Digital hospital-oriented data fusion system

Abstract

The invention relates to the field of hospital data fusion systems, and discloses a digital hospital-oriented data fusion system, which comprises a data acquisition module, a data preprocessing module, a data analysis module, a data fusion engine module, a data storage module and a data access and application interface module, according to the system, the medical service quality is improved, the diagnosis time is shortened and the diagnosis accuracy is improved through deep fusion of data and intelligent pushing and other functions, medical staff can master the condition of a patient more quickly, comprehensively and accurately so as to provide higher-quality and higher-efficiency medical service for the patient, the adaptability and expansibility of the system are improved, and the medical service quality is improved. And by virtue of real-time, accurate and comprehensive data fusion and an intelligent data display function aiming at management personnel, a hospital manager can accurately insight into the hospital operation condition, timely adjust a management strategy and optimize resource configuration.

Description

Technical Field

[0001] The present invention relates to the technical field of hospital data fusion systems, and in particular to a data fusion system for digital hospitals. Background Art

[0002] With the continuous advancement of hospital informatization construction, various medical information systems such as hospital information system (HIS), laboratory information management system (LIS), medical picture archiving and communication system (PACS) have been widely used. In the current process of digital hospital construction, many medical information systems have been widely deployed and put into use.

[0003] However, some current medical information systems have many shortcomings when it comes to integrating and utilizing these multi-source heterogeneous data, such as: 1) Low level of data integration: Most hospitals have simply implemented partial data interaction between systems and lack a deep fusion mechanism; 2) Insufficient versatility and accuracy of fusion algorithms: Some existing data fusion solutions use entity recognition and matching algorithms that rely too much on fixed rules set by humans, and have poor processing capabilities for ambiguity and inconsistencies in the data; 3) It is difficult to ensure the timeliness and consistency of data updates: When the existing system updates data, there is a lack of effective linkage mechanisms between data sources, which easily leads to asynchronous updates of data in different systems; 4) Limited system scalability and compatibility: Large-scale architectural adjustments and redevelopment are required, which is costly and inefficient.

[0004] Based on this, a data fusion system for digital hospitals is proposed here to solve the above problems. Summary of the invention

[0005] In order to overcome the above technical problems, the purpose of the present invention is to provide a data fusion system for digital hospitals to improve the quality of medical services and enhance the adaptability and scalability of the system. The purpose of the present invention can be achieved through the following technical solutions:

[0006] A data fusion system for digital hospitals, comprising:

[0007] Data collection module: It has an intelligent adaptation mechanism that can automatically identify the type of data source and dynamically load the corresponding collection plug-in to obtain data;

[0008] Data preprocessing module: Integrates intelligent data verification and automatic repair functions. When cleaning data, in addition to conventional rule verification, it also uses machine learning models to learn normal and abnormal patterns in historical data, so that it can make more accurate judgments on suspected abnormal data. Moreover, when a repairable error is found, it can automatically repair it or prompt the operator to confirm the repair.

[0009] Data analysis module: provides a variety of data analysis tools and algorithms, supports in-depth mining and analysis of integrated data, and discovers patterns and trends in the data;

[0010] Data fusion engine module, including:

[0011] Entity recognition and matching: Adopt an adaptive entity recognition and matching algorithm to perform preliminary screening based on the initial general rules, then use the graph neural network (GNN) technology in deep learning to construct the data entities in the data source into a graph structure, dynamically capture the complex relationship between entities through node and edge feature learning, and continuously optimize the matching results;

[0012] Semantic fusion: Using a semantic fusion algorithm based on knowledge graph enhancement, we first build a knowledge graph that covers a wealth of medical knowledge, map and associate medical concepts in different data sources with entities in the knowledge graph, and use the predefined semantic relationships and logical reasoning capabilities in the knowledge graph to accurately determine the semantic similarity between medical concepts expressed in different texts, thus achieving more accurate semantic fusion.

[0013] Data storage module: Combining the large-capacity storage advantages of distributed file systems (such as HDFS) and the relational query advantages of graph databases (such as Neo4j), massive large file data such as images and texts are stored in the distributed file system, while the complex relationships between the fused data are stored and efficiently queried using the graph database. While meeting the large data storage requirements, it can quickly respond to complex relational query requests, providing strong support for subsequent deep data mining and analysis.

[0014] Data access and application interface module: Provides a unified data access interface for various application systems within the hospital and user terminals such as medical staff, managers, etc. to call and obtain corresponding integrated data according to permissions and needs.

[0015] Furthermore, the data collection module includes a compatible mode for multiple types of data sources: in actual data application scenarios, data sources are often diverse, and a flexible and efficient collection mode: different types of data sources usually require different collection modes and tools.

[0016] Furthermore, the data preprocessing module performs operations such as cleaning, conversion and normalization on the collected raw data, wherein the cleaning operation mainly removes noise in the data, the conversion performs unified format conversion on data of different formats, and the normalization operation standardizes data of different dimensions.

[0017] Furthermore, the data fusion engine module is based on entity recognition technology to accurately identify data representing the same object in different data sources, and through the patient's unique identification, all relevant data about the patient from HIS, LIS, PACS, EMR and other systems are linked to build a complete patient data view, presenting the overall picture of the patient's diagnosis and treatment in the hospital.

[0018] Furthermore, the data analysis module provides auxiliary diagnosis and treatment functions for clinical doctors. By analyzing the fused patient data and using machine learning, data mining and other technologies, it provides doctors with reference suggestions for disease diagnosis.

[0019] Furthermore, the data analysis module also assists hospital management decision-making by analyzing data such as department business volume, patient satisfaction, and medical resource utilization, providing data basis for hospital management to rationally allocate human and material resources, optimize department layout, and formulate development strategies.

[0020] Furthermore, the data access and application interface module supports multiple interface forms such as RESTful API for easy integration.

[0021] Beneficial effects of the present invention:

[0022] (1) The present invention significantly improves the quality of medical services: through deep data integration and intelligent push functions, it reduces diagnosis time and improves diagnosis accuracy. Medical staff can understand the patient's condition more quickly, comprehensively and accurately, thereby providing patients with better quality and more efficient medical services.

[0023] (2) Improving system adaptability and scalability: In the face of the ever-changing information needs of hospitals, the data fusion system of the present invention can easily cope with the access of new data sources and the expansion of new business scenarios due to its intelligent adaptive collection, adaptive fusion algorithm, and flexible storage and interface design. There is no need to reconstruct the system on a large scale, which greatly reduces the cost of system upgrades and maintenance.

[0024] (3) Enhance the scientific nature of hospital management decisions: With real-time, accurate, and comprehensive data and intelligent data display functions for managers, hospital managers can accurately understand the hospital's operating conditions, adjust management strategies in a timely manner, and optimize resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be further described below in conjunction with the accompanying drawings.

[0026] Figure 1 This is a schematic diagram of system module connections provided by the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] See also Figure 1 As shown, a data fusion system for digital hospitals includes:

[0029] Data acquisition module: It has an intelligent adaptation mechanism that can automatically identify the type of data source and dynamically load the corresponding acquisition plug-in for data acquisition. For example, for the data generated by a newly connected niche but industry-standard inspection device, the system can automatically identify its data output interface specification and call the corresponding acquisition plug-in. It can quickly realize data acquisition without a lot of manual configuration, which greatly improves the access efficiency and compatibility of new data sources. After the system is started, the data acquisition module will continuously monitor the network status, including the access status of new devices. When a new device is detected to be connected to the network, the module will respond immediately and prepare for the next step of device identification;

[0030] Data preprocessing module: intelligent data verification and automatic repair functions. When cleaning data, in addition to conventional rule verification (such as value range, format legitimacy, etc.), machine learning models are also used to learn normal and abnormal patterns in historical data, so that more accurate judgments can be made on suspected abnormal data. Moreover, when a repairable error is found (such as a partial digit error in the patient's contact information due to input errors, which can be repaired intelligently by comparing the contact information of patients in the same region and age group with the statistical rules of the patients), it can be repaired automatically or the operator can be prompted to confirm the repair, effectively improving data quality and reducing manual intervention costs;

[0031] Data analysis module: provides a wealth of data analysis tools and algorithms, supports in-depth mining and analysis of integrated data, discovers patterns and trends in the data, and uses advanced data analysis tools and algorithms to conduct in-depth mining and analysis of integrated data, discovers patterns and trends in the data, and provides a scientific basis for the hospital's decision-making;

[0032] Data fusion engine module, including:

[0033] Entity recognition and matching: Adopting an adaptive entity recognition and matching algorithm, it first performs preliminary screening based on the initial general rules, and then uses the graph neural network (GNN) technology in deep learning to construct the data entities in the data source into a graph structure. It dynamically captures the complex relationship between entities through node and edge feature learning, and continuously optimizes the matching results. For example, for the patient identity information registered by the hospital using different coding rules at different times, the algorithm can adaptively learn the potential associations between different codes and accurately identify all relevant data belonging to the same patient, even if some information is ambiguous or inconsistent.

[0034] Semantic fusion: Using the semantic fusion algorithm based on knowledge graph enhancement, we first build a knowledge graph that covers a wealth of medical knowledge, map and associate medical concepts in different data sources with entities in the knowledge graph, and use the predefined semantic relationships and logical reasoning capabilities in the knowledge graph to accurately determine the semantic similarity between medical concepts expressed in different texts, and achieve more accurate semantic fusion. For example, for the various abbreviations and common names that may exist in different systems for "coronary atherosclerotic heart disease", the semantic analysis of the knowledge graph can accurately identify and fuse related data, avoiding the semantic misjudgment problem that is prone to occur in traditional simple similarity calculations based on word vectors;

[0035] The data fusion engine module is also based on entity recognition technology to accurately identify data representing the same object in different data sources. Through the patient's unique identifier, it links all relevant data about the patient from HIS, LIS, PACS, EMR and other systems to build a complete patient data view, presenting the patient's full picture of the hospital's diagnosis and treatment.

[0036] Data storage module: Combining the large-capacity storage advantages of distributed file systems (such as HDFS) and the relational query advantages of graph databases (such as Neo4j), massive large file data such as images and texts are stored in the distributed file system, while the complex relationships between the fused data are stored and efficiently queried using the graph database. While meeting the large data storage requirements, it can quickly respond to complex relational query requests, providing strong support for subsequent deep data mining and analysis.

[0037] Data access and application interface module: Provides a unified data access interface for various application systems within the hospital and user terminals such as medical staff, managers, etc. to call and obtain corresponding integrated data according to permissions and needs.

[0038] The data collection module includes compatible methods for multiple types of data sources: In actual data application scenarios, data sources are often diverse, as well as flexible and efficient collection methods: Different types of data sources usually require different collection methods and tools.

[0039] The data preprocessing module performs operations such as cleaning, conversion and normalization on the collected raw data. The cleaning operation is mainly to remove noise in the data, the conversion is to convert data of different formats into a unified format, and the normalization operation is to standardize data of different dimensions.

[0040] The data fusion engine module is based on entity recognition technology, which accurately identifies data representing the same object in different data sources. Through the patient's unique identifier, it links all relevant data about the patient from HIS, LIS, PACS, EMR and other systems, builds a complete patient data view, and presents the patient's overall diagnosis and treatment in the hospital.

[0041] The data analysis module provides auxiliary diagnosis and treatment functions for clinicians. By analyzing the integrated patient data and using machine learning, data mining and other technologies, it provides doctors with reference suggestions for disease diagnosis.

[0042] The data analysis module also assists hospital management decision-making by analyzing data such as department business volume, patient satisfaction, and medical resource utilization, providing data basis for hospital management to rationally allocate human and material resources, optimize department layout, and formulate development strategies.

[0043] The data access and application interface module supports multiple interface forms such as RESTful API for easy integration.

[0044] Embodiment 1

[0045] An example of the data collection process: Take the case of a newly connected medical device with a special data output format. After the system is started, the data collection module automatically detects that the new device is connected to the network, and identifies its data type and corresponding collection rules by analyzing the protocol header information and other features sent by its data port. Then, it dynamically loads the appropriate plug-in from the pre-configured collection plug-in library (if there is no complete match in the plug-in library, the administrator can be prompted to perform a simple configuration to generate a new plug-in), establishes a connection with the device in the manner set by the plug-in, collects the data generated by the device in real time (such as physiological monitoring data in a specific format generated by the device, etc.), and preliminarily encapsulates the collected data and passes it to the data preprocessing module. The entire process does not require manual complex interface development and configuration. Compared with the existing technology that requires manual writing of a large amount of collection code and debugging interfaces for new devices, it greatly shortens the access cycle of new data sources.

[0046] Embodiment 2

[0047] Data preprocessing operation example: After receiving data from the acquisition module, the data preprocessing module first uses the anomaly detection model trained by machine learning to scan the data. For example, for a batch of newly collected patient vital signs data, the model determines whether there are abnormal values ​​based on the distribution pattern of historical normal signs data and the correlation between different signs. If it is found that a patient's heart rate data is obviously beyond the normal range of people of the same age and health status, and combined with other relevant signs data, it is judged that it may be an input error (rather than a real abnormal physiological condition), then an attempt is made to automatically repair the erroneous heart rate data by comparing and analyzing the patient's other recent examination results and the average signs data of patients of the same type. The repaired data is then converted to the conventional format to meet the requirements of subsequent fusion processing. Compared with the existing method of only performing simple format and range verification, this preprocessing method of intelligent verification and repair effectively improves the accuracy and availability of the data.

[0048] Embodiment 3

[0049] Specific operation examples of the data fusion engine:

[0050] 1. Entity recognition and matching operations: Assuming that a batch of data to be integrated is obtained from multiple data sources (HIS, LIS, PACS, EMR, etc.), the adaptive entity recognition and matching algorithm is first started. The algorithm first performs preliminary grouping according to the basic patient identity information matching rules (such as name, ID number, etc.), and grouping the records that may belong to the same patient. Then, these grouped data are constructed into a graph structure, with patient-related data entities as nodes and the associations between entities (such as a test result corresponding to a patient's diagnosis and treatment activity, etc.) as edges, which are input into the graph neural network for training and learning. As the learning deepens, the network can dynamically capture the hidden associations between entities in different data sources. For example, it is found that although the patient address information entered at different times is slightly different, it can be judged to be the same address through the surrounding geographic information, associated contacts and other features, and then accurately match all data belonging to the same patient. Even if some information is missing or vague, a high matching accuracy rate can be guaranteed, which is difficult for existing entity recognition algorithms that rely on fixed rules to do.

[0051] 2. Semantic fusion operation: When performing semantic fusion, for text description data about disease diagnosis, treatment methods, etc. in different systems, first map the medical concepts in these texts with the entities in the constructed medical knowledge graph. For example, the "myocardial infarction" mentioned in the text will correspond to the superordinate concept of "coronary atherosclerotic heart disease" in the knowledge graph and its related detailed knowledge nodes such as pathology and treatment. Through the semantic relationships and logical reasoning defined in the knowledge graph, the semantic similarity of similar concept expressions in different texts is judged, and the semantically similar content is fused, and the relevant diagnostic basis, treatment process and other information are linked according to reasonable logic to generate a fused semantically complete and accurate description, which overcomes the problem of semantic deviation that is prone to occur in traditional calculations based only on simple text similarity.

[0052] Embodiment 3

[0053] Data access and application interface call example: When a clinician logs into the system and opens a patient's diagnosis and treatment interface, the data access and application interface module first identifies the doctor's role and the disease type of the patient being treated (such as diabetes based on the patient's existing preliminary diagnosis information), and then retrieves the patient's fused data (including previous blood sugar test results, medication records, dietary instructions, etc.) from the data storage module according to the preset intelligent service rules. At the same time, it also pushes the latest treatment plan recommendations in the relevant diabetes diagnosis and treatment guidelines, successful treatment cases of patients with similar conditions in the hospital, and other auxiliary information to facilitate doctors to quickly refer to and formulate the best diagnosis and treatment plan. For hospital managers, when entering the operation analysis interface, the system automatically extracts relevant data from the data storage module based on the indicators they are concerned about (such as the inventory turnover rate of drugs in each department), and displays it in intuitive visual forms such as bar charts and line charts after aggregation analysis, helping managers quickly gain insight into the management problems reflected behind the data. These personalized and intelligent interface services are not available in existing ordinary data query interfaces.

[0054] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A data fusion system for digital hospitals, characterized in that: include: Data collection module: It has an intelligent adaptation mechanism that can automatically identify the type of data source and dynamically load the corresponding collection plug-in to obtain data; Data preprocessing module: Integrates intelligent data verification and automatic repair functions. When cleaning data, in addition to conventional rule verification, it also uses machine learning models to learn normal and abnormal patterns in historical data, so that it can make more accurate judgments on suspected abnormal data. Moreover, when a repairable error is found, it can automatically repair it or prompt the operator to confirm the repair. Data analysis module: provides a variety of data analysis tools and algorithms, supports in-depth mining and analysis of integrated data, and discovers patterns and trends in the data; Data fusion engine module, including: In terms of entity recognition and matching: an adaptive entity recognition and matching algorithm is used; in terms of semantic fusion: a semantic fusion algorithm based on knowledge graph enhancement is used; Data storage module: Combining the large-capacity storage advantages of distributed file systems (such as HDFS) and the relational query advantages of graph databases (such as Neo4j), massive image, text and other large file data are stored in the distributed file system, while the complex relationships between the fused data are stored and efficiently queried using the graph database. While meeting the large data storage requirements, it can quickly respond to complex relational query requests, providing strong support for subsequent deep data mining and analysis; Data access and application interface module: Provides a unified data access interface for various application systems within the hospital and user terminals such as medical staff, managers, etc. to call and obtain corresponding integrated data according to permissions and needs.

2. A data fusion system for digital hospitals according to claim 1, characterized in that: The data collection module includes a multi-type data source compatible mode: in actual data application scenarios, data sources are often diverse, and a flexible and efficient collection mode: different types of data sources usually require different collection methods and tools.

3. The data fusion system for digital hospitals according to claim 1 is characterized in that: The data preprocessing module performs operations such as cleaning, conversion and normalization on the collected raw data, wherein the cleaning operation mainly removes noise in the data, the conversion converts data of different formats into a unified format, and the normalization operation standardizes data of different dimensions.

4. The data fusion system for digital hospitals according to claim 1 is characterized in that: The data fusion engine module is based on entity recognition technology to accurately identify data representing the same object in different data sources, and associates all relevant data about the patient from HIS, LIS, PACS, EMR and other systems through the patient's unique identification, to build a complete patient data view, and present a comprehensive picture of the patient's diagnosis and treatment in the hospital.

5. The data fusion system for digital hospitals according to claim 1 is characterized in that: The data analysis module provides auxiliary diagnosis and treatment functions for clinicians. By analyzing the fused patient data and using machine learning, data mining and other technologies, it provides doctors with reference suggestions for disease diagnosis.

6. The data fusion system for digital hospitals according to claim 1 is characterized in that: The data analysis module also assists hospital management decision-making by analyzing data such as department business volume, patient satisfaction, and medical resource utilization, providing data basis for hospital management to rationally allocate human and material resources, optimize department layout, and formulate development strategies.

7. The data fusion system for digital hospitals according to claim 1 is characterized in that: The data access and application interface module supports multiple interface forms such as RESTful API for easy integration.

Images