Chest pain patient medical data intelligent management method and system based on artificial intelligence

Abstract

The invention discloses a chest pain patient medical data intelligent management method and system based on artificial intelligence, and the method comprises the steps: obtaining multi-source heterogeneous data in real time according to an electronic medical record, image data and an inspection report of a chest pain patient, converting the data from different sources into a standardized data format in a unified manner, and generating a time-synchronized multi-source data set; for the multi-source data set, identifying missing, wrong or incomplete parts in the data, and generating a patient data file through a data restoration model of the generative adversarial network; for the patient data file, recording the change history of the data in real time, and generating a real-time synchronous patient data file through a distributed data synchronization algorithm; and aiming at the real-time synchronized patient data file, generating a diagnosis suggestion and a treatment scheme, providing a basis and confidence of a diagnosis result, and generating a final intelligent diagnosis support report. According to the embodiment of the invention, the efficiency and accuracy of data management can be improved, and the diagnosis and treatment effects of patients with chest pain are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data management, in particular to a chest pain patient medical data intelligent management method and system based on artificial intelligence. BACKGROUND

[0002] With the increasing demand for chest pain treatment, the medical data management of chest pain patients faces many challenges, including the diversity and heterogeneity of data sources, data quality problems, information update lag, and the limitations of artificial intelligence applications. The electronic medical records, image data and test reports generated by chest pain patients during the treatment process come from different medical devices and systems, with different formats, and traditional data management is difficult to efficiently integrate. In addition, there are often missing, incorrect or incomplete parts in the data, which affects the accuracy of medical decision-making, and existing systems often lack real-time update mechanism, so that doctors cannot obtain the latest patient information. SUMMARY

[0003] The purpose of the present application is to provide a chest pain patient medical data intelligent management method and system based on artificial intelligence, to solve the problems in the prior art, and to improve the efficiency and accuracy of data management, and to improve the diagnosis and treatment effect of chest pain patients.

[0004] One embodiment of the present application provides a chest pain patient medical data intelligent management method based on artificial intelligence, which comprises: According to the electronic medical records, image data and test reports of chest pain patients, a data acquisition framework based on edge computing is adopted to obtain real-time multi-source heterogeneous data, and through an adaptive data format conversion algorithm, different sources of data are uniformly converted into a standardized data format to generate a time-synchronized multi-source data set; For the time-synchronized multi-source data set, a deep learning-based anomaly detection algorithm is used to identify missing, incorrect or incomplete parts in the data, and a data repair model based on a generative adversarial network is used to repair and complete low-quality data to generate a patient data archive; For the patient data archive, a data update mechanism based on blockchain is used to record the change history of the data in real time, and a distributed data synchronization algorithm is used to ensure that doctors obtain the latest patient information to generate a real-time synchronized patient data archive; For the real-time synchronized patient data archive, a multi-modal fusion-based artificial intelligence analysis model is used to combine the clinical characteristics and historical case data of chest pain patients to generate diagnosis suggestions and treatment plans, and an explainability module of the artificial intelligence analysis model is used to provide the basis and confidence of the diagnosis results to generate a final intelligent diagnosis support report.

[0005] Optionally, according to the electronic medical record, image data and test report of the patient with chest pain, a data acquisition framework based on edge computing is adopted to obtain multi-source heterogeneous data in real time, through an adaptive data format conversion algorithm, different sources of data are uniformly converted into a standardized data format, a time-synchronized multi-source data set is generated, including: According to the electronic medical record, image data and test report of the patient with chest pain, a data acquisition framework based on edge computing is adopted to obtain multi-source heterogeneous data in real time through distributed edge nodes; each edge node is equipped with a lightweight data caching mechanism to ensure the real-time and continuity of data acquisition; For the collected multi-source heterogeneous data, a format recognition model based on deep learning is used to automatically identify the format type of different data sources, and through an adaptive parsing algorithm, data in different formats is parsed into a structured intermediate representation to generate a preliminary parsed data set; For the preliminary parsed data set, a standardized conversion algorithm based on the combination of rule engine and machine learning is used to uniformly convert data from different sources into a standardized data format, and through a context-aware mapping rule library, the conversion logic is dynamically adjusted to generate a standardized data set; For the standardized data set, a time synchronization algorithm based on dynamic time warping is used to eliminate the timestamp differences of different data sources, and through an interpolation filling method, missing data is supplemented to generate a time-synchronized multi-source data set.

[0006] Optionally, for the time-synchronized multi-source data set, a deep learning-based anomaly detection algorithm is used to identify missing, incorrect or incomplete parts of the data, and through a data repair model based on a generative adversarial network, low-quality data is repaired and completed to generate a patient data archive, including: For the time-synchronized multi-source data set, an anomaly detection algorithm based on a deep autoencoder is used to identify missing, incorrect or incomplete parts of the data through reconstruction error, and through a dynamic threshold adjustment mechanism, a data quality evaluation report is generated to mark low-quality data points; For the marked low-quality data points, a data repair model based on a generative adversarial network is trained, wherein the generator is used to generate reasonable completion values for missing data, and the discriminator is used to evaluate the authenticity of the generated data, and through a conditional constraint mechanism, the generated data is ensured to conform to medical logic and context consistency; The low-quality data points are input into the trained data repair model to generate completion values for missing or incorrect data, and through a multi-round iterative optimization method, the accuracy and reliability of the repaired data are gradually improved to generate a repaired data set; For the repaired data set, a verification method based on medical expert rules is used to verify the reasonableness of the repaired data in combination with historical case data and clinical guidelines, and through a data quality scoring mechanism, a high-quality patient data archive is generated.

[0007] Optionally, the patient data archive adopts a blockchain-based data update mechanism to record the change history of the data in real time, ensures that doctors obtain the latest patient information through a distributed data synchronization algorithm, and generates a real-time synchronized patient data archive, including: For high-quality patient data archives, a blockchain-based data update mechanism is adopted to record each data change as a transaction on the blockchain, automatically verify the legality of data changes through a smart contract, and generate an unalterable data change history; According to the internal network environment and external Internet connection of the hospital, a data synchronization framework based on distributed ledger technology is constructed, a lightweight consensus algorithm is used to ensure the data consistency between multiple nodes, and a preliminary distributed synchronization network is generated; For nodes in the distributed synchronization network, a real-time data synchronization algorithm based on timestamps is used to ensure that doctors' terminals can obtain the latest patient information, and a conflict detection and resolution mechanism is used to dynamically handle data conflicts between multiple nodes, and generate conflict-free synchronized data; For synchronized data, a version control-based data integration method is used to integrate the latest data with historical change records, and a dynamic permission control mechanism is used to ensure that only authorized personnel can access sensitive information, and a real-time synchronized patient data archive is generated.

[0008] Optionally, the real-time synchronized patient data archive adopts a multi-modal fusion-based artificial intelligence analysis model, combines the clinical features and historical case data of chest pain patients to generate diagnosis suggestions and treatment plans, and provides the basis and confidence of the diagnosis results through the explainability module of the artificial intelligence analysis model. Generate the final intelligent diagnosis support report, including: For real-time synchronized patient data archives, a multi-modal feature extraction network based on multi-head attention mechanism is used to extract text features of electronic medical records, visual features of image data, and numerical features of test reports, and through a cross-modal attention mechanism, the features of different modalities are weighted and fused to generate multi-modal fusion feature representation; For multi-modal fusion feature representation, a diagnosis model based on deep neural network is used to combine the clinical features and historical case data of chest pain patients to generate preliminary diagnosis suggestions, and through the explainability module of the diagnosis model, the basis and confidence of the diagnosis results are provided; For preliminary diagnosis suggestions, a treatment plan optimization algorithm based on reinforcement learning is used to dynamically adjust the treatment plan in combination with the individual characteristics of patients and the current situation of hospital resources, and through a multi-objective optimization model, the treatment effect and side effect risk are balanced to generate optimized diagnosis suggestions and treatment plans; For the optimized diagnosis suggestion and treatment plan, the report generation method based on natural language generation technology is adopted to convert the analysis result into an easy-to-understand text description, and through the visualization technology, the final intelligent diagnosis support report containing the diagnosis basis, treatment plan and confidence is generated.

[0009] Still another embodiment of the present application provides an artificial intelligence-based intelligent management system for medical data of chest pain patients, comprising: The acquisition module is configured to acquire multi-source heterogeneous data in real time according to the electronic medical records, image data and test reports of the chest pain patients by using an edge computing-based data acquisition framework, and convert the data of different sources into a standardized data format by using a self-adaptive data format conversion algorithm to generate a time-synchronized multi-source data set. The recognition module is configured to identify missing, erroneous or incomplete parts in the data by using a deep learning-based anomaly detection algorithm for the time-synchronized multi-source data set, repair and complete the low-quality data by using a data repair model of a generative adversarial network, and generate a patient data archive. The generation module is configured to record the change history of the data in real time by using a blockchain-based data update mechanism for the patient data archive, ensure that the doctor obtains the latest patient information by using a distributed data synchronization algorithm, and generate a real-time synchronized patient data archive. The management module is configured to generate diagnosis suggestions and treatment plans by using a multi-modal fusion-based artificial intelligence analysis model for the real-time synchronized patient data archive, combining the clinical features and historical case data of the chest pain patients, provide the basis and confidence of the diagnosis result by using an explainability module of the artificial intelligence analysis model, and generate a final intelligent diagnosis support report.

[0010] Still another embodiment of the present application provides a storage medium having a computer program stored therein, wherein the computer program is configured to execute the method described in any one of the above embodiments when running.

[0011] Still another embodiment of the present application provides an electronic device comprising a memory and a processor, wherein the memory has a computer program stored therein, and the processor is configured to execute the computer program to execute the method described in any one of the above embodiments.

[0012] Compared with the prior art, the application provides a kind of medical data intelligent management method for chest pain patient based on artificial intelligence, according to the electronic medical record, image data and test report of chest pain patient, real-time acquisition of multi-source heterogeneous data, different sources of data are uniformly converted into standardized data format, and time-synchronized multi-source data set is generated;For multi-source data set, identify the missing, error or incomplete part in the data, generate patient data archives by generating the data repair model of generative adversarial network;For patient data archives, real-time record the change history of data, generate real-time synchronized patient data archives by distributed data synchronization algorithm;For real-time synchronized patient data archives, generate diagnosis suggestion and treatment plan, provide the basis and confidence of diagnosis result, and generate the final intelligent diagnosis support report, so as to improve the efficiency and accuracy of data management, improve the diagnosis and treatment effect of chest pain patient. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 The hardware structure block diagram of the computer terminal of the artificial intelligence-based medical data intelligent management method for chest pain patient provided by the embodiment of the application is shown in the figure. Figure 2 The flowchart of the artificial intelligence-based medical data intelligent management method for chest pain patient provided by the embodiment of the application is shown in the figure. Figure 3 The structure diagram of the artificial intelligence-based medical data intelligent management system for chest pain patient provided by the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0014] The embodiments described below with reference to the accompanying drawings are exemplary and are used to explain the application, but cannot be interpreted as a limitation of the application.

[0015] The embodiment of the application first provides an artificial intelligence-based medical data intelligent management method for chest pain patient, which can be applied to electronic equipment, such as computer terminal, specifically, such as ordinary computer, etc.

[0016] The following will be described in detail taking the running on computer terminal as an example. Figure 1 The hardware structure block diagram of the computer terminal of the artificial intelligence-based medical data intelligent management method for chest pain patient provided by the embodiment of the application is shown in the figure. Figure 1 As shown in the figure, the computer device includes a processor, a memory and a network interface connected by a system bus, wherein the memory can include a non-volatile storage medium and an internal memory.

[0017] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, which, when executed, can make the processor execute any kind of artificial intelligence-based medical data intelligent management method for chest pain patient.

[0018] The processor is configured to provide computing and control capabilities to support the operation of the entire computer device.

[0019] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which can make the processor execute any kind of artificial intelligence-based intelligent management method for medical data of chest pain patients when the computer program is executed by the processor.

[0020] The network interface is configured to perform network communication, such as sending assigned tasks, etc. Those skilled in the art can understand that, Figure 1 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0021] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0022] Referring to Figure 2 The embodiments of the present application provide an artificial intelligence-based intelligent management method for medical data of chest pain patients, which can include the following steps: S201, according to the electronic medical record, image data and test report of the chest pain patient, a data acquisition framework based on edge computing is adopted to acquire multi-source heterogeneous data in real time, different sources of data are uniformly converted into a standardized data format through an adaptive data format conversion algorithm, and a time-synchronized multi-source data set is generated; Specifically, the edge computing framework allows data processing on nearby devices or edge nodes where the patient is located, reducing data transmission latency and improving real-time performance. For example, when first aid personnel receive a chest pain patient on site, they can collect the patient's electrocardiogram, vital sign monitoring data, medical record information, etc. through mobile devices, and upload these data to the edge node in real time. In this process, through the adaptive data format conversion algorithm, the system can automatically identify and convert different sources of data formats, such as converting the DICOM format of medical images, the JSON format of electronic medical records, and the CSV format of test reports into standardized data formats. This data integration method not only supports compatibility of different types of data, but also ensures consistency of data architecture, laying a foundation for subsequent data analysis.

[0023] This step plays an important role in the medical management of chest pain patients. First, through edge computing, real-time data collection and processing are realized, ensuring that medical staff can quickly obtain multi-source heterogeneous data of patients, so as to make rapid response and correct decision in emergency situations. Second, the standardized data format unifies data from different sources, making subsequent analysis, processing and storage more efficient and convenient, avoiding the difficulty of data integration due to inconsistent formats. In addition, the generated time-synchronized multi-source data set provides a reliable data foundation for subsequent anomaly detection, diagnosis suggestions and treatment plan generation, ensuring the effectiveness and accuracy of the entire intelligent management system.

[0024] Specifically, based on the electronic medical records, image data and test reports of chest pain patients, a data collection framework based on edge computing can be used to obtain multi-source heterogeneous data in real time through distributed edge nodes; each edge node is equipped with a lightweight data caching mechanism to ensure the real-time and continuity of data collection; In this step, the data collection framework based on edge computing realizes real-time acquisition of multi-source heterogeneous data of chest pain patients through a distributed edge node architecture. Each edge node can be regarded as a small data processing center that can quickly respond and process data in an environment close to the data generation site. In order to ensure the real-time and continuity of data collection, these edge nodes are equipped with a lightweight data caching mechanism to temporarily store data in case of network delay or sudden situations, preventing data loss. For example, when an ambulance receives the vital sign data of a chest pain patient, the edge node can process these data in real time and cache them until the network is restored to normal.

[0025] The implementation of this step can significantly improve the efficiency and accuracy of data collection, especially in emergency medical situations, real-time acquisition of patient information can help medical staff make quick and reasonable decisions. This real-time nature also enhances the ability to respond quickly to changes in patient condition, thereby improving treatment outcomes.

[0026] In this step, we first design an edge computing-based data collection architecture to ensure that we can quickly obtain multi-source heterogeneous data for chest pain patients. Specifically, edge nodes are deployed in ambulances, hospital emergency departments, and other important locations, which enables data to be processed and stored closer to the patient, reducing data transmission delays. For example, when an ambulance receives a chest pain patient, the on-board edge node will promptly collect the patient's vital signs (such as heart rate, blood pressure) and other relevant information, such as electrocardiogram data, which not only improves data collection speed but also provides timely data support for subsequent processing.

[0027] To ensure the continuity and real-time nature of data collection, each edge node is equipped with a lightweight data caching mechanism. When network connectivity is unstable or delayed, the caching mechanism can temporarily store collected data until network conditions return to normal. For example, in an emergency scene, data transmission may be interrupted due to unstable signals, at which point the caching mechanism will automatically take over to ensure that data is not lost due to network problems. After the network is restored, the edge node will upload the stored data in batches to the central server or cloud, ensuring data integrity and consistency.

[0028] In addition, edge computing not only improves the real-time nature of data collection but also enables preliminary data processing on edge nodes, such as data preprocessing and format conversion. This local processing approach not only reduces bandwidth consumption but also reduces the burden on the central server, significantly improving the overall system's operational efficiency.

[0029] For the collected multi-source heterogeneous data, a deep learning-based format recognition model is used to automatically identify the format types of different data sources, and through an adaptive parsing algorithm, the data in different formats is parsed into a structured intermediate representation, generating a preliminary parsed data set. In this step, the system uses a deep learning-based format recognition model to identify the format of the collected multi-source heterogeneous data. This process first requires analyzing different sources of data (such as electronic medical records, image files, and test reports) to automatically identify their format types (such as DICOM, XML, or CSV). Through an adaptive parsing algorithm, the system can select the appropriate parsing method based on the identified format type to convert these multi-source heterogeneous data into a unified structured intermediate representation. For example, for image data, the system may extract key information such as image size, resolution, and file type, while for electronic medical records, the system may focus on the text content and structure.

[0030] The successful implementation of this step ensures that data in various formats can be effectively parsed and integrated, laying the foundation for subsequent data standardization. This not only improves data processing efficiency but also ensures data integrity and accuracy, providing reliable data support for subsequent analysis processes.

[0031] In this step, the system automatically identifies the collected, heterogeneous data from multiple sources using a deep learning-based format recognition model. Data from different sources may use different formats. For example, electronic medical records may be in JSON or XML, while imaging data is typically in DICOM format, and laboratory reports may be in CSV format. To efficiently process this data, the system first uses a trained deep learning model to analyze the structural characteristics of each piece of data and automatically identify its format type. This process ensures that subsequent data processing strategies can be tailored to the specific format.

[0032] After recognition, the system uses an adaptive parsing algorithm to parse data in various formats into a unified structured intermediate representation. For example, DICOM files require the extraction of image information, patient name, age, and other metadata. Each field must be parsed for further analysis and processing. This parsing process utilizes natural language processing (NLP) technology to analyze text content. For example, it can extract information such as a patient's medical history and current symptoms from electronic medical records and convert it into a structured form, facilitating subsequent data processing.

[0033] Ultimately, the system integrates all parsed data into a preliminarily parsed dataset. With a unified format, this data can be directly used for subsequent standardized processing and analysis. This dataset not only reduces data processing complexity but also ensures that inconsistent formats prevent misleading data during analysis. For example, through structured representation, clinicians can more clearly view all relevant patient information, thereby improving the quality of medical decision-making.

[0034] For the initially parsed dataset, a standardized conversion algorithm based on a combination of a rule engine and machine learning is used to uniformly convert data from different sources into a standardized data format. Through a context-aware mapping rule library, the conversion logic is dynamically adjusted to generate a standardized dataset. In this step, the system will use a standardized conversion algorithm based on a combination of rule engines and machine learning for the previously parsed data set. This process first requires the construction of a mapping relationship containing a rule library. By defining the mapping rules for each data format, the system can automatically convert data from different sources into a unified standardized data format. The context-aware mapping rule library can dynamically adjust the conversion logic according to the source and context information of the data, ensuring the conversion effect in different situations. For example, when processing medical record formats from different hospitals, the system will automatically map and convert various fields according to the rules defined in advance.

[0035] This process ensures that the final generated data remains consistent in structure, reducing the complexity of subsequent data analysis and processing. By unifying the standardized data format, different systems and tools can efficiently utilize these data, further enhancing the collaboration capabilities and efficiency of the overall medical management system.

[0036] In this step, the system runs a standardized conversion algorithm based on a combination of rule engines and machine learning to achieve the uniform conversion of collected data into a standardized format. First, the system needs to build a context-aware mapping rule library that will contain conversion rules for different data sources. When the system receives the preliminary parsed data set, it can quickly determine which fields need to be converted and how to convert them through these rules. For example, for medical records from different hospitals, there may be cases where the same information uses different field names, and the mapping rules will ensure that these fields are unified into standardized names.

[0037] When processing data, the rule engine first analyzes the characteristics of the input data and looks up the corresponding mapping relationship in the context rule library. For example, if the "patient age" field in a hospital's medical record data is labeled as "Age", while another data uses "Patient Age", the rule engine will standardize it to the unified "Patient_Age". At the same time, combined with machine learning technology, the system will continuously optimize these mapping rules through historical data in the process of continuously processing data, to improve the accuracy and efficiency of subsequent conversion.

[0038] Through the context-aware adjustment mechanism, the system can also dynamically adjust the conversion logic when the data sources are complex or new data formats appear. For example, if a new format of data source is added, the system will automatically update the mapping rules to ensure the compatibility of data conversion. This flexible and powerful standardization processing capability not only reduces the workload of data processing, but also enhances the system's ability to adapt to new data sources, providing clear and standardized data support for clinical doctors.

[0039] For the standardized dataset, a time synchronization algorithm based on dynamic time warping (DTW) is employed to eliminate the timestamp discrepancies across different data sources. Interpolation methods are used to fill in missing data, generating a time-synchronized multi-source dataset.

[0040] In this step, the system uses a dynamic time warping (DTW) algorithm to synchronize the time of the standardized dataset. Since data from different sources often have different formats and precision timestamps, the dynamic time warping technique can effectively eliminate these timestamp discrepancies, allowing records from different data sources to be accurately aligned in chronological order. At the same time, if some data is found to be missing during the time synchronization process, the system will use interpolation methods to estimate the reasonable value of the missing data based on the existing data, ensuring the integrity of the time-synchronized dataset.

[0041] Ensuring the time synchronization of data is crucial in medical applications, as it directly affects the monitoring and analysis of patient condition changes. By eliminating timestamp discrepancies, medical personnel can more accurately understand the evolution of the patient's condition when reviewing patient information. In addition, the interpolation method to fill in missing data improves the integrity of the data, allowing more information to be obtained in subsequent analysis, thereby improving the accuracy of decision-making.

[0042] In this step, the system uses a dynamic time warping (DTW) algorithm to synchronize the time of the standardized dataset. Time synchronization is a very important step in medical data management, especially in emergency and clinical monitoring, where accurate time order is crucial for data analysis. The DTW algorithm can effectively deal with the problem of inconsistent timestamps in different data sources, automatically adjusting the time series so that data from different sources can be aligned in chronological order. For example, an electrocardiogram may record once a minute, while a vital sign monitoring device may record once every ten seconds. The DTW algorithm helps to unify these timestamps into a stable timeline.

[0043] If some data is found to be missing during the time synchronization process, the system will use interpolation methods to estimate the reasonable value of the missing data based on the existing data. For example, if the heart rate data is missing during a certain period, the system can perform linear interpolation based on the heart rate trend before and after the missing data to estimate the missing heart rate value, ensuring the integrity of the dataset. This interpolation method is not only simple and effective, but also plays an important role in maintaining data consistency, especially in emergency situations, where ensuring the integrity of relevant medical information is crucial.

[0044] Finally, after a series of processing, the system generates a time-synchronized multi-source data set. This data set can support subsequent analysis and decision-making, helping doctors make decisions with clear and consistent information in a rapidly changing clinical environment, ensuring that chest pain patients can receive timely and effective medical intervention.

[0045] S202, for the time-synchronized multi-source data set, an abnormality detection algorithm based on deep learning is used to identify missing, erroneous or incomplete parts of the data, and a data repair model based on a generative adversarial network is used to repair and complete low-quality data, generating a patient data profile; In this process, first, the time-synchronized multi-source data set is analyzed using an abnormality detection algorithm based on deep learning, which identifies missing, erroneous or incomplete parts of the data. This algorithm usually performs abnormality detection based on reconstruction error, that is, the model learns the characteristics of normal data and establishes a standard for their representation, and then detects points that deviate from this standard during the testing phase and marks them as abnormal. In addition, the system adjusts the dynamic threshold at any time to improve the sensitivity to data anomalies, ensuring that low-quality data can be effectively identified.

[0046] This step has far-reaching significance in the entire patient data management system. By using the abnormality detection capabilities of deep learning, the system can ensure the quality and integrity of the data, which is the basis and guarantee for medical decision-making. In particular, when dealing with emergency information for chest pain patients, accurate patient data can directly affect the treatment effect. In addition, the data repair model based on a generative adversarial network can effectively improve the accuracy of data repair, ensuring that the generated patient data profile accurately reflects the patient's health status and enhances the doctor's confidence in the reliability of the data when making clinical decisions.

[0047] Specifically, for the time-synchronized multi-source data set, an abnormality detection algorithm based on deep autoencoders can be used to identify missing, erroneous or incomplete parts of the data through reconstruction error, and a dynamic threshold adjustment mechanism can be used to generate a data quality assessment report and mark low-quality data points. In this step, the system utilizes a deep autoencoder (DAE) based anomaly detection algorithm to analyze the synchronized multi-source dataset. The deep autoencoder model learns from the input data, attempts to reconstruct the original input, and calculates the reconstruction error. The size of the reconstruction error can be used to assess the degree of abnormality of a data point. If the reconstruction error of a certain data point exceeds the pre-set threshold, it will be marked as low-quality data. In this process, the introduction of a dynamic threshold adjustment mechanism can adjust the sensitivity of anomaly detection in real time according to the characteristics of different time periods or data sources. For example, in some cases, there may be abnormal equipment measurements or user input errors. Through real-time learning for dynamic adjustment, the system can more flexibly identify abnormal data points.

[0048] This step ensures the accuracy and reliability of the data by effectively identifying and marking low-quality data. In medical applications, low-quality data can lead to incorrect diagnoses or unnecessary treatments. Therefore, through timely anomaly detection and processing, the effectiveness and safety of the patient management system can be significantly improved. In addition, the generated data quality assessment report provides a solid foundation for subsequent data repair and decision-making, helping medical personnel quickly locate problems.

[0049] In this step, the system first needs to preprocess the time-synchronized multi-source dataset to ensure the standardization of input data. Data preprocessing may include denoising, normalization, and missing value filling operations. Then, a deep autoencoder (DAE) based anomaly detection algorithm is used for analysis. The structure of the autoencoder usually includes an input layer, a hidden layer, and an output layer. The input layer accepts the original data, the hidden layer learns the potential representation of the data, and the output layer reconstructs the input data. The system calculates the reconstruction error of each data point. The larger the reconstruction error, the higher the degree of deviation from the normal pattern.

[0050] For example, suppose in the monitoring data of a patient with chest pain, the heart rate should be within a reasonable range, but a certain data point shows a heart rate of 250 beats per minute, which is obviously unreasonable. After processing by the autoencoder, the system will calculate the reconstruction error of this data point. If it exceeds the set dynamic threshold, the data point will be marked as low-quality data. In addition, the dynamic threshold adjustment mechanism can optimize based on real-time data distribution information to improve the sensitivity and accuracy of detection.

[0051] Finally, the system generates a data quality assessment report detailing all low-quality data points, including the reasons for marking and reconstruction error values. This report will provide clear direction for subsequent data repair, allowing medical subjects to quickly identify and handle data quality issues.

[0052] For the marked low-quality data points, a data repair model based on a generative adversarial network is trained, where the generator is used to generate reasonable completion values for missing data, and the discriminator is used to evaluate the authenticity of the generated data, and through conditional constraint mechanisms, ensure that the generated data conforms to medical logic and contextual consistency. In this step, the system trains a generative adversarial network (GAN) model for the data points marked as low-quality in the first step. The generative adversarial network consists of two main parts: the generator and the discriminator. The role of the generator is to generate reasonable completion values for missing or low-quality data, while the discriminator is responsible for evaluating the authenticity of these generated data. In order to ensure that the generated data is not only reasonable but also conforms to medical logic, the system will introduce conditional constraint mechanisms, that is, consider the medical context information such as the age, gender, medical history and other related information of the patient during the generation process, so that the generated data is consistent with the true data in terms of features.

[0053] This process ensures the scientificity and rationality of data completion. Since the accuracy of data is crucial in the medical field, the generated data not only needs to conform to statistical characteristics, but also needs to be consistent with the specific situation of the patient. In this way, the system can effectively improve the quality of data completion, and thus improve the completeness and reliability of patient records, so that subsequent medical decisions can be based on more comprehensive data.

[0054] In this step, the system will create and train a generative adversarial network (GAN) model based on the marked low-quality data points. The generative adversarial network consists of two main components: the generator and the discriminator. The task of the generator is to receive low-quality data and its related context information (such as the basic information and medical history of the patient), and then generate reasonable completion values. The task of the discriminator is to evaluate the generated data and judge its authenticity and reasonableness.

[0055] For example, for a patient with heart disease, the generator may generate a missing heart rate data based on the patient's age, gender and clinical history. In order to ensure that the generated data conforms to medical logic, the system will introduce conditional constraint mechanisms, for example, the generated heart rate value must be within the normal heart rate range (60-100 times / minute) and consistent with the patient's condition. The generator will consider these conditional constraints while generating data, making the generated data more reliable.

[0056] During the entire training process, the generator and the discriminator will continuously optimize through adversarial learning. Each time the generator generates new data, the discriminator will evaluate the quality of the generated data based on the characteristics of the true data and give feedback. Through continuous iteration, the generator can learn how to generate more accurate and reasonable data completion values, while the discriminator improves its ability to identify generated data. Ultimately, this training process based on generative adversarial networks will effectively improve the quality of data repair.

[0057] The low-quality data points are input into the trained data repair model to generate completed values for missing or incorrect data. Through multiple rounds of iterative optimization, the accuracy and reliability of the repaired data are gradually improved, generating a repaired data set. The key to this step is to apply the trained generative adversarial network model to the labeled low-quality data points. The trained generator is used to actually complete these low-quality data, generating reasonable missing data under contextual conditions. Multiple rounds of iterative optimization are the core of data repair. The system will run the generative adversarial network multiple times, gradually adjusting the parameters of the generator to improve the accuracy and reliability of the generated data. After each generation, the completed data is evaluated by the discriminator to determine its authenticity and feedback to the generator, continuously improving the quality of the generated data.

[0058] Through multiple rounds of iterative optimization, the quality and credibility of the repaired data set can be continuously improved. This process not only improves the completeness of patient data files, but also ensures that doctors can rely on more accurate and authentic data to develop treatment plans in medical decision-making. This has a positive impact on improving the overall quality of medical services and reducing medical risks.

[0059] In this step, the system inputs the labeled low-quality data points into the trained generative adversarial network model. First, the generator will use these low-quality data as input, combining other relevant information about the patient, such as disease type, symptom description, etc., to generate possible completed values. During this process, the generator will ensure that the generated data meets medical logic based on the features and constraints learned during previous training. For example, if there is missing ECG data, the generator will generate a reasonable ECG completion value based on the patient's heart rate trend, past medical history, etc.

[0060] After generating the completed values, the system will input these generated data into the discriminator again, which will evaluate the authenticity and reasonableness of the generated data. If the generated data does not meet the expected quality standards, the system can automatically adjust the parameters of the generator and continue the new generation process. Through multiple rounds of iterative optimization, the generator gradually adjusts its generation strategy, continuously improving the accuracy and reliability of the generated data.

[0061] Finally, after several rounds of iteration, the system will form a repaired data set, which will be considered as high-quality data. This data set will not only be used for patient file updates, but also serve as an important basis for subsequent analysis and decision support. This process ensures that the data of patients with chest pain is effectively supplemented while maintaining its scientificity and practicality.

[0062] For the repaired dataset, a verification method based on medical expert rules is used, combined with historical case data and clinical guidelines, to verify the rationality of the repaired data. Through a data quality scoring mechanism, a high-quality patient data archive is generated.

[0063] In this step, the system verifies the repaired dataset to ensure that the generated data meets medical logic and practical application requirements. Through a verification method based on medical expert rules, the system compares the repaired data with historical case data and clinical guidelines to check the rationality and consistency of the data. For example, the system can verify whether the repaired blood pressure value of a patient is within the normal range and consistent with the historical data of patients with the same disease. If the repaired value deviates significantly from the medical standard, the system will mark these data as abnormal.

[0064] This process ensures the scientificity and medical applicability of the repaired data, minimizing misleading information caused by data repair. By combining clinical guidelines and historical case data, the system can improve the rigor of the data quality scoring mechanism and generate high-quality patient data archives, providing reliable basis for medical decision-making, thereby improving the safety and effectiveness of medical services.

[0065] In this step, first, the system will define a series of verification rules based on medical expert experience and clinical guidelines. These rules usually include the setting of normal ranges for key indicators in patient data and the logical relationship that meets these standards. For example, the normal range of heart rate, blood pressure, body temperature, and other vital signs, as well as the changes in indicators under different disease conditions.

[0066] The system compares the repaired dataset with these verification rules to check the rationality of the data. Suppose the repaired data of a heart disease patient shows a blood pressure value of 180 / 120 mmHg, the system will automatically compare it with the set rules and find that the value is significantly higher than the normal range, and mark it as abnormal. This combination of rules is not limited to the comparison of a single indicator, but can also be combined with other health records of the patient for comprehensive analysis to ensure the consistency and rationality of the data.

[0067] In addition, the system will also use historical case data for cross-validation. If the repaired data significantly differs from the historical data of similar patients, the system will mark it as an abnormal situation that needs further confirmation. The data quality scoring mechanism will assign a score to each repaired data based on the above verification process, forming a detailed data quality report. The verified dataset will be considered as a high-quality patient data archive, ensuring a true reflection of the patient's health status and providing a solid basis for subsequent medical decision-making.

[0068] S203, for the patient data file, a blockchain-based data update mechanism is adopted to record the change history of the data in real time, and through a distributed data synchronization algorithm, the doctor can obtain the latest patient information and generate a real-time synchronized patient data file; In this step, the system introduces blockchain technology as the update mechanism for the patient data file. Whenever patient data changes (such as adding medical records, updating test results, or modifying treatment plans), these changes will be recorded as transactions on the blockchain. The immutability of the blockchain ensures data integrity, and any data changes at any point in time can be traced back, providing transparency and trust for the medical process. At the same time, the system will use smart contracts to automatically verify the legality of data changes, ensuring that only authorized users can make changes to patient data. Through this mechanism, the patient's data architecture forms a dynamic, real-time updating system.

[0069] The use of blockchain-based data update mechanisms can significantly improve the security and reliability of patient data. In the medical industry, data accuracy and integrity are related to patient safety and treatment effectiveness. By using blockchain to ensure that every data change is effectively recorded, medical personnel can obtain the latest patient information in real time, avoiding misdiagnosis or missed diagnosis due to outdated information. In addition, transparent data change history records can also provide better rights protection for patients, improving their trust in medical services, and thus promoting the improvement of the doctor-patient relationship.

[0070] Specifically, for high-quality patient data files, a blockchain-based data update mechanism can be used to record each data change as a transaction on the blockchain, and through smart contracts, automatically verify the legality of data changes, and generate an immutable data change history. In this step, the patient's high-quality data file will be protected by blockchain technology. Each data change (such as updating laboratory results, modifying medical orders, etc.) will be recorded as a transaction and written to the blockchain. Smart contracts will be used to automatically verify the legality and compliance of these changes, ensuring that any changes comply with the hospital's standard operating procedures. During the transaction recording process, the immutability of the blockchain ensures data integrity, making historical records always traceable, and any modification to the data can be clearly audited.

[0071] This mechanism greatly enhances the transparency and security of medical data. Blockchain ensures the authenticity and immutability of data change records, reducing the risk of counterfeit data due to human error or malicious attacks. At the same time, automated smart contracts can reduce the reliance on manual audits in medical management processes, improving efficiency and reducing delays. This mechanism with transparent traceability not only improves the trust relationship between doctors and patients, but also provides legal compliance for medical data.

[0072] In implementing a blockchain-based data update mechanism, first, a suitable blockchain platform for the medical industry needs to be selected, such as Hyperledger Fabric or Ethereum, which can meet the privacy and security requirements of medical data. Next, work with the hospital IT department to build a medical data management system integrated with blockchain. In this system, whenever a patient's medical data changes (e.g., uploading new test results or adjusting treatment plans), the system will automatically generate a corresponding blockchain transaction, ensuring the authenticity and time reliability of data changes through encryption and timestamping of transaction content. This process involves structured processing of data to ensure that transaction data can be written to the blockchain in a standard format.

[0073] Secondly, the design of smart contracts is crucial, as they act as automatically executed programs to ensure the legality of data changes. Strict rules are set in smart contracts, such as only authorized doctors can update specific types of data, ensuring that only compliant operations can generate blockchain transactions. Smart contracts not only verify transactions when they are generated, but also provide automated feedback during data review. Once irregular operations are detected, the system will immediately reject execution and record abnormal information for subsequent review.

[0074] Finally, hospitals can use blockchain browser tools to visualize the review of data change history. Through these tools, medical staff can easily view detailed information of each transaction, including the reason for the change, the time, and the operator. This transparent history record can enhance the trust between patients and medical staff, and provide strong data support in medical disputes or legal reviews, helping hospitals maintain their legal rights and interests.

[0075] According to the hospital's internal network environment and external internet connection, build a data synchronization framework based on distributed ledger technology, ensure data consistency between multiple nodes through lightweight consensus algorithm, and generate a preliminary distributed synchronization network; The focus of this step is to build a distributed ledger technology framework to achieve real-time synchronization of data within the hospital and when connected to the outside. Unlike traditional centralized data storage, distributed ledgers allow multiple nodes to jointly maintain data consistency and integrity. According to the hospital's network conditions (such as the stability of the local area network and the internet), design a reasonable node architecture and use lightweight consensus algorithms to quickly reach an agreement to ensure that the data on each node is the same, thereby avoiding data islands and update lag phenomena.

[0076] By constructing a distributed ledger technology synchronization framework, hospitals can achieve real-time data updating and sharing, providing doctors with the latest patient information and supporting the timeliness and accuracy of clinical decision-making. This architecture decentralizes data storage, reducing the risk of single-point failure and improving system robustness. Meanwhile, the application of lightweight consensus algorithms reduces resource consumption and speeds up data processing, ensuring the efficiency of information flow between multiple departments and roles.

[0077] In building a distributed ledger technology data synchronization framework, first, the hospital's network environment needs to be evaluated in detail to determine the internal network's bandwidth, latency, and connection capabilities with external networks. This step involves collaboration with network architects who will plan the locations of each medical node based on the hospital's geographical distribution, such as emergency rooms, wards, and laboratories, ensuring that each node can connect to the main blockchain network through a secure virtual private network (VPN).

[0078] Next, the hospital needs to choose an appropriate lightweight consensus algorithm, such as Practical Byzantine Fault Tolerance (PBFT) or Raft. This choice will affect the network's performance and data consistency. Through lightweight consensus algorithms, hospitals can ensure that the system can quickly reach consensus when data changes occur, thus achieving rapid data synchronization. For example, when a laboratory updates a patient's test results, the data will be sent to all relevant nodes, and after consensus algorithm processing, it will quickly reach agreement among multiple nodes, while avoiding the problem of data desynchronization caused by network latency.

[0079] After the framework is established, the system needs to be monitored and optimized. Through real-time monitoring tools, hospitals can continuously track the performance of the network and the status of each node, ensuring the high availability and stability of the system. In actual operation, if a node connection is abnormal or fails, the system can automatically isolate it from the network, ensuring the healthy operation of the overall network. This dynamic management can minimize the impact on the hospital's daily clinical activities, ensuring the reliability of the system and the timely updating of data.

[0080] For nodes in the distributed synchronization network, a real-time data synchronization algorithm based on timestamps is used to ensure that doctors' terminals can access the latest patient information. Through conflict detection and resolution mechanisms, dynamic handling of data conflicts between multiple nodes is achieved, generating conflict-free synchronized data. In this step, the focus is on ensuring that nodes in the distributed synchronization network can access the latest patient information in real-time. A timestamp-based real-time data synchronization algorithm is employed, where any data update is accompanied by a timestamp to determine its recency. Additionally, to address potential data conflicts among multiple nodes, such as simultaneous updates to the same data, a conflict detection and resolution mechanism is implemented to ensure consistent and conflict-free data generation, thereby guaranteeing accurate and reliable information for doctors' terminals.

[0081] Through timestamp synchronization and conflict resolution mechanisms, hospitals can effectively address data consistency issues, ensuring that medical decisions are based on the latest information and improving clinical outcomes. This mechanism not only prevents medical accidents caused by data errors but also enhances doctors' confidence in their decision-making basis. Meanwhile, this process improves the response speed of the medical information system, enabling faster transmission of new patient information to clinical decision-makers and reducing the risk of medical risks due to information delays.

[0082] When implementing a timestamp-based real-time data synchronization algorithm in a distributed synchronization network, an efficient data update notification mechanism needs to be established in each node. When data within a node changes, the system automatically generates a timestamp and attaches it to the data item. Next, through message queue technology (such as Kafka), data change events and timestamps are pushed to other related nodes, so that all nodes can obtain the latest data update information in a timely manner. When doctors access patient information, the system will display the latest data based on the timestamp, ensuring that their decisions are based on the current true situation.

[0083] Then, to prevent data conflicts among multiple nodes, hospitals need to design a conflict detection and resolution mechanism. For example, rules are set so that when multiple nodes attempt to update the same patient's medical record simultaneously, the system will compare the data versions based on timestamps. The system will preferentially retain the latest update and notify other nodes for synchronization. If there is a conflict in data versions, the system will automatically record the conflict and use predefined rules to resolve it, such as prioritizing the doctor's permission level or the timestamp of the update. At the same time, the system should provide a manual intervention interface for medical staff to review and make decisions in special cases.

[0084] Finally, to ensure the integrity of real-time synchronized data, hospitals can set up a regular full data checking mechanism to ensure the consistency of all node data by comparing it with the main ledger data. If the system finds that the data in a node has not been updated in time, it will trigger a resynchronization process to ensure that it is consistent with the main ledger data. This efficient data synchronization and conflict resolution mechanism helps maintain doctors' trust in the system, enabling them to accurately use the latest patient information in clinical practice.

[0085] For synchronized data, a version control-based data integration method is used to integrate the latest data with historical change records. Through dynamic permission control mechanisms, only authorized personnel can access sensitive information, generating real-time synchronized patient data archives.

[0086] In this final step, the focus is on integrating and managing synchronized data. Using version control methods, the latest data and historical change records are reasonably combined to form a complete patient data archive. Through dynamic permission control mechanisms, only authorized medical personnel can access sensitive information, which not only guarantees patient privacy but also enhances data management security.

[0087] Using version control methods, all patient data can be effectively traced, even during data updates, and historical records can be preserved for doctors to review patient medical history and previous treatment processes at any time. At the same time, dynamic permission control mechanisms effectively protect sensitive information from unauthorized access and tampering. This integration and management process enhances the integrity and security of patient data archives and enhances the overall data governance capabilities of the hospital.

[0088] When performing version control and integration on synchronized data, first, implement version control functions in the data management system. Each time the data changes, the system not only updates existing data but also records new data and its version information, forming historical change records. For example, when a doctor modifies a patient's prescription, not only the latest prescription information is stored, but also the old version of the prescription and the reason for the change. In the future, when reviewing patient medical records, medical personnel can easily access historical versions to understand patient medical history and treatment processes, providing comprehensive information support for clinical decision-making.

[0089] Then, the hospital needs to design a dynamic permission control mechanism to protect sensitive information. Through a role management system, the hospital can set different access permissions for different roles (such as doctors, nurses, administrative personnel, etc.). For example, only doctors from relevant departments and authorized administrators can access complete patient medical records, while ordinary nurses can only view basic information that needs to be handled. The system monitors permission settings in real time and verifies each data access request to ensure that unauthorized users cannot view sensitive data. This dynamic control ensures the protection of patient privacy and enhances data management security.

[0090] Finally, to generate a real-time synchronized patient data profile, the hospital can utilize a data integration tool to combine the latest data with historical versions, forming a complete patient profile. This profile will include the patient's basic information, historical visit records, test results, prescription records, etc., with each piece of data marked with its version and update time. This information will be automatically generated and stored in the hospital's electronic medical record system, allowing doctors to quickly retrieve it when needed for diagnosis and treatment decisions. Through this comprehensive management, the hospital not only improves the availability and completeness of data, but also builds a safe and efficient medical data management system.

[0091] S204, for the real-time synchronized patient data profile, an artificial intelligence analysis model based on multi-modal fusion is used to generate diagnosis suggestions and treatment plans by combining the clinical characteristics of chest pain patients and historical case data. Through the explainability module of the artificial intelligence analysis model, the basis and confidence of the diagnosis result are provided, and the final intelligent diagnosis support report is generated.

[0092] In this process, an artificial intelligence analysis model based on multi-modal fusion is used to comprehensively integrate information from different data sources (such as electronic medical records, image data, and test reports) to construct a comprehensive feature representation. These features include not only the clinical characteristics of patients (such as age, gender, medical history, etc.), but also the visual features of imaging examination results (such as chest X-ray or CT scan images) and the numerical features of test results (such as electrocardiogram data). By introducing a cross-modal attention mechanism, the model can effectively identify important associations between different features, and then generate a unified multi-modal feature representation. This process not only improves the relevance and completeness of the data, but also provides a rich information base for subsequent diagnosis and treatment plan generation.

[0093] By using a multi-modal fusion artificial intelligence analysis model, doctors can obtain more comprehensive and accurate diagnosis support, and the explainability module of the model can provide transparent decision-making basis to enhance the doctor's trust in the model output results. In addition, this method can significantly improve the accuracy and individualization of diagnosis, ensuring that the treatment plan is more in line with the specific situation of the patient. For example, by analyzing historical case data, the model can identify that certain patient groups have a more positive response to a particular treatment plan, thereby providing more targeted treatment recommendations for doctors. This not only improves the quality of clinical decision-making, but also improves the treatment effect and satisfaction of patients.

[0094] Specifically, for the real-time synchronized patient data profile, a multi-modal feature extraction network based on multi-head attention mechanism can be used to extract text features from electronic medical records, visual features from image data, and numerical features from test reports. Through a cross-modal attention mechanism, the features of different modalities are weighted and fused to generate a multi-modal fusion feature representation. In this step, a multi-head attention mechanism is introduced into the multi-modal feature extraction network to extract features from different types of data. First, the text features in the electronic medical record are analyzed using natural language processing techniques, such as using word embedding models to convert text into vector representations. For image data, the model uses a convolutional neural network to extract features from the image and identify potential lesions in the chest image. The numerical features of the test report are extracted through simple numerical encoding or normalization processing. After these feature extractions are completed, a cross-modal attention mechanism is used to weight and fuse the features of each modality to ensure that the final generated feature representation can comprehensively reflect the important information of different data sources.

[0095] Through multi-modal feature extraction and weighted fusion, the model's overall understanding of complex conditions such as chest pain can be significantly improved. Different modalities of data complement each other, making the model more accurate and reliable when making decisions. In addition, this process helps identify potential lesions or abnormalities and assess the patient's health status more comprehensively based on multi-angle information. Ultimately, doctors can rely on more accurate and detailed diagnostic information when developing clinical strategies, thereby providing more effective individualized treatment plans.

[0096] In this step, a multi-modal feature extraction network is first established, which can process data from different sources, including electronic medical records (EMR), image data (such as X-ray or CT images), and laboratory test reports. For text features in electronic medical records, natural language processing techniques such as BERT or Word2Vec are used to convert text data into vector representations. This process mainly includes denoising, tokenization, and word embedding of medical record text to extract valuable information. For example, patient complaints, medical history, and treatment records can be extracted through pre-trained language models to achieve feature vectorization.

[0097] Next, when processing image data, a convolutional neural network (CNN) such as ResNet or EfficientNet is used for feature extraction. The design of this network can effectively capture key visual information in the image and convert it into a feature vector. For example, for a chest X-ray of a patient with chest pain, the network can automatically identify potential lung lesions, heart size and shape, and other important features. At the same time, the numerical features of the test report are simply normalized to ensure that these features can be reasonably compared and analyzed when integrated.

[0098] Finally, all the extracted features will be input into a multi-head attention mechanism. This mechanism can weight and fuse features from different modalities to identify the relevance between different features. For example, a doctor might find that a specific clinical symptom (extracted from electronic medical records) is closely related to an imaging finding (such as cardiac enlargement). By weighting and fusing these features through the cross-modal attention mechanism, a multi-modal fusion feature representation can be generated, which comprehensively reflects the health status of the chest pain patient. This representation will provide rich information support for the subsequent diagnosis model.

[0099] For the multi-modal fusion feature representation, a deep neural network-based diagnosis model is used to generate preliminary diagnosis suggestions by combining the clinical features of the chest pain patient and historical case data. Through the explainability module of the diagnosis model, the basis and confidence of the diagnosis results are provided. In this step, the constructed multi-modal fusion features will be input into a deep neural network-based diagnosis model, which will analyze the clinical features of the chest pain patient and historical case data. The multi-layer structure of the deep neural network enables it to effectively learn complex feature relationships and generate preliminary diagnosis suggestions. The explainability module provides the necessary decision basis by analyzing the weight distribution and feature influence within the model, clearly defining the contribution of each input feature in the final diagnosis. This transparent analysis process not only improves the credibility of the diagnosis, but also enhances the understanding and trust of the model by doctors.

[0100] The implementation of this step significantly improves the intelligent level of diagnosis, helping doctors quickly obtain scientific and systematic diagnosis suggestions. This process helps reduce human error, especially in the handling of complex cases, supporting doctors to make more stable clinical decisions. In addition, through the support of the explainability module, doctors can intuitively understand the decision-making process of the model, thereby improving the efficiency of communication with patients and enhancing the confidence of patients in the treatment plan.

[0101] In this step, a diagnosis model is constructed based on deep neural networks, taking multi-modal fusion features as input. The model structure usually uses fully connected layers and activation functions (such as ReLU) to form non-linear characteristics, and may also add convolutional layers to further extract spatial features. The hierarchical design of this network structure helps the model capture complex hidden patterns. For example, when the input contains data about patient symptoms, imaging examinations, and test results, the model can learn the complex relationship between symptoms and underlying pathological states.

[0102] During the training process, historical case data is used as labeled data to train the model in a supervised learning manner. By fitting these data, the model learns how to map the input features to the corresponding diagnosis results. During the training process, methods such as cross-validation are used to evaluate the accuracy and robustness of the model. For example, when inputting the specific data of a chest pain patient, the model may output a preliminary diagnosis suggestion of "acute myocardial infarction", and its accuracy depends on the learning and training effect of the model on historical data.

[0103] To ensure the credibility of the diagnosis results, an explainability module is introduced into the model. This module uses techniques such as LIME (Local Interpretable Model-agnostic Explanations) or SHAP (SHapley Additive exPlanations) to analyze the specific impact of each input feature on the final diagnosis result. For example, through SHAP values, doctors can see which features (such as ECG changes, slight chest imaging abnormalities, etc.) played a key role in making a certain diagnosis suggestion. This transparency not only helps medical staff verify the rationality of the model output results, but also enhances the trust in the model's decision-making, providing strong support for subsequent clinical decision-making.

[0104] For the preliminary diagnosis suggestion, a reinforcement learning-based treatment plan optimization algorithm is used to dynamically adjust the treatment plan based on the individual characteristics of the patient and the current situation of hospital resources, and through a multi-objective optimization model, to balance the treatment effect and the risk of side effects, to generate an optimized diagnosis suggestion and treatment plan; In this stage, the preliminary diagnosis suggestion is input into the reinforcement learning-based treatment plan optimization algorithm. This algorithm simulates the effects of different treatment plans and compares their performances in terms of treatment effect, patient satisfaction, and side effect risk, etc., to achieve dynamic adjustment and optimization. Combined with the individual characteristics of the patient (such as medical history, drug allergy, etc.) and the current situation of hospital resources (such as drug availability, doctors' professional fields, etc.), the system will be able to generate the most suitable treatment plan.

[0105] The introduction of reinforcement learning makes the adjustment of treatment plans more intelligent and personalized. Through learning from historical treatment data, the model can identify which treatment plans perform well in specific patient groups, and then provide scientific optimization suggestions for doctors. This dynamic adjustment capability ensures that patients can obtain effective treatment while minimizing potential side effects, improving the overall safety and effectiveness of treatment. This provides a solid foundation for individualized treatment in clinical practice, and improves patient satisfaction and treatment effect.

[0106] In this step, a reinforcement learning framework needs to be constructed, starting with defining multiple dimensions of treatment plans, including drug selection, dosage, treatment duration, and treatment approach (such as conservative treatment or surgical intervention). This framework will integrate historical case data and individualized characteristics of the current patient to obtain the performance of multiple treatment options under different conditions. For example, in past cases, a certain drug performed well in patients of a certain age group, but had no obvious effect on patients of another age group. This information will be integrated into the model through training data.

[0107] Next, reinforcement learning algorithms such as Q-learning or deep Q-networks are used for model training, based on reward functions to evaluate the effectiveness of different treatment plans. The reward function will integrate treatment effectiveness, patient feedback, side effects during treatment, and other factors, encouraging the algorithm to choose those plans that reduce side effects while improving treatment effectiveness. For example, if a patient experiences severe side effects during chemotherapy, the model will record this feedback and adjust its score for that plan accordingly, giving it a lower priority in future recommendations, guiding doctors to choose other safer treatment options.

[0108] During the actual treatment of the patient, the system will analyze the patient's clinical data and the hospital's resource situation (such as whether the drug is sufficient, the doctor's expertise, etc.) in real time, providing personalized treatment recommendations. For example, for a young patient with chest pain, if the system finds that they have an allergic reaction to a certain drug, it will integrate previous learning results and preferentially recommend safer alternative options for the patient. This dynamic adjustment mechanism ensures continuous learning and optimization of the model, ultimately generating optimized diagnosis recommendations and treatment plans, enabling doctors to make more scientific and reasonable decisions.

[0109] For the optimized diagnosis recommendations and treatment plans, a report generation method based on natural language generation technology is used to convert the analysis results into easy-to-understand text descriptions, and through visualization technology, the final intelligent diagnosis support report is generated, including diagnosis basis, treatment plan, and confidence.

[0110] In this step, based on natural language generation technology, the optimized diagnosis recommendations and treatment plans are converted into easy-to-understand text reports. The system expresses the analysis results in simple language, through a systematic information structure (such as introduction, analysis results, diagnosis basis, treatment plan, confidence, etc.), ensuring that both doctors and patients can easily understand the information provided. In addition, the report will combine visualization technology to display important data and analysis results in the form of charts, further enhancing the readability and professionalism of the report.

[0111] This form of report generation can significantly improve the efficiency of communication between doctors and patients, helping patients better understand their health status and treatment options. At the same time, easy-to-understand reports reduce the burden on doctors when explaining analysis results, allowing them to focus more on clinical decision-making and further communication with patients. On the basis of information transparency, patient participation and satisfaction will be improved, enhancing patients' trust in the treatment process.

[0112] In this step, a report generation system based on natural language generation (NLG) technology is implemented. First, the system will establish a template library to define the basic structure of different types of diagnostic support reports, usually including the introduction, analysis results, diagnostic basis, treatment options, expected effects and risk assessment, etc. For example, the introduction part of the report can briefly introduce the patient's basic information and the reason for the visit, and the analysis results part can describe the accurate diagnosis obtained through the model in detail.

[0113] Subsequently, the system will use data-driven methods to convert analysis results into natural language text. This process includes parsing structured data and dynamically filling generated text templates. For example, if the model recommends a new treatment option, the system will automatically generate a text: "Based on the patient's clinical characteristics and historical case data, a low-dose aspirin is recommended as the initial treatment option, which has shown good efficacy in similar cases." Such text output allows doctors to quickly understand the analysis results and recommendations.

[0114] In addition, by integrating visualization technology, the system can also display data in the form of charts, making the information in the report more vivid. For example, the success rate and side effect risk of different treatment options can be expressed in a graphical way, allowing patients to better understand their health status and treatment options when reading the report. The final intelligent diagnostic support report will be pushed to patients and doctors through email or mobile application, ensuring timely delivery and effective communication of information. This process not only improves the professionalism and readability of the report, but also enhances patient participation, making them more confident and understanding during the treatment process.

[0115] It can be seen that, according to the electronic medical record, image data and test report of the chest pain patient, multi-source heterogeneous data is acquired in real time, data of different sources is uniformly converted into a standardized data format, and a time-synchronized multi-source data set is generated; for the multi-source data set, missing, error or incomplete parts in the data are identified, a data repair model of a generative adversarial network is generated, and a patient data archive is generated; for the patient data archive, the change history of the data is recorded in real time, a distributed data synchronization algorithm is used to generate a real-time synchronized patient data archive; for the real-time synchronized patient data archive, a diagnosis suggestion and a treatment plan are generated, a basis and a confidence of a diagnosis result are provided, and a final intelligent diagnosis support report is generated, so that the efficiency and accuracy of data management can be improved, and the diagnosis and treatment effect of the chest pain patient can be improved.

[0116] Another embodiment of the present application provides an artificial intelligence-based intelligent management system for medical data of a chest pain patient, referring to Figure 3 The system can be set to execute the steps in any of the above method embodiments when running.

[0117] The present application also provides a storage medium having a computer program stored therein, wherein the computer program is set to execute the steps in any of the above method embodiments when running.

[0118] The present application also provides an electronic device comprising a memory and a processor, wherein the memory has a computer program stored therein, and the processor is set to run the computer program to execute the steps in any of the above method embodiments.

[0119] Specifically, the above electronic device can further comprise a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0120] The above embodiments according to the drawings illustrate the structure, features and effects of the present application, and the above description is only the preferred embodiment of the present application, but the present application is not limited to the embodiments shown in the drawings, any changes or modifications made in accordance with the concept of the present application, or equivalent embodiments with equivalent changes, shall be within the scope of the present application.

Claims

1. An artificial intelligence-based intelligent management method for medical data of chest pain patients, characterized in that: The method comprises: Based on the electronic medical records, imaging data, and laboratory reports of chest pain patients, an edge computing-based data acquisition framework is used to acquire multi-source heterogeneous data in real time. An adaptive data format conversion algorithm is used to convert data from different sources into a standardized data format, generating a time-synchronized multi-source data set. For time-synchronized multi-source data sets, we use a deep learning-based anomaly detection algorithm to identify missing, erroneous, or incomplete data. We then use a generative adversarial network-based data repair model to repair and complete low-quality data and generate patient data archives. For patient data archives, a blockchain-based data update mechanism is used to record data change history in real time. Through a distributed data synchronization algorithm, it ensures that doctors obtain the latest patient information and generates real-time synchronized patient data archives. For real-time synchronized patient data archives, an artificial intelligence analysis model based on multimodal fusion is used, combined with the clinical characteristics and historical case data of chest pain patients to generate diagnostic recommendations and treatment plans. Through the interpretability module of the artificial intelligence analysis model, the basis and confidence of the diagnostic results are provided, and the final intelligent diagnosis support report is generated.

2. The method according to claim 1, characterized in that Based on the electronic medical records, imaging data, and test reports of chest pain patients, the system uses an edge computing-based data acquisition framework to acquire multi-source heterogeneous data in real time. Through an adaptive data format conversion algorithm, the data from different sources is uniformly converted into a standardized data format to generate a time-synchronized multi-source data set, including: Based on the electronic medical records, imaging data, and laboratory reports of chest pain patients, an edge computing-based data collection framework is used to acquire multi-source heterogeneous data in real time through distributed edge nodes. Each edge node is equipped with a lightweight data caching mechanism to ensure real-time and continuous data collection. For the collected multi-source heterogeneous data, a deep learning-based format recognition model is used to automatically identify the format types of different data sources. Through an adaptive parsing algorithm, data in different formats are parsed into structured intermediate representations to generate a preliminary parsed data set. For the initially parsed dataset, a standardized conversion algorithm based on a combination of a rule engine and machine learning is used to uniformly convert data from different sources into a standardized data format. Through a context-aware mapping rule library, the conversion logic is dynamically adjusted to generate a standardized dataset. For the standardized data set, a time synchronization algorithm based on dynamic time warping is used to eliminate the timestamp differences between different data sources. The missing data is supplemented by interpolation filling method to generate a time-synchronized multi-source data set.

3. The method according to claim 2, characterized in that The aforementioned deep learning-based anomaly detection algorithm for time-synchronized multi-source datasets identifies missing, erroneous, or incomplete data. The data repair model generated by the Generative Adversarial Network is then used to repair and complete low-quality data, generating patient data archives. This includes: For time-synchronized multi-source datasets, an anomaly detection algorithm based on a deep autoencoder is used to identify missing, erroneous, or incomplete parts of the data through reconstruction errors. A dynamic threshold adjustment mechanism is used to generate a data quality assessment report and mark low-quality data points. For labeled low-quality data points, a data restoration model based on a generative adversarial network is trained. The generator is used to generate reasonable complement values ​​for missing data, and the discriminator is used to evaluate the authenticity of the generated data. A conditional constraint mechanism is used to ensure that the generated data conforms to medical logic and contextual consistency. Input low-quality data points into the trained data repair model to generate complementary values ​​for missing or erroneous data. Through multiple rounds of iterative optimization methods, the accuracy and reliability of the repaired data are gradually improved to generate a repaired dataset. For the repaired data set, a verification method based on medical expert rules is used, combined with historical case data and clinical guidelines, to verify the rationality of the repaired data, and a high-quality patient data archive is generated through a data quality scoring mechanism.

4. The method according to claim 3, characterized in that The patient data archive uses a blockchain-based data update mechanism to record data change history in real time. Through a distributed data synchronization algorithm, it ensures that doctors obtain the latest patient information and generates real-time synchronized patient data archives, including: For high-quality patient data archives, a blockchain-based data update mechanism is used to record each data change as a transaction on the blockchain. Through smart contracts, the legitimacy of the data change is automatically verified, generating an unalterable data change history. Based on the hospital's internal network environment and external Internet connectivity, a data synchronization framework based on distributed ledger technology was constructed. A lightweight consensus algorithm was used to ensure data consistency across multiple nodes, generating a preliminary distributed synchronization network. For nodes in the distributed synchronization network, a real-time data synchronization algorithm based on timestamps is used to ensure that the doctor terminal can obtain the latest patient information. Through the conflict detection and resolution mechanism, data conflicts between multiple nodes are dynamically handled to generate conflict-free synchronized data. For the synchronized data, a data integration method based on version control is adopted to integrate the latest data with historical change records. Through the dynamic permission control mechanism, it is ensured that only authorized personnel can access sensitive information, generating real-time synchronized patient data archives.

5. The method according to claim 4, characterized in that The AI ​​analysis model based on multimodal fusion is used for real-time synchronized patient data archives. It combines the clinical characteristics and historical case data of chest pain patients to generate diagnostic recommendations and treatment plans. The AI ​​analysis model's interpretability module provides the basis and confidence level for the diagnostic results and generates a final intelligent diagnostic support report, including: For real-time synchronized patient data archives, a multimodal feature extraction network based on a multi-head attention mechanism is used to extract text features from electronic medical records, visual features from imaging data, and numerical features from test reports. Through a cross-modal attention mechanism, features from different modalities are weightedly fused to generate a multimodal fusion feature representation. For multimodal fusion feature representation, a diagnostic model based on a deep neural network is used, combining the clinical characteristics of chest pain patients and historical case data to generate preliminary diagnostic recommendations. The interpretability module of the diagnostic model provides the basis and confidence level of the diagnostic results. For preliminary diagnostic recommendations, a reinforcement learning-based treatment plan optimization algorithm is used. This algorithm dynamically adjusts treatment plans based on individual patient characteristics and hospital resource availability. Through a multi-objective optimization model, it balances treatment efficacy with the risk of side effects to generate optimized diagnostic recommendations and treatment plans. For optimized diagnostic recommendations and treatment plans, a report generation method based on natural language generation technology is used to convert the analysis results into easy-to-understand text descriptions. Through visualization technology, a final intelligent diagnostic support report containing diagnostic basis, treatment plan and confidence level is generated.

6. An artificial intelligence-based intelligent management system for medical data of chest pain patients, characterized by: The system comprises: The acquisition module is used to acquire multi-source heterogeneous data in real time based on the electronic medical records, imaging data, and laboratory reports of chest pain patients using an edge computing-based data acquisition framework. Using an adaptive data format conversion algorithm, the data from different sources is uniformly converted into a standardized data format to generate a time-synchronized multi-source dataset. The recognition module uses a deep learning-based anomaly detection algorithm to identify missing, erroneous, or incomplete data from time-synchronized multi-source datasets. It then uses a data repair model generated by a generative adversarial network to repair and complete low-quality data and generate patient data archives. The generation module is used to update patient data archives using a blockchain-based data update mechanism to record data change history in real time. Through a distributed data synchronization algorithm, it ensures that doctors obtain the latest patient information and generates real-time synchronized patient data archives. The management module is used to generate diagnostic recommendations and treatment plans for real-time synchronized patient data archives using an artificial intelligence analysis model based on multimodal fusion, combined with the clinical characteristics and historical case data of chest pain patients. The interpretability module of the artificial intelligence analysis model provides the basis and confidence level of the diagnostic results and generates the final intelligent diagnosis support report.

7. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 5 when executed.

8. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 5.

Images