensures that each patient can be treated timely and appropriately, and embodies the concept of accurate medical treatment.

Step S106, data optimization is carried out on the personalized medical scheme and the resource optimization strategy through the pre-acquired patient flow data, and a target management strategy is obtained, wherein the target management strategy comprises the following steps: resource allocation sub-policies, patient management sub-policies, and service optimization sub-policies.

Specifically, wavelet transformation is performed on pre-acquired patient flow data, and the time series data is decomposed into components with different frequencies by using a wavelet analysis technology, so that a multi-scale time series decomposition result is obtained. The wavelet transformation can effectively capture various periodic modes and emergencies in the patient flow data, and provides a rich information basis for subsequent analysis. Autoregressive analysis is carried out on the multi-scale time series decomposition result, and a time series prediction model, such as an autoregressive integrated moving average model (ARIMA), is used in the process to analyze the data change trend on different time scales so as to obtain a short-term prediction value of the patient flow. The predicted value reflects the change trend of the patient flow in a future period (such as 24 hours in the future), and provides prospective guidance for resource allocation.

And calculating the resource utilization rate of each department based on the short-term predicted value of the patient flow, and comparing the predicted patient flow with the resource capacity of each department to obtain a dynamic resource load distribution diagram. This profile intuitively shows the resource pressure conditions for different time periods and different departments. And carrying out hot spot analysis on the dynamic resource load distribution map, and identifying areas and time periods with abnormally high or low resource utilization rate by using a space statistical method such as hot spot analysis (Getis-Ord Gi) statistics to obtain a resource bottleneck identification report. This report details potential resource bottleneck points, providing an explicit goal for subsequent resource allocation. Generating a dynamic allocation instruction set of the resources based on the resource bottleneck identification report, wherein a heuristic algorithm or an optimization algorithm, such as a genetic algorithm, is used in the process to prepare a series of resource allocation schemes, so as to obtain a resource allocation sub-strategy. This sub-strategy contains specific resource allocation instructions such as increasing the number of medical staff in a department or adjusting the equipment usage plan at a specific time.

At the same time, queue theory analysis is carried out on the pre-acquired patient flow data, and a queuing theory model, such as an M/M/c model, is used for analyzing the distribution characteristics of the patient waiting time to obtain a patient waiting time distribution function. The function describes probability distribution of patient waiting time under different conditions, and provides theoretical basis for optimizing patient management. And constructing triage priority rules based on the patient waiting time distribution function, wherein the process considers factors such as the emergency degree, waiting time, resource availability and the like of the patient, and an intelligent triage scheme is obtained. The scheme can dynamically adjust the treatment sequence of the patient and balance the use efficiency of medical resources and the waiting experience of the patient. Combining the intelligent triage scheme with the individual characteristic data of the patient, wherein a machine learning algorithm such as a decision tree or a random forest is used in the step, and the triage rule is combined with the specific condition of the patient to obtain a personalized treatment path. This path customizes the optimal flow of medical visits for each patient, including appointment time, examination order, etc. A patient guiding mechanism is built based on the personalized treatment path, and a set of intelligent navigation and reminding system is built in the process to obtain a patient management sub-strategy. This sub-strategy can guide the patient in real time, optimizing his experience of seeking medical advice. And carrying out cooperative optimization on the resource allocation sub-strategy and the patient management sub-strategy, wherein a multi-objective optimization algorithm such as pareto optimization is used in the step, so that balance is sought between resource efficiency and patient experience, and a service flow reconstruction scheme is obtained. The proposal comprehensively considers the resource utilization and the patient demand, and optimizes the medical service flow comprehensively. The service quality monitoring index is constructed based on a service flow reconstruction scheme, and a series of Key Performance Indexes (KPIs) such as patient satisfaction, average waiting time and the like are defined in the process to obtain a service optimization sub-strategy. This sub-strategy provides a complete set of quality of service assessment and improvement mechanisms.

Integrating the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy, wherein the step uses a system integration technology to organically combine the three sub-strategies to form a unified management framework, and the target management strategy is obtained. The target management strategy is a comprehensive, dynamic and intelligent medical management scheme, and can respond to the flow change of the patient in real time, optimize the resource allocation and improve the service quality.

For example: emergency departments in a comprehensive hospital are using this system to optimize their operation. First, the system performs wavelet transform on patient flow data for the past year, decomposing the data into high frequency (hour level), medium frequency (day level) and low frequency (week level) components. By autoregressive analysis, the system predicts patient flow for the next 24 hours, and finds that peak visits will occur 8-10 a.m. on monday, with an expected 30% increase in patient count over usual. Based on the prediction, the system calculates the resource utilization rate of each area of the emergency, and discovers that the resource utilization rate of the sub-diagnosis area and the medical diagnosis room exceeds 90% in the predicted peak period, thereby forming a potential bottleneck.

For this case, the resource orchestration strategy suggests adding 2 triage nurses and 1 physician in monday morning and opening 2 additional clinics. At the same time, through queue theory analysis, the system finds that the average waiting time of patients during peak hours may exceed 60 minutes. Based on this, the patient management sub-strategy constructs a dynamic triage scheme, triages patients according to urgency, and generates personalized triage paths for each patient. For example, for a chest pain patient aged 65, the system recommends that he or she prefer to perform an electrocardiographic examination, and then immediately schedule a physician diagnosis. The service optimization sub-strategy constructs a set of monitoring indexes including average waiting time of patients, diagnosis and treatment time utilization rate of doctors and the like, and sets corresponding target values. Through the collaborative implementation of the strategies, the system can reduce the average waiting time of the peak period from 60 minutes to 30 minutes, and simultaneously reduce the average workload of doctors from 95% to 85%, thereby greatly improving the operation efficiency and the service quality of emergency departments. The comprehensive target management strategy not only optimizes the resource allocation, but also obviously improves the medical experience of patients, and embodies the modern hospital management concept of refinement and individuation.

In the embodiment of the application, the effective integration of heterogeneous data is realized through the pretreatment and standardization of the multi-source medical data, the problem of information island in the traditional system is solved, and a solid data base is laid for the subsequent intelligent analysis. Secondly, by intelligently classifying and labeling the standardized medical data set, the structured medical knowledge data is obtained, which not only improves the understandability and usability of the data, but also provides more comprehensive and accurate information support for medical decision. The virtual medical twin constructed based on the structured medical knowledge data provides a comprehensive and dynamic digital mapping for medical management, so that medical institutions can perform various simulations and predictions in a virtual environment, and decision risks are greatly reduced. The virtual medical twin body is subjected to data mining and feature extraction, and the obtained medical analysis data provides scientific basis for the establishment of personalized medical schemes and the generation of resource optimization strategies. The decision support based on the depth data analysis remarkably improves the accuracy of medical services and the efficiency of resource utilization. By comprehensively analyzing the patient individual data and the hospital operation data acquired in real time, the method can generate a truly personalized medical scheme, meets the requirement of accurate medical treatment, can dynamically adjust resource allocation according to real-time conditions, and improves the operation efficiency of hospitals. In particular, the personalized medical scheme and the resource optimization strategy are further optimized through the pre-collected patient flow data, the obtained target management strategy covers multiple aspects of resource allocation, patient management, service optimization and the like, and a comprehensive intelligent solution is provided for hospital management. The dynamic and intelligent management mode can effectively cope with the complexity and variability of the medical environment, and the medical service quality and the patient satisfaction are obviously improved. In addition, the application of the method can also reduce the medical cost, reduce the waste of medical resources and create greater economic benefits for medical institutions. The method relates to the processing of a large amount of sensitive medical data, fully considers the problems of data safety and privacy protection in the construction and implementation processes, adopts multiple safety measures such as data desensitization, access control and the like, can effectively protect the privacy of patients, and improves the efficiency and the accuracy of medical data management based on artificial intelligence.

In a specific embodiment, the process of executing step S101 may specifically include the following steps:

(1) Performing data cleaning on the multi-source medical data to obtain initial cleaning data, and performing format consistency processing on the initial cleaning data to obtain uniform format data;

(2) Performing data desensitization processing on the unified format data to obtain privacy protection data, and performing time sequence alignment on the privacy protection data to obtain time-consistent data;

(3) Interpolation processing is carried out on the missing values in the time-consistent data to obtain a complete data set, and abnormal value detection and processing are carried out on the complete data set to obtain abnormal correction data;

(4) Carrying out data normalization processing on the abnormal correction data to obtain normalized data, and carrying out dimension reduction processing on the normalized data through a principal component analysis algorithm to obtain dimension reduction data;

(5) Carrying out data segmentation and coding on the reduced data to obtain coded data;

(6) And (5) performing metadata labeling and index creation on the encoded data to obtain a standardized medical data set.

Specifically, the multi-source medical data is subjected to data cleansing, which includes removing duplicate data, correcting obvious error values, unifying data formats, and the like. For example, the date formats of the different sources are unified as "YYYY-MM-DD", or abnormal body temperature records (e.g., 50 ℃) are adjusted to the normal range. Through this process, initial clean-up data is obtained, which has been freed from significant errors and inconsistencies. And carrying out format consistency processing on the initial cleaning data, and converting the data from different sources into a uniform format and unit. For example, all blood pressure data is unified into mmHg units, or electrocardiographic data output from different devices is converted into a standard ECG format. After the processing, unified format data is obtained, and a foundation is laid for subsequent analysis.

Data desensitization of unified format data is a key step in protecting patient privacy. This process involves encrypting or replacing sensitive information such as name, identification card number, etc. For example, the patient ID is encrypted using a hash algorithm, or the real name is replaced with a random code. By the processing, the privacy protection data is obtained, so that the privacy of a patient is protected, and the subsequent data analysis is not influenced. The privacy-preserving data is time-series aligned, which ensures that data from different sources remains consistent in the time dimension. For example, body temperature data recorded once per hour is aligned with blood pressure data recorded once per day on a time axis. The process obtains time consistent data, and provides a basis for subsequent time sequence analysis.

Interpolation of missing values in time-consistent data is an important step in improving data integrity. Common methods include linear interpolation, spline interpolation, or predicting missing values using machine learning algorithms. For example, for missing blood glucose data, linear interpolation may be performed based on values at the previous and subsequent time points. By this processing, a complete data set is obtained, improving the continuity and analyzability of the data. And detecting and processing abnormal values of the complete data set. This step identifies and processes outliers using statistical methods or machine learning algorithms. For example, abnormal blood pressure values are detected using a box-plot method, or abnormal heart rate patterns are identified using a clustering algorithm. Through the process, the abnormal correction data is obtained, and the reliability of the data is improved.

The data normalization processing is performed on the abnormality correction data in order to eliminate dimensional differences between different features. Common methods include min-max normalization or Z-score normalization. For example, all physiological indices are normalized to the range of 0-1. After such processing, normalized data is obtained so that direct comparisons can be made between different features. The dimension reduction processing of the normalized data through the principal component analysis algorithm is an important step for reducing data redundancy and extracting key features. The PCA algorithm converts the original features into a new set of features that are uncorrelated with each other by linear transformation. By selecting the first few principal components, the data dimension can be significantly reduced while retaining a large portion of the information. The process results in reduced data, which reduces computational complexity and retains the main features of the data.

The data segmentation and encoding of the reduced data is done to better organize and manage the data. This step may involve segmenting the continuous time series data into segments of fixed length, or according to specific events (e.g., hospitalization, surgery, etc.). Each data segment is assigned a unique code for subsequent retrieval and analysis. By the processing, the coded data is obtained, and the structuring degree of the data is improved.

And finally, marking metadata and creating indexes for the coded data. Metadata includes information about the source of the data, the time of acquisition, the type of data, etc., while indexes are used to speed up the querying and retrieval of the data. This step adds rich descriptive information for each data segment and builds an efficient index structure. Through this process, a standardized medical dataset is ultimately obtained, which is not only highly structured, but also easy to manage and analyze.

For example: a hospital collected 5 year follow-up data for 1000 diabetics, including daily blood glucose levels, monthly glycosylated hemoglobin (HbA 1 c) measurements, quarterly complications checks, etc. First, 5% of the glycemic data found during the data cleansing process were significantly abnormal (e.g., 0 mg/dL or 1000 mg/dL), and these data were marked and deleted. The format unification process unifies the blood glucose units recorded by different devices to mg/dL. Data desensitization replaces the patient name with a randomly generated ID. Time series alignment aligns the daily blood glucose data, the monthly HbA1c and the quarterly check data on a time axis, creating a uniform time index. For 20% of missing blood glucose data, a linear interpolation method was used for filling. Abnormal value detection found that 3% of HbA1c values were abnormally high (> 15%), and these values were adjusted to the normal range (4-14%). Data normalization scales both blood glucose and HbA1c values to the range of 0-1. PCA dimension reduction reduces the original 20 features (including various complication indexes) to 8 main components, and the 95% information quantity is reserved. Data segmentation the 5 year data for each patient was divided into 20 quarter segments, each segment being assigned a unique code. Finally, metadata is added for each data segment, including data source, acquisition time, patient base information (age, gender, etc.).

In a specific embodiment, the process of executing step S102 may specifically include the following steps:

(1) Performing multidimensional feature extraction on the standardized medical data set to obtain a medical feature vector set, and performing unsupervised cluster analysis on the medical feature vector set to obtain an initial medical data category;

(2) Semantic mapping is carried out on the initial medical data category based on a preset medical ontology to obtain semantic enhanced medical data, and multi-modal relation extraction is carried out on the semantic enhanced medical data to obtain a medical entity association map;

(3) Medical term standardization processing is carried out on the medical entity association map to obtain a standardized medical vocabulary network, and a medical event time sequence chain is constructed based on the standardized medical vocabulary network to obtain medical time sequence association data;

(4) Performing pattern recognition and mining on the medical time sequence associated data to obtain a medical management pattern library;

(5) Performing simulation evaluation on the medical decision flow based on the medical management mode library to obtain a decision influence index set;

(6) Carrying out data fusion on the decision-making influence index set and the medical time sequence associated data to obtain a fused medical knowledge graph;

(7) And carrying out structural processing and index establishment on the fusion medical knowledge graph to obtain structural medical knowledge data.

In particular, multi-dimensional feature extraction of standardized medical datasets involves a variety of feature extraction techniques including statistical feature extraction (e.g., mean, variance, kurtosis, etc.), frequency domain feature extraction (e.g., fourier transform coefficients), and time domain feature extraction (e.g., trend, periodicity, etc.). By these techniques, raw medical data is converted into a series of numerical features, forming a set of medical feature vectors. The vector set contains main characteristic information of the original data, and lays a foundation for subsequent analysis. An unsupervised cluster analysis is performed on the set of medical feature vectors, which uses a clustering algorithm (e.g., K-means or hierarchical clustering) to classify similar feature vectors into one class. The purpose of cluster analysis is to discover the inherent structure in the data, grouping medical data with similar characteristics. Through this process, initial medical data categories are derived that reflect the natural grouping of data, but also lack explicit medical semantics. To assign a clear medical meaning to these data categories, the initial medical data category is semantically mapped based on a preset medical ontology. A medical ontology is a formalized knowledge representation that contains concepts, relationships, and rules of the medical field. Semantic enhanced medical data is obtained by mapping data categories to concepts in a medical ontology. This process not only gives a clear medical meaning to the data category, but also establishes a link between the data and the specialized medical knowledge. The multi-modal relation extraction of the semantically enhanced medical data is a key step of constructing a medical entity association map. This process uses natural language processing and graph analysis techniques to identify and extract relationships between entities from different types of medical data (e.g., text, images, numerical values, etc.). For example, the relationship between symptoms and diseases is extracted from the medical record text, or the relationship between physiological indexes and diagnoses is extracted from the examination results. Through this step, a medical entity association graph is obtained, which graph shows a complex association network between medical concepts. In order to ensure consistency and standardization of medical terms, medical term standardization processing is performed on medical entity association maps. This process unifies the differently expressed medical terms into a standardized medical vocabulary system, such as the international disease classification (ICD) or the systematic medical terms (SNOMED CT). Through this process, a normalized medical vocabulary network is obtained that contains not only standardized medical terms, but also retains their relationships.

Constructing a medical event timing chain based on a normalized medical vocabulary network is an important step in capturing medical event time sequence and causal relationships. This process uses time series analysis techniques, such as time series pattern mining algorithms, to extract the temporal order and interrelationships of events from the medical data. Medical time series associated data are obtained, and the data reflect the association and evolution rule of medical events in the time dimension. Pattern recognition and mining of medical time series associated data is the process of finding rules and patterns hidden in the data. This step discovers recurring patterns or laws from a large volume of medical temporal data using various data mining techniques, such as frequent pattern mining, sequential pattern mining, and the like. Through this process, a medical management pattern library is obtained, which contains various medical management-related patterns, such as common disease development patterns, treatment response patterns, and the like.

The simulation evaluation of the medical decision flow based on the medical management model library is an important link for applying the extracted knowledge to the actual decision. This process uses techniques such as decision tree analysis or monte carlo simulation to evaluate the possible outcomes and effects of different decision paths. Through such simulation evaluation, a decision impact index set is obtained, which quantifies various impacts that may be brought about by different decisions. The data fusion of the decision impact index set and the medical time sequence related data is to combine the static decision evaluation result with the dynamic time sequence data. The fusion process uses data integration techniques, such as feature fusion or decision level fusion, to organically combine information from both data sources. Through the step, a fused medical knowledge graph is obtained, and the graph is a complex knowledge network integrating time sequence relations, decision influence and medical knowledge.

And carrying out structural processing and index establishment on the fusion medical knowledge graph, wherein the step converts the complex knowledge graph into a structural data format which is easy to store, retrieve and apply. By establishing the multidimensional index structure, the data retrieval efficiency is improved. The process finally obtains structured medical knowledge data, and the data set not only contains rich medical knowledge and relations, but also has good structural property and operability.

For example: a hospital collected 5 year follow-up data for 1000 type 2 diabetics, including daily blood glucose levels, monthly HbA1c tests, quarterly complications checks, etc. Firstly, the data are subjected to multidimensional feature extraction, 20 key features such as blood sugar fluctuation amplitude, hbA1c change trend and the like are extracted, and feature vectors of each patient are formed. These feature vectors were clustered using the K-means algorithm, setting k=5, and the patients were classified into 5 initial categories. Then, based on a preset diabetes medical ontology, semantic mapping is carried out on the 5 categories, and medical meaningful categories such as 'good blood sugar control', 'high risk of complications', and the like are obtained. And then, analyzing the medication records and blood sugar changes of the patient, extracting the medicine-effect relationship, and constructing a medical entity association map. In the term normalization process, different expressions (such as "diabetic nephropathy" and "diabetic nephropathy") used by different doctors are unified into standard terms in ICD-10. Based on the standardized data, a disease development time sequence chain of each patient is constructed, wherein the time sequence chain comprises key time points of first diagnosis, complication occurrence, treatment adjustment and the like. By pattern mining these time-series chains, several typical disease progression patterns are found, such as "rapid complication development", "stable control", etc. Based on these patterns, a simulated evaluation is performed for different treatment regimens, such as evaluating the effect of early insulin intervention on the development of complications. The evaluation results show that early insulin intervention may reduce the incidence of complications from 60% to 40% within 5 years for "rapid complication-developing" patients. And fusing the evaluation results with the original time sequence data to form a comprehensive diabetes knowledge graph. Finally, this knowledge graph is converted into a structured data format, creating a multidimensional index based on the stage of disease progression, treatment regimen and risk of complications.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

(1) Carrying out semantic analysis on the structured medical knowledge data to obtain a medical entity relation network, and constructing a three-dimensional space model based on the medical entity relation network to obtain an initial virtual medical environment;

(2) Performing dynamic parameter mapping on the initial virtual medical environment to obtain an interactable medical scene;

(3) Performing real-time data stream access on the interactive medical scene to obtain a data-driven virtual medical model;

(4) Establishing a multi-dimensional evaluation index based on the data driving type virtual medical model to obtain a medical twin evaluation index set, and performing simulation verification on the medical twin evaluation index set to obtain a twin body operation rule set;

(5) Performing medical procedure digital mapping based on the twin body operation rule set to obtain virtual medical operation data;

(6) And performing multi-scene test and optimization on the virtual medical operation data to obtain a virtual medical twin body.

In particular, concepts, entities, and relationships in data are understood in depth using natural language processing and knowledge graph techniques. Semantic coding is carried out on medical terms by applying a Word vector model (such as Word2Vec or BERT), and then a relation extraction algorithm (such as remote supervision learning or attention mechanism) is utilized to identify the relation among entities, so that a complex medical entity relation network is finally formed. This network includes not only medical entities (e.g., diseases, symptoms, medications, etc.), but also various relationships between them (e.g., "cause," "treat," "concurrence," etc.).

The three-dimensional space model is constructed based on the medical entity relation network, and abstract medical knowledge is converted into a visual three-dimensional structure. This process involves graphics and spatial modeling techniques such as using Force-directed algorithms (Force-DIRECTED GRAPH DRAWING) to transform network relationships into spatial locations, which are then visualized by three-dimensional rendering techniques such as OpenGL or WebGL. The constructed three-dimensional spatial model includes the physical layout of the hospital, the equipment distribution, the patient flow path, etc., forming an initial virtual medical environment. Dynamic parameter mapping of the initial virtual medical environment is critical to "live" the digital twin. This step maps real-time data, historical data, and predicted data into the virtual environment, associating the data with the virtual object using data binding techniques. For example, patient admission, transfer, discharge, etc. information is updated in real time into the virtual environment, or data such as device usage status, healthcare worker workload, etc. is dynamically displayed. Through the process, the interactive medical scene is obtained, the current state can be displayed, and the system response under different conditions can be simulated. Real-time data stream access to interactive medical scenarios is critical to keep digital twins synchronized with the actual medical system. This step uses stream processing techniques (e.g., APACHE KAFKA or APACHE FLINK) to establish a data bridge between the actual medical system and the virtual environment, ensuring that the virtual environment can reflect changes in the actual situation in real time. This includes accessing data streams for various medical devices, real-time data for patient monitoring systems, updates for hospital information systems, and the like. By the real-time data access, a data-driven virtual medical model is obtained, which can accurately reflect the real-time state and dynamic change of a medical system.

The multi-dimensional evaluation index is established based on the data-driven virtual medical model to comprehensively evaluate the performance and efficiency of the medical system. This process involves index building and data analysis, using statistical analysis and machine learning techniques to evaluate system performance from multiple perspectives. Evaluation indicators include, but are not limited to, patient waiting time, bed turnover rate, medical device utilization, healthcare worker workload, and the like. Through the indexes, a medical twin evaluation index set is obtained, and a quantitative basis is provided for the management and optimization of a medical system. Simulation verification of the medical twin evaluation index set is an important step of ensuring that the digital twin can accurately reflect and predict the actual medical system behavior. This process simulates various medical scenarios using discrete event simulation techniques (e.g., simPy or AnyLogic), continually adjusting and optimizing parameters and algorithms of the digital twin by comparing the predicted results in the virtual environment to the performance of the actual medical system. Through repeated verification and adjustment, a twin body operation rule set is obtained, and the rule set defines how the virtual medical twin body operates and responds to various conditions.

The medical procedure digital mapping based on the twin body operation rule set is to convert the abstract rule into a specific virtual operation procedure. This process uses workflow modeling techniques (e.g., BPMN or UML activity diagrams) to simulate various procedures in a medical system, such as patient visits, medical resource scheduling, emergency response procedures, and the like. Through this mapping virtual medical operation data is obtained, which data reflects the operation state and performance of the medical system under different conditions. The multi-scenario test and optimization of the virtual medical operation data are key steps for perfecting the virtual medical twin. This process uses scenario analysis and optimization algorithms (e.g., genetic algorithms or reinforcement learning) to simulate various possible medical scenarios such as daily operation, peak management, emergency handling, etc., and continuously optimize twin parameters and algorithms based on test results. Through such repeated testing and optimization, a highly accurate and reliable virtual medical twin is finally obtained.

For example: a trimethyl hospital decides to construct a digital twin to optimize the operation of its emergency department. First, semantic analysis is performed on structured medical knowledge data of emergency departments, and key entities such as a patient, a doctor, a nurse, a consulting room, an examination device and the like, and relationships between the key entities such as a patient-waiting-consulting room, a doctor-using-examination device and the like are identified to form a medical entity relationship network. Based on the network, a three-dimensional space model of emergency departments is constructed, including physical layouts of waiting areas, consulting rooms, rescue rooms and the like, and the distribution of medical staff and equipment. Dynamic parameters such as patient latency, physician workload, etc. are then mapped into this model, creating an interactable virtual emergency scenario. The virtual scene can reflect the actual situation in real time by accessing a real-time data stream, such as the arrival rate of the patient updated every 15 minutes, the use frequency of the examination equipment 10 times per hour on average, and the like. Based on this model, an evaluation index set including average waiting time, doctor utilization, bed turnover rate, and the like is established. Through simulation verification, the average waiting time of the predicted peak period in the virtual environment is found to be 120 minutes, and the actual data is displayed to be 110 minutes, so that the prediction model is finely adjusted. Based on the optimized rules, the whole emergency treatment process is digitally mapped, including the whole process from arrival, triage, waiting, treatment to departure of the patient. Finally, the virtual twin is continuously adjusted and optimized by simulating the operation conditions under different conditions (such as common workdays, holiday peaks, sudden public health events, and the like). For example, during the simulation of holiday peaks, it was found that the average waiting time could be reduced from 90 minutes to 60 minutes by adding a triage nurse. The highly accurate virtual medical twin body not only can reflect the operation conditions of emergency departments in real time, but also can predict the influence of different decisions, and provides a powerful decision support tool for hospital management.

In a specific embodiment, the process of executing step S104 may specifically include the following steps:

(1) Performing space-time data decomposition on the virtual medical twin body to obtain a multi-dimensional medical data stream, and performing data cleaning and standardization processing on the multi-dimensional medical data stream to obtain a standardized medical data set;

(2) Performing time sequence correlation analysis on the normalized medical data set to obtain a medical event correlation chain;

(3) Constructing a medical decision tree based on the medical event association chain to obtain an initial medical decision rule set;

(4) Cross-verifying and pruning optimization are carried out on the initial medical decision rule set to obtain a simplified medical decision rule;

(5) Mining a medical resource utilization mode based on the simplified medical decision rule to obtain a resource allocation feature map;

(6) Performing cluster analysis on the resource allocation feature map to obtain a medical resource utilization efficiency evaluation index;

(7) Constructing a multi-objective optimization function based on the medical resource utilization efficiency evaluation index to obtain a resource allocation strategy set;

(8) Performing sensitivity analysis on the resource allocation strategy set to obtain a key influence factor set;

(9) And performing parameter tuning on the virtual medical twin body based on the key influence factor set to obtain medical analysis data.

In particular, this process splits complex twin volume data in time and space dimensions using multidimensional data analysis techniques, such as Principal Component Analysis (PCA) or tensor decomposition, resulting in a multi-dimensional medical data stream. These data streams include time series (e.g., patient flow per hour), spatial distribution (e.g., resource utilization at different departments), and attribute information (e.g., patient physiological metrics). For time series data, the decomposition can be performed using the following formula:

X(t)=T(t)+S(t)+R(t)；

where X (T) is the original time series, T (T) is the trend component, S (T) is the seasonal component, and R (T) is the residual.

And then, carrying out data cleaning and standardization processing on the multi-dimensional medical data streams, identifying and processing abnormal values by using an abnormality detection algorithm (such as an isolated forest), and simultaneously converting the data with different scales under the same standard by adopting a Z-score standardization method and the like to finally obtain a standardized medical data set. The Z-score normalization formula is as follows:

；

Where z is the normalized value, x is the original value, Is the mean value of the values,Is the standard deviation.

Time-series correlation analysis of normalized medical data sets is a key step in mining potential associations between medical events. This process uses time series analysis techniques such as cross-correlation analysis or the gladhand causal test to explore the time dependence between different medical events. The cross-correlation function can be expressed as:

；

Wherein, Is the hysteresis quantityIs used for the cross-correlation of the (c) and (d),AndThere are two time sequences of which,AndIs their mean value and n is the total length of the time series.

Constructing a medical decision tree based on a medical event association chain is an important step in converting the results of the time series analysis into operational decisions. This process uses a decision tree algorithm, such as ID3 or C4.5, to extract the decision rule from the medical event association chain. The information gain of the decision tree can be calculated by the following formula:

；

Wherein, Is the information gain of feature a,Is the entropy of the data set T,Is the conditional entropy for a given condition of feature a.

Cross-validation and pruning optimization of the initial medical decision rule set is key to improving decision tree accuracy and generalization capability. In pruning, the cost complexity function may be expressed as:

；

Wherein, Is the cost and complexity of the implementation,Is a leaf nodeThe number of samples in (a) is set,Is a leaf nodeIs used as a reference to the entropy of (a),Is the number of leaf nodes of the tree,Is a complexity parameter.

The mining of medical resource utilization modes based on the simplified medical decision rules is an important link for optimizing resource allocation. This step uses an association rule mining algorithm, such as Apriori or FP-Growth, to discover potential patterns of resource utilization from a large volume of medical data. The support and confidence of the association rule can be calculated by the following formulas:

；

wherein A is a first item, B is a second item, Is the total number of transactions that are to be performed,Indicating the frequency of the simultaneous occurrence of items a and B,Representing the frequency of occurrence of item a, the Support measures the frequency of occurrence of the rule in the dataset, and the confidence measure (Confidence) measures the reliability of the rule. The clustering analysis of the resource allocation feature map is a step of further generalizing and summarizing the resource utilization patterns. This process uses clustering algorithms, such as K-means or hierarchical clustering, to classify similar resource allocation patterns into one class.

The multi-objective optimization function is constructed based on the medical resource utilization efficiency evaluation index in order to find a balance among the plurality of objectives. The multi-objective optimization problem can be expressed as:

Wherein, First, theThe utilization rate of seed resources; Total available resources; wait time for jth patient; First of all A medical quality index; Total medical cost; b, budget limitation; ,,, Weights of all targets; decision variables representing allocation to the first The number of seed resources; First of all The maximum available amount of seed resources; maximum allowable waiting time; the minimum medical quality requirement index value.

The sensitivity analysis of the resource allocation policy set is an important step in assessing policy stability and adaptability. The Sobol index in the global sensitivity analysis can be expressed as:

；

Wherein, Is the firstSobol index of the individual parameter(s),Is the output variance caused by the ith parameter,Is the total output variance.

The multi-objective optimization function is constructed based on the medical resource utilization efficiency evaluation index in order to find a balance among the plurality of objectives. This function takes into account a number of objectives such as resource utilization efficiency, patient satisfaction, medical quality, etc., and uses a multi-objective optimization algorithm, such as NSGA-II (non-dominant ranking genetic algorithm II), to derive a set of resource allocation policies. This policy set contains the optimal resource allocation schemes under different conditions.

The sensitivity analysis of the resource allocation policy set is an important step in assessing policy stability and adaptability. This process uses local or global sensitivity analysis methods, such as Morris method or Sobol index, to observe the effect of these changes on the effects of the strategy by changing the values of the different parameters, thereby yielding a set of key influencing factors. These factors are variables that affect the resource allocation effect most significantly, and need to be focused and controlled in actual operation. And (3) performing parameter tuning on the virtual medical twin body based on the key influence factor set, wherein an optimization algorithm such as gradient descent or genetic algorithm is used in the step, and various parameters of the twin body are adjusted so that the actual medical system behaviors can be reflected and predicted more accurately. Through this process, medical analysis data is obtained which contains not only the in-depth analysis results of the current medical system, but also predictions and optimization suggestions for future conditions.

For example: the emergency department virtual medical twins of a trimethyl hospital contain detailed operational data of the past year. First, these data are spatially and spatially resolved, and the data are split in the dimensions of hours, days, zhou Dengshi, and in the dimensions of space such as different clinics and examination rooms. After data cleansing and normalization, a normalized data set is obtained that contains 10 key metrics (e.g., patient latency, doctor workload, device usage, etc.). Through the time series correlation analysis, it is found that there is a hysteresis effect of 2 hours between the patient arrival rate and the average waiting time, i.e. the arrival rate at the present moment affects the waiting time after 2 hours. Based on this association, a decision tree is constructed for predicting peak resource demand. Through cross verification and pruning, the node number of the decision tree is reduced from 50 to 15, but the accuracy is reduced from 85% to 83%. Through analysis of these rules, it was found that adding a triage nurse between 9 and 11 monday morning reduced the average waiting time from 40 minutes to 25 minutes. Clustering the resource allocation patterns by using a K-means algorithm, wherein the K value is set to be 3, and 3 typical resource utilization patterns are obtained: high efficiency, balanced and overloaded. Based on these patterns, a multi-objective optimization function is constructed that takes into account resource utilization, patient latency, and medical quality. Solving through NSGA-II algorithm to obtain a strategy set containing 50 optimal resource allocation schemes under different conditions. Sensitivity analysis shows that the number of triage nurses and the working time of doctors are two most critical factors affecting the system performance, and the sensitivity indexes are 0.6 and 0.5 respectively. Finally, based on the analysis results, parameters of the virtual medical twin are optimized, so that the system performance under different conditions can be accurately predicted, and real-time suggestions are provided for resource allocation. For example, during the simulated peak influenza period, twins suggest adding 2 doctors and 1 nurse on an original basis, and it is expected that the average patient waiting time can be reduced from 90 minutes to 45 minutes while reducing the average workload of doctors from 85% to 70%. These medical analysis data provide valuable decision support for hospital administrators, helping them to better cope with a variety of complex medical scenarios.

In a specific embodiment, the process of executing step S105 may specifically include the following steps:

(1) Performing multidimensional decomposition on the medical analysis data to obtain a patient characteristic vector and a hospital resource state matrix;

(2) Performing cluster analysis on the patient feature vectors to obtain patient grouping data, and constructing personalized risk assessment indexes based on the patient grouping data to obtain a patient risk distribution map;

(3) Carrying out dynamic load analysis on the hospital resource state matrix to obtain a resource utilization rate curve, and carrying out peak prediction based on the resource utilization rate curve to obtain a resource demand prediction table;

(4) Cross matching is carried out on the patient risk distribution diagram and the resource demand prediction table to obtain an initial resource allocation scheme, and a multi-constraint optimization rule set is constructed based on the initial resource allocation scheme to obtain a resource scheduling strategy;

(5) Performing discrete event simulation on the resource scheduling strategy to obtain a strategy evaluation result, and performing iterative optimization on the resource scheduling strategy based on the strategy evaluation result to obtain an optimized resource allocation scheme;

(6) And carrying out association analysis on the optimized resource allocation scheme and the patient individual data to obtain a personalized medical scheme and a resource optimization strategy.

Specifically, the high-dimensional data is converted into a low-dimensional representation using techniques such as Principal Component Analysis (PCA) or tensor decomposition. For patient data, the PCA method can extract key features to obtain a patient feature vector, and meanwhile, a similar method is applied to hospital resource data to obtain a hospital resource state matrix. These two results reflect the core characteristics of the patient and the dynamic state of the hospital resources, respectively. And carrying out cluster analysis on the patient characteristic vectors, and carrying out iterative computation on the similar patient characteristics by a common K-means algorithm to obtain patient grouping data. Based on these groupings, personalized risk assessment indicators are constructed, and a logistic regression model may be used to calculate the probability that each patient belongs to different risk levels, thereby obtaining a patient risk profile. This profile intuitively demonstrates the health risk level for different patient populations. And (3) carrying out dynamic load analysis on the hospital resource state matrix, and analyzing the dynamic change trend of resource use by adopting a time sequence analysis method, such as an autoregressive integral moving average model (ARIMA), so as to obtain a resource utilization rate curve. Based on the curve, a peak detection algorithm and a time sequence prediction model, such as Prophet, are used for predicting the resource demand in a period of time in the future, so as to obtain a resource demand prediction table. This predictive table provides prospective guidance for the rational allocation of resources.

Cross-matching the patient risk profile with the resource demand prediction table is a key step, and the process uses heuristic algorithms or linear programming methods, such as simplex methods, to optimally match the patient's demand with the hospital's resource supply, resulting in an initial resource allocation scheme. Based on the scheme, a multi-constraint optimization rule set is constructed, a plurality of constraint conditions such as medical quality, cost benefit, patient satisfaction and the like are considered, and a mathematical programming method such as integer programming is used to obtain a resource scheduling strategy. This strategy not only meets the basic requirements of resource allocation, but also seeks a balance among multiple objectives. Performing discrete event simulation on the resource scheduling policy is an important step in verifying the effectiveness of the policy. The process uses discrete event simulation software to simulate the execution effect of strategies under different conditions to obtain strategy evaluation results. These results reflect the behavior of the strategy in various possible situations, providing basis for further optimization. And (3) carrying out iterative optimization on the resource scheduling strategy based on a strategy evaluation result, wherein a heuristic algorithm, such as a genetic algorithm or a simulated annealing algorithm, is used for continuously adjusting and improving the strategy, and finally, an optimized resource allocation scheme is obtained. The scheme is optimized and verified for multiple rounds, and can achieve a good effect in practical application.

And carrying out association analysis on the optimized resource allocation scheme and the individual data of the patient, wherein an association rule mining algorithm, such as an Apriori algorithm, is used in the process, and the resource allocation scheme is combined with the requirements of the specific patient to obtain a personalized medical scheme and a resource optimization strategy. This final output not only allows for optimal configuration of the overall resources, but also provides personalized treatment advice for each patient's specific situation.

For example: the cardiovascular department of certain trimethyl hospitals applies this set of methods to optimize their medical services. First, medical data of the past year is subjected to multidimensional decomposition, and feature vectors (each vector contains 50 features such as age, blood pressure, blood fat and the like) of 10000 patients and a state matrix reflecting the use condition of 30 key resources of cardiovascular department are obtained. Patients were divided into 5 major groups, such as hypertension groups, coronary heart disease groups, etc., by the K-means clustering algorithm. Based on these groupings, a risk assessment model was constructed, yielding an average risk score (1-10 score) for each group of patients. Meanwhile, the resource state matrix is subjected to time sequence analysis, the ARIMA model is used for predicting the resource demand of the future week, and the fact that the use peak of the inspection equipment occurs at 9-11 am on monday is found, and the estimated use rate reaches 95%. The system matches the patient risk distribution with the resource demand forecast, and initially establishes a resource allocation scheme, such as suggesting that high risk patients (risk score > 8) are preferentially scheduled for treatment in a period of low resource utilization (such as 2-4 pm on tuesday, with a predicted use rate of 60%). Based on the initial scheme, a multi-objective optimization model considering medical quality, waiting time and cost is constructed, and a better resource scheduling strategy is obtained through integer programming solution. This strategy was validated by discrete event simulation, and simulation results showed that the new strategy could reduce the average patient latency from 45 minutes to 30 minutes while reducing the average workload of the physician from 85% to 75%.

After multiple rounds of iterative optimization, the system finally generates an optimized resource allocation scheme. This approach not only optimally allocates the overall resources, but also generates personalized visit advice for each patient. For example, for a 65 year old hypertensive patient (risk score 9.2), the system recommends a routine examination at 10 am on wednesday (expected device usage 70%) and schedules a 15 minute consultation with specialists at 2 pm on the same day, while recommending a personalized medication regimen.

In a specific embodiment, the process of executing step S106 may specifically include the following steps:

(1) Performing wavelet transformation on pre-acquired patient flow data to obtain a multi-scale time sequence decomposition result;

(2) Autoregressive analysis is carried out on the multi-scale time series decomposition result to obtain a short-term predicted value of the patient flow;

(3) Calculating the resource utilization rate of each department based on the short-term predicted value of the patient flow to obtain a dynamic resource load distribution diagram, and performing hot spot analysis on the dynamic resource load distribution diagram to obtain a resource bottleneck identification report;

(4) Generating a resource dynamic allocation instruction set based on the resource bottleneck identification report to obtain a resource allocation sub-strategy;

(5) Performing queue theoretical analysis on pre-acquired patient flow data to obtain a patient waiting time distribution function, and constructing triage priority rules based on the patient waiting time distribution function to obtain an intelligent triage scheme;

(6) Combining the intelligent diagnosis scheme with the individual characteristic data of the patient to obtain a personalized diagnosis path;

(7) Constructing a patient guiding mechanism based on the personalized diagnosis path to obtain a patient management sub-strategy;

(8) Performing collaborative optimization on the resource allocation sub-strategy and the patient management sub-strategy to obtain a service flow reconstruction scheme, and constructing a service quality monitoring index based on the service flow reconstruction scheme to obtain a service optimization sub-strategy;

(9) And integrating the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy to obtain the target management strategy.

In particular, wavelet transformation of pre-acquired patient flow data is a key step in achieving multi-scale time series decomposition. The wavelet transform can decompose time series data into components of different frequencies, effectively capturing various periodic patterns and incidents in the patient flow data. This process uses a Discrete Wavelet Transform (DWT) algorithm to decompose the raw patient flow data into high frequency detail coefficients and low frequency approximation coefficients, resulting in a multi-scale time series decomposition result. These decomposition results reflect the varying characteristics of patient flow over different time scales, providing a rich information basis for subsequent analysis. Autoregressive analysis is performed on the multi-scale time series decomposition results, and the process uses an autoregressive integrated moving average (ARIMA) model. The ARIMA model predicts patient flow in the future short term by analyzing the autocorrelation and moving average characteristics of the time series. Parameters of the model are determined through a maximum likelihood estimation method, and a short-term predicted value of the patient flow is obtained. The predicted value reflects the change trend of the patient flow in a future period (such as 24 hours in the future), and provides prospective guidance for resource allocation.

And calculating the resource utilization rate of each department based on the short-term predicted value of the patient flow, wherein the step compares the predicted patient flow with the resource capacity of each department, and calculates the resource utilization rate of different departments in different time periods, thereby obtaining a dynamic resource load distribution diagram. This profile intuitively shows the resource pressure at different time points for each department of the hospital. And carrying out hot spot analysis on the dynamic resource load distribution diagram, and identifying areas and time periods with abnormally high or low resource utilization rate by using a space statistical method such as Getis-Ord Gi statistics to obtain a resource bottleneck identification report. This report details potential resource bottleneck points, providing an explicit goal for subsequent resource allocation. Generating a dynamic allocation instruction set of the resources based on the resource bottleneck identification report, wherein a heuristic algorithm or an optimization algorithm, such as a genetic algorithm, is used in the process to prepare a series of resource allocation schemes, so as to obtain a resource allocation sub-strategy. This sub-strategy contains specific resource allocation instructions such as increasing the number of medical staff in a department or adjusting the equipment usage plan at a specific time.

And simultaneously, carrying out queue theoretical analysis on pre-acquired patient flow data, and analyzing the distribution characteristics of the patient waiting time by using an M/M/c queuing model to obtain a patient waiting time distribution function. The function describes probability distribution of patient waiting time under different conditions, and provides theoretical basis for optimizing patient management. And constructing triage priority rules based on the patient waiting time distribution function, wherein the process considers factors such as the emergency degree, waiting time, resource availability and the like of the patient, and an intelligent triage scheme is obtained. The scheme can dynamically adjust the treatment sequence of the patient and balance the use efficiency of medical resources and the waiting experience of the patient. Combining the intelligent triage scheme with the individual characteristic data of the patient, wherein a machine learning algorithm such as a decision tree or a random forest is used in the step, and the triage rule is combined with the specific condition of the patient to obtain a personalized treatment path. This path customizes the optimal flow of medical visits for each patient, including appointment time, examination order, etc. A patient guiding mechanism is built based on the personalized treatment path, and a set of intelligent navigation and reminding system is built in the process to obtain a patient management sub-strategy. This sub-strategy can guide the patient in real time, optimizing his experience of seeking medical advice.

And carrying out cooperative optimization on the resource allocation sub-strategy and the patient management sub-strategy, wherein a multi-objective optimization algorithm such as pareto optimization is used in the step, so that balance is sought between resource efficiency and patient experience, and a service flow reconstruction scheme is obtained. The proposal comprehensively considers the resource utilization and the patient demand, and optimizes the medical service flow comprehensively. The service quality monitoring index is constructed based on a service flow reconstruction scheme, and a series of Key Performance Indexes (KPIs) such as patient satisfaction, average waiting time and the like are defined in the process to obtain a service optimization sub-strategy. This sub-strategy provides a complete set of quality of service assessment and improvement mechanisms. Integrating the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy, wherein the step uses a system integration technology to organically combine the three sub-strategies to form a unified management framework so as to obtain the target management strategy. The target management strategy is a comprehensive, dynamic and intelligent medical management scheme, and can respond to the flow change of the patient in real time, optimize the resource allocation and improve the service quality.

For example: the emergency department of a complex hospital is using this set of methods to optimize its operation. First, the patient flow data of the past year is subjected to wavelet transformation, and the data is decomposed into high frequency (hour level), intermediate frequency (day level) and low frequency (week level) components. By autoregressive analysis, patient flow is predicted for the next 24 hours, and peak visits are found at 8-10 monday morning, with an expected 30% increase in patient numbers over usual. Based on the prediction, the resource utilization rate of each area of the emergency is calculated, and the resource utilization rate of the sub-diagnosis area and the internal medicine diagnosis room is found to exceed 90% in the predicted peak period, so that a potential bottleneck is formed.

For this case, the resource orchestration strategy suggests adding 2 triage nurses and 1 physician in monday morning and opening 2 additional clinics. Meanwhile, through queue theory analysis, it was found that the average waiting time of patients during peak hours may exceed 60 minutes. Based on this, the patient management sub-strategy constructs a dynamic triage scheme, triages patients according to urgency, and generates personalized triage paths for each patient. For example, for a chest pain patient aged 65, the system recommends that he or she prefer to perform an electrocardiographic examination, and then immediately schedule a physician diagnosis.

The service optimization sub-strategy constructs a set of monitoring indexes including average waiting time of patients, diagnosis and treatment time utilization rate of doctors and the like, and sets corresponding target values. Through the collaborative implementation of the strategies, the average waiting time of the peak period can be expected to be reduced from 60 minutes to 30 minutes, and the average workload of doctors is reduced from 95% to 85%, so that the operation efficiency and the service quality of emergency departments are greatly improved.

The medical data management method based on artificial intelligence in the embodiment of the present application is described above, and the medical data management system based on artificial intelligence in the embodiment of the present application is described below, referring to fig. 2, an embodiment of the medical data management system based on artificial intelligence in the embodiment of the present application includes:

A processing module 201, configured to preprocess multi-source medical data collected from the electronic medical record system, the medical device, and the wearable device, to obtain a standardized medical data set;

the classification module 202 is configured to intelligently classify and label the standardized medical data set to obtain structured medical knowledge data;

A construction module 203 for constructing a digital twin based on the-structured medical knowledge data to obtain a virtual medical twin;

The extraction module 204 is used for performing data mining and feature extraction on the virtual medical twin body to obtain medical analysis data;

The analysis module 205 is configured to comprehensively analyze patient individual data and hospital operation data collected in real time based on the medical analysis data, so as to obtain a personalized medical solution and a resource optimization strategy;

The optimizing module 206 is configured to perform data optimization on the personalized medical solution and the resource optimization policy through pre-acquired patient flow data, so as to obtain a target management policy, where the target management policy includes: resource allocation sub-policies, patient management sub-policies, and service optimization sub-policies.

Through the cooperation of the components, the effective integration of heterogeneous data is realized through the pretreatment and standardization of the multi-source medical data, the problem of information island in the traditional system is solved, and a solid data foundation is laid for the subsequent intelligent analysis. Secondly, by intelligently classifying and labeling the standardized medical data set, the structured medical knowledge data is obtained, which not only improves the understandability and usability of the data, but also provides more comprehensive and accurate information support for medical decision. The virtual medical twin constructed based on the structured medical knowledge data provides a comprehensive and dynamic digital mapping for medical management, so that medical institutions can perform various simulations and predictions in a virtual environment, and decision risks are greatly reduced. The virtual medical twin body is subjected to data mining and feature extraction, and the obtained medical analysis data provides scientific basis for the establishment of personalized medical schemes and the generation of resource optimization strategies. The decision support based on the depth data analysis remarkably improves the accuracy of medical services and the efficiency of resource utilization. By comprehensively analyzing the patient individual data and the hospital operation data acquired in real time, the method can generate a truly personalized medical scheme, meets the requirement of accurate medical treatment, can dynamically adjust resource allocation according to real-time conditions, and improves the operation efficiency of hospitals. In particular, the personalized medical scheme and the resource optimization strategy are further optimized through the pre-collected patient flow data, the obtained target management strategy covers multiple aspects of resource allocation, patient management, service optimization and the like, and a comprehensive intelligent solution is provided for hospital management. The dynamic and intelligent management mode can effectively cope with the complexity and variability of the medical environment, and the medical service quality and the patient satisfaction are obviously improved. In addition, the application of the method can also reduce the medical cost, reduce the waste of medical resources and create greater economic benefits for medical institutions. The method relates to the processing of a large amount of sensitive medical data, fully considers the problems of data safety and privacy protection in the construction and implementation processes, adopts multiple safety measures such as data desensitization, access control and the like, can effectively protect the privacy of patients, and improves the efficiency and the accuracy of medical data management based on artificial intelligence.

The present application also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, and may also be a volatile computer readable storage medium, in which instructions are stored which, when executed on a computer, cause the computer to perform the steps of the artificial intelligence-based medical data management method.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.

The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A medical data management method based on artificial intelligence is characterized in that, the medical data management method based on artificial intelligence comprises the following steps:

preprocessing multi-source medical data acquired from an electronic medical record system, medical equipment and wearable equipment to obtain a standardized medical data set;

Performing intelligent classification and labeling on the standardized medical data set to obtain structured medical knowledge data;

constructing a digital twin body based on the structured medical knowledge data to obtain a virtual medical twin body;

Performing data mining and feature extraction on the virtual medical twin body to obtain medical analysis data;

Based on the medical analysis data, comprehensively analyzing patient individual data and hospital operation data acquired in real time to obtain a personalized medical scheme and a resource optimization strategy, wherein the method specifically comprises the following steps of: performing multidimensional decomposition on the medical analysis data to obtain a patient characteristic vector and a hospital resource state matrix; performing cluster analysis on the patient feature vectors to obtain patient grouping data, and constructing personalized risk assessment indexes based on the patient grouping data to obtain a patient risk distribution map; performing dynamic load analysis on the hospital resource state matrix to obtain a resource utilization rate curve, and performing peak prediction based on the resource utilization rate curve to obtain a resource demand prediction table; cross matching is carried out on the patient risk distribution diagram and the resource demand prediction table to obtain an initial resource allocation scheme, and a multi-constraint optimization rule set is constructed based on the initial resource allocation scheme to obtain a resource scheduling strategy; performing discrete event simulation on the resource scheduling strategy to obtain a strategy evaluation result, and performing iterative optimization on the resource scheduling strategy based on the strategy evaluation result to obtain an optimized resource allocation scheme; performing association analysis on the optimized resource allocation scheme and patient individual data to obtain the personalized medical scheme and a resource optimization strategy;

Data optimization is carried out on the personalized medical scheme and the resource optimization strategy through pre-acquired patient flow data to obtain a target management strategy, and the method specifically comprises the following steps: performing wavelet transformation on pre-acquired patient flow data to obtain a multi-scale time sequence decomposition result; performing autoregressive analysis on the multi-scale time series decomposition result to obtain a short-term predicted value of the patient flow; calculating the resource utilization rate of each department based on the short-term predicted value of the patient flow to obtain a dynamic resource load distribution diagram, and performing hot spot analysis on the dynamic resource load distribution diagram to obtain a resource bottleneck identification report; generating a resource dynamic allocation instruction set based on the resource bottleneck identification report to obtain a resource allocation sub-strategy; performing queue theoretical analysis on pre-acquired patient flow data to obtain a patient waiting time distribution function, and constructing diagnosis priority rules based on the patient waiting time distribution function to obtain an intelligent diagnosis scheme; combining the intelligent diagnosis scheme with individual characteristic data of the patient to obtain a personalized diagnosis path; constructing a patient guiding mechanism based on the personalized diagnosis path to obtain a patient management sub-strategy; performing collaborative optimization on the resource allocation sub-strategy and the patient management sub-strategy to obtain a service flow reconstruction scheme, and constructing a service quality monitoring index based on the service flow reconstruction scheme to obtain a service optimization sub-strategy; and integrating the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy to obtain the target management strategy.

2. The medical data management method based on artificial intelligence of claim 1, wherein preprocessing the multi-source medical data collected from the electronic medical record system, the medical device and the wearable device to obtain a standardized medical data set comprises:

performing data cleaning on the multi-source medical data to obtain initial cleaning data, and performing format consistency processing on the initial cleaning data to obtain uniform format data;

performing data desensitization processing on the unified format data to obtain privacy protection data, and performing time sequence alignment on the privacy protection data to obtain time-consistent data;

Performing interpolation processing on the missing values in the time-consistent data to obtain a complete data set, and performing outlier detection and processing on the complete data set to obtain outlier correction data;

Carrying out data normalization processing on the abnormal correction data to obtain normalized data, and carrying out dimension reduction processing on the normalized data through a principal component analysis algorithm to obtain dimension reduction data;

Carrying out data segmentation and coding on the reduced data to obtain coded data;

and performing metadata annotation and index creation on the encoded data to obtain the standardized medical data set.

3. The artificial intelligence based medical data management method according to claim 1, wherein the intelligent classification and labeling of the standardized medical data set to obtain the structured medical knowledge data comprises:

performing multidimensional feature extraction on the standardized medical data set to obtain a medical feature vector set, and performing unsupervised cluster analysis on the medical feature vector set to obtain an initial medical data category;

Semantic mapping is carried out on the initial medical data category based on a preset medical ontology to obtain semantic enhanced medical data, and multi-modal relation extraction is carried out on the semantic enhanced medical data to obtain a medical entity association map;

Medical term standardization processing is carried out on the medical entity association map to obtain a standardized medical vocabulary network, and a medical event time sequence chain is constructed based on the standardized medical vocabulary network to obtain medical time sequence association data;

Performing pattern recognition and mining on the medical time sequence associated data to obtain a medical management pattern library;

performing simulation evaluation on the medical decision flow based on the medical management mode library to obtain a decision influence index set;

carrying out data fusion on the decision-making influence index set and the medical time sequence associated data to obtain a fused medical knowledge graph;

And carrying out structural processing and index establishment on the fusion medical knowledge graph to obtain the structural medical knowledge data.

4. The artificial intelligence based medical data management method of claim 1, wherein the constructing a digital twin based on the structured medical knowledge data to obtain a virtual medical twin comprises:

Carrying out semantic analysis on the structured medical knowledge data to obtain a medical entity relation network, and constructing a three-dimensional space model based on the medical entity relation network to obtain an initial virtual medical environment;

performing dynamic parameter mapping on the initial virtual medical environment to obtain an interactable medical scene;

performing real-time data stream access on the interactive medical scene to obtain a data-driven virtual medical model;

Establishing a multi-dimensional evaluation index based on the data driving type virtual medical model to obtain a medical twin evaluation index set, and performing simulation verification on the medical twin evaluation index set to obtain a twin body operation rule set;

Performing medical procedure digital mapping based on the twin body operation rule set to obtain virtual medical operation data;

and performing multi-scene test and optimization on the virtual medical operation data to obtain the virtual medical twin.

5. The artificial intelligence based medical data management method according to claim 1, wherein the performing data mining and feature extraction on the virtual medical twin body to obtain medical analysis data comprises:

performing space-time data decomposition on the virtual medical twin body to obtain a multi-dimensional medical data stream, and performing data cleaning and standardization processing on the multi-dimensional medical data stream to obtain a standardized medical data set;

Performing time sequence correlation analysis on the normalized medical data set to obtain a medical event correlation chain;

constructing a medical decision tree based on the medical event association chain to obtain an initial medical decision rule set;

cross-verifying and pruning optimization are carried out on the initial medical decision rule set to obtain a simplified medical decision rule;

Mining a medical resource utilization mode based on the simplified medical decision rule to obtain a resource allocation feature map;

Performing cluster analysis on the resource allocation feature map to obtain a medical resource utilization efficiency evaluation index;

Constructing a multi-objective optimization function based on the medical resource utilization efficiency evaluation index to obtain a resource allocation strategy set;

Performing sensitivity analysis on the resource allocation strategy set to obtain a key influence factor set;

and performing parameter tuning on the virtual medical twin body based on the key influence factor set to obtain the medical analysis data.

6. An artificial intelligence based medical data management system for performing the artificial intelligence based medical data management method of any one of claims 1-5, the artificial intelligence based medical data management system comprising:

The analysis module is used for comprehensively analyzing the patient individual data and the hospital operation data acquired in real time based on the medical analysis data to obtain a personalized medical scheme and a resource optimization strategy, and specifically comprises the following steps: performing multidimensional decomposition on the medical analysis data to obtain a patient characteristic vector and a hospital resource state matrix; performing cluster analysis on the patient feature vectors to obtain patient grouping data, and constructing personalized risk assessment indexes based on the patient grouping data to obtain a patient risk distribution map; performing dynamic load analysis on the hospital resource state matrix to obtain a resource utilization rate curve, and performing peak prediction based on the resource utilization rate curve to obtain a resource demand prediction table; cross matching is carried out on the patient risk distribution diagram and the resource demand prediction table to obtain an initial resource allocation scheme, and a multi-constraint optimization rule set is constructed based on the initial resource allocation scheme to obtain a resource scheduling strategy; performing discrete event simulation on the resource scheduling strategy to obtain a strategy evaluation result, and performing iterative optimization on the resource scheduling strategy based on the strategy evaluation result to obtain an optimized resource allocation scheme; performing association analysis on the optimized resource allocation scheme and patient individual data to obtain the personalized medical scheme and a resource optimization strategy;

The optimizing module is used for carrying out data optimization on the personalized medical scheme and the resource optimizing strategy through the pre-acquired patient flow data to obtain a target management strategy, wherein the target management strategy comprises the following steps: the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy specifically comprise: performing wavelet transformation on pre-acquired patient flow data to obtain a multi-scale time sequence decomposition result; performing autoregressive analysis on the multi-scale time series decomposition result to obtain a short-term predicted value of the patient flow; calculating the resource utilization rate of each department based on the short-term predicted value of the patient flow to obtain a dynamic resource load distribution diagram, and performing hot spot analysis on the dynamic resource load distribution diagram to obtain a resource bottleneck identification report; generating a resource dynamic allocation instruction set based on the resource bottleneck identification report to obtain a resource allocation sub-strategy; performing queue theoretical analysis on pre-acquired patient flow data to obtain a patient waiting time distribution function, and constructing diagnosis priority rules based on the patient waiting time distribution function to obtain an intelligent diagnosis scheme; combining the intelligent diagnosis scheme with individual characteristic data of the patient to obtain a personalized diagnosis path; constructing a patient guiding mechanism based on the personalized diagnosis path to obtain a patient management sub-strategy; performing collaborative optimization on the resource allocation sub-strategy and the patient management sub-strategy to obtain a service flow reconstruction scheme, and constructing a service quality monitoring index based on the service flow reconstruction scheme to obtain a service optimization sub-strategy; and integrating the resource allocation sub-strategy, the patient management sub-strategy and the service optimization sub-strategy to obtain the target management strategy.

7. A computer readable storage medium having instructions stored thereon, which when executed by a processor, implement the artificial intelligence based medical data management method of any of claims 1-5.