Emergency treatment field knowledge extraction system and method based on multi-source data

Abstract

The invention relates to the technical field of emergency treatment field knowledge extraction, in particular to an emergency treatment field knowledge extraction system and method based on multi-source data. The method comprises the following steps: obtaining historical electronic health record data of a patient, and obtaining electronic health record desensitization data through desensitization processing; extracting medical record patient signs and historical ultrasonic image data from the desensitization data; performing symptom expression correlation estimation based on the ultrasonic image data and the physical sign data, performing historical symptom correlation analysis in combination with the electronic health record desensitization data, and generating symptom correlation data; constructing an emergency treatment field knowledge graph by using the disease associated data, carrying out emergency treatment field research risk assessment on the electronic health record data of the patient through the knowledge graph, and optimizing the knowledge graph in combination with a risk assessment result; according to the invention, through optimization of the knowledge graph in the emergency treatment field, the risk assessment precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency field knowledge extraction, and in particular to an emergency field knowledge extraction system and method based on multi-source data. Background Art

[0002] The patient's electronic health record (EHR), vital sign data, ultrasound images and other information sources together constitute a complex medical knowledge system. How to extract useful knowledge from these multi-source data and integrate them effectively is an urgent problem to be solved in the current medical field. To this end, the knowledge extraction method in the emergency field should be able to analyze data through multi-level and multi-dimensional data. Ultrasound images are one of the common medical images. Through the texture feature analysis and edge feature extraction of ultrasound images, the shape and nature of the lesion area can be identified, providing an intuitive basis for emergency diseases. However, a single image data analysis cannot fully reflect the patient's clinical condition. It must be combined with multiple data sources such as medical record patient vital sign data and historical medical condition information. This method extracts relevant symptom manifestation data and historical disease treatment related data through desensitization and intelligent analysis of the patient's historical electronic health records, ultrasound images, vital sign data, etc., and constructs an emergency field knowledge graph. However, the traditional emergency field knowledge extraction system and method based on multi-source data has the problem of inaccurate risk analysis of central venous catheters and inaccurate analysis of disease research related data. Summary of the invention

[0003] Based on this, it is necessary to provide an emergency field knowledge extraction system and method based on multi-source data to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for extracting emergency field knowledge based on multi-source data includes the following steps:

[0005] Step S1: Acquire the patient's historical electronic health record data; perform record desensitization processing according to the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitized data; collect the patient's physical signs according to the patient's historical electronic health record desensitized data to obtain the patient's physical signs data; acquire the patient's historical ultrasound image according to the patient's historical electronic health record desensitized data to obtain the patient's historical ultrasound image data;

[0006] Step S2: performing a symptom manifestation correlation estimation based on the patient's historical ultrasound image data and the patient's medical record vital sign data to obtain symptom manifestation correlation data; performing a historical symptom treatment correlation analysis based on the symptom manifestation correlation data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment correlation data;

[0007] Step S3: constructing an emergency field knowledge graph based on historical disease treatment related data and disease manifestation related data to obtain emergency field knowledge graph data; estimating emergency field research plans based on the emergency field knowledge graph data to obtain emergency field research plan data;

[0008] Step S4: Conduct emergency field research risk assessment based on the emergency field research plan data to obtain emergency field research risk data; optimize the emergency field knowledge graph data based on the emergency field research risk data to obtain emergency field knowledge graph optimized data.

[0009] The present invention obtains and desensitizes the patient's historical electronic health record data to ensure the privacy of the data and make it suitable for further analysis. By extracting the patient's physical sign data and ultrasound image data from the medical record, a comprehensive patient health record can be constructed, thereby providing high-quality input for subsequent data analysis. In this process, the combined analysis of medical record data and image data not only reveals the patient's clinical characteristics, but also provides a key basis for in-depth understanding of symptom manifestations and associations. In the process of symptom manifestation association estimation, through in-depth analysis of patient historical data, the potential association between different symptom manifestations can be revealed, which provides accurate support for subsequent disease pattern recognition and derivation of emergency research programs, and provides comprehensive knowledge support for research in the emergency field. By constructing a knowledge graph in the emergency field, these data are integrated in a structured manner to form an in-depth knowledge base to help identify key trends and potential research directions in the field of emergency medicine. The estimation of research programs in the emergency field is supported by the knowledge graph, so that research work can be carried out under a more accurate framework, thereby effectively identifying research hotspots and potential risk points in the emergency field. Based on these research plan data, an emergency field research risk assessment is conducted, so that potential risks in the research process can be discovered and controlled in a timely manner, the research direction and methods can be optimized, and the effectiveness and credibility of the research can be further improved. By optimizing the knowledge graph in the emergency field, the knowledge system in the emergency field can be continuously updated and improved, and the continuous progress of emergency medicine can be promoted. This series of steps enhances the depth and breadth of emergency field research through multi-dimensional data analysis, and improves the ability to understand and explore emergency problems. Therefore, the present invention is an optimization of the traditional emergency field knowledge extraction method based on multi-source data, which solves the problem of inaccurate central venous catheter risk analysis and inaccurate disease research-related data in the traditional emergency field knowledge extraction method based on multi-source data, and improves the accuracy of central venous catheter risk analysis and the accuracy of disease research-related data.

[0010] Preferably, step S1 comprises the following steps:

[0011] Step S11: Obtaining patient historical electronic health record data;

[0012] Step S12: performing record desensitization processing according to the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitization data;

[0013] Step S13: Collecting medical record patient vital signs based on the patient's historical electronic health record desensitized data to obtain medical record patient vital sign data;

[0014] Step S14: Acquire the patient's historical ultrasound image based on the patient's historical electronic health record desensitized data to obtain the patient's historical ultrasound image data.

[0015] The present invention obtains the patient's historical electronic health record data and performs desensitization processing to ensure that the patient's privacy is effectively protected, while providing clear and secure basic data for subsequent analysis. This process not only eliminates the hidden dangers of data privacy, but also improves the credibility of the data. Based on the desensitized electronic health record data, the vital signs of medical record patients are collected, so that the vital signs information of different patients can be comprehensively collected, ensuring the accuracy and comprehensiveness of the data relied on for subsequent analysis. At the same time, the acquisition of the patient's historical ultrasound images provides rich visual information for image data analysis, and combined with the detailed content of the electronic health record, disease analysis and lesion detection can be better performed. The integration and processing of these data not only optimizes the data utilization efficiency in emergency medicine, but also enhances the multi-dimensional understanding of complex diseases and promotes the process of medical data intelligence.

[0016] Preferably, step S2 comprises the following steps:

[0017] Step S21: Analyze the patient's symptom manifestations according to the patient's historical ultrasound image data and the patient's physical sign data in the medical history to obtain the patient's symptom manifestation data;

[0018] Step S22: collecting historical patient conditions based on the desensitized data of the patient's historical electronic health records to obtain historical patient condition data;

[0019] Step S23: estimating the correlation between symptom manifestations based on historical patient condition data and patient symptom manifestation data to obtain symptom manifestation correlation data;

[0020] Step S24: Perform historical symptom treatment association analysis based on the symptom manifestation association data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment association data.

[0021] The present invention can effectively extract the patient's symptom data through comprehensive analysis of the patient's historical ultrasound image data and the patient's physical sign data in the medical record, laying the foundation for further disease association analysis. This process enables the potential connection between different symptoms to be revealed through the combination of multi-source data, which helps to deeply understand the clinical manifestations of patients in emergency situations and provides high-quality input information for subsequent disease prediction and association analysis. The collection of historical patient condition data, based on desensitized electronic health record data, can ensure the privacy protection of data and provide an important historical background for the prediction of symptom association. The prediction of symptom association is carried out on the basis of historical condition data and symptom data, which helps to reveal the symptom development pattern and potential association in the disease process, thereby providing a more detailed reference for future emergency research. This analysis enables an in-depth understanding of patient symptoms and provides more comprehensive background information for researchers in the emergency field. In addition, by further combining and analyzing symptom data with historical electronic health record data, valuable insights can be provided for research program optimization and disease-related research in the emergency field. This process effectively connects the patient's historical condition and symptom data, providing accurate data support for a comprehensive analysis of the relationship between symptoms and symptoms.

[0022] Preferably, step S21 includes the following steps:

[0023] Step S211: constructing a gray-level co-occurrence matrix according to the patient's historical ultrasound image data to obtain ultrasound image gray-level co-occurrence matrix data;

[0024] Step S212: performing ultrasonic image texture feature recognition on the ultrasonic image gray level co-occurrence matrix data to obtain ultrasonic image texture feature data;

[0025] Step S213: performing a physical sign detection on the patient in the medical record based on the physical sign data of the patient in the medical record to obtain the physical sign data of the patient in the medical record;

[0026] Step S214: Analyze the patient's symptom manifestation based on the ultrasound image texture feature data and the patient's internal physical sign data in the medical record to obtain the patient's symptom manifestation data.

[0027] The present invention reveals the texture features in the image by constructing a grayscale co-occurrence matrix based on the patient's historical ultrasound image data, providing rich image information for subsequent analysis. These grayscale co-occurrence matrix data provide a basis for texture feature recognition, which can capture subtle changes in the image, thereby identifying more accurate texture features and deeply understanding the imaging manifestations of the disease. At the same time, in vivo physical sign detection based on the patient's physical sign data in the medical record helps to comprehensively collect and analyze the patient's physical signs, laying a data foundation for subsequent symptom manifestation analysis. By combining ultrasound image texture features with in vivo physical sign data, symptom manifestation analysis can be performed with the support of multidimensional data, thereby revealing the complexity and diversity of the disease. The comprehensiveness and reliability of data fusion are improved, which helps to build more refined and comprehensive patient historical symptom manifestation data.

[0028] Preferably, step S212 includes the following steps:

[0029] Performing image grayscale value change intensity detection on the ultrasound image grayscale co-occurrence matrix data, thereby obtaining ultrasound image grayscale value change intensity data;

[0030] Performing ultrasound image contrast analysis according to the ultrasound image gray value change intensity data, thereby obtaining ultrasound image contrast data;

[0031] Acquiring the ultrasonic image entropy value according to the ultrasonic image gray level co-occurrence matrix data, thereby obtaining the ultrasonic image entropy value data;

[0032] Performing ultrasound image texture complexity analysis based on ultrasound image entropy value data and ultrasound image gray-level co-occurrence matrix data, thereby obtaining ultrasound image texture complexity data;

[0033] Ultrasonic image texture feature analysis is performed according to the ultrasonic image contrast data and the ultrasonic image texture complexity data, so as to obtain ultrasonic image texture feature data.

[0034] The present invention effectively extracts the detail changes in the image by detecting the intensity of grayscale value changes on the grayscale co-occurrence matrix data of the ultrasonic image, reflecting the contrast and texture information between different regions. This process reveals the intensity of changes in brightness and color in the image, providing a basis for subsequent contrast analysis. By performing contrast analysis on the intensity data of grayscale value changes in the ultrasonic image, the clarity and detail presentation of the image can be quantified, providing important parameters for further imaging research. At the same time, obtaining entropy values ​​from the grayscale co-occurrence matrix can measure the randomness and information complexity of the image, providing intuitive data support for the tissue structure characteristics of the image. By combining entropy values ​​and grayscale co-occurrence matrix data, texture complexity analysis is performed, which not only reveals the detail complexity of the image, but also identifies abnormal areas in the image. Combining contrast data with texture complexity data to perform texture feature analysis of ultrasonic images helps to have a deeper understanding of the potential structures and pathological features in the image, comprehensively improve the accuracy and meticulousness of image analysis, and thus provide more reliable image feature data support for further medical analysis.

[0035] Preferably, step S214 includes the following steps:

[0036] Calculate the gradient change of texture feature direction according to the texture feature data of the ultrasonic image to obtain the gradient change data of the texture feature direction of the image;

[0037] Estimating the edge feature of the ultrasound image based on the directional gradient change data of the image texture feature, thereby obtaining the edge feature data of the ultrasound image;

[0038] Acquiring boundary features of the lesion region according to the texture feature data of the ultrasound image and the edge feature data of the ultrasound image to obtain boundary feature data of the lesion region of the ultrasound image;

[0039] Performing geometric feature analysis on the lesion region boundary feature data and the ultrasound image texture feature data to obtain geometric shape feature data of the lesion region;

[0040] The pain level of internal body signs is assessed according to the internal body sign data of the patient in the medical record, thereby obtaining the pain level data of the internal body signs of the patient in the medical record;

[0041] Based on the patient's physical sign pain level data and the lesion area geometric shape feature data in the medical record, the patient's symptom manifestation analysis is performed to obtain the patient's symptom manifestation data.

[0042] The present invention calculates the gradient change of the texture feature direction of the ultrasonic image texture feature data, effectively extracts the texture change in different directions in the image, and provides detailed information for further analyzing the spatial distribution characteristics of the image. This process can not only reveal the local structural differences of the image, but also provide accurate data support for the estimation of edge features. Estimating edge features based on the gradient change data of the texture feature direction helps to accurately locate the boundary information in the image and enhance the recognition ability of the lesion area. Combining the ultrasonic image texture feature data and edge feature data to collect the boundary features of the lesion area helps to extract the specific boundary data of the lesion area and provide a clear boundary division for the lesion area in the image. Through the combined analysis of the boundary feature data and texture feature data of the lesion area, the geometric shape features of the lesion area can be further analyzed, and the specific morphology and complexity of the lesion can be fully described, thereby improving the analysis depth of the image. At the same time, based on the in vivo physical sign data of the patient in the medical record, the pain level is evaluated, the patient's in vivo response is quantified, and combined with the geometric feature data of the lesion area, more accurate background information can be provided for the comprehensive analysis of symptom manifestations, thereby optimizing the understanding of the lesion and its impact.

[0043] Preferably, step S3 comprises the following steps:

[0044] Step S31: extracting entity relationships based on historical disease treatment related data and disease manifestation related data to obtain disease entity relationship data;

[0045] Step S32: constructing an emergency field knowledge graph based on the disease entity relationship data to obtain emergency field knowledge graph data;

[0046] Step S33: Estimate the disease type in the emergency field based on the patient's historical electronic health record desensitized data according to the emergency field knowledge graph data to obtain emergency field disease type estimation data;

[0047] Step S34: Estimate the research plan in the emergency field based on the disease type estimation data in the emergency field and the knowledge graph data in the emergency field to obtain the research plan data in the emergency field.

[0048] The present invention can extract entity relationships based on historical disease processing related data and disease manifestation related data, and can dig out the potential relationship between key diseases and symptoms from a huge data set, and clarify the relationship between different diseases and their manifestations by extracting entities. This process provides a clear framework and effective data support for the subsequent construction of knowledge graphs, thereby enhancing the understanding of diseases and their manifestations in the emergency field. Based on the extracted disease entity relationship data, the knowledge graph in the emergency field is constructed, which not only integrates multi-source information such as historical disease data and symptom manifestation data, but also makes disease-related knowledge more structured and visualized, which helps to reveal disease patterns, processes and potential associations, and lays a solid knowledge foundation for in-depth research in the emergency field. Further, based on the data of the knowledge graph in the emergency field, disease type estimation is performed, and the constructed knowledge graph system is used to accurately analyze and predict the disease type, and the desensitized data of the patient's historical electronic health record is combined to obtain the disease type estimation results in a specific emergency environment. This process effectively improves the accuracy of disease prediction in the emergency field and promotes the scientificity and pertinence of emergency research. Through the further application of knowledge graphs, it is possible to predict research plans in the emergency field based on the estimated data of disease types, thereby providing new research directions and data support for researchers in the emergency field and promoting innovation and development in the field of emergency medicine.

[0049] Preferably, step S4 comprises the following steps:

[0050] Step S41: conducting emergency field research risk assessment according to emergency field research program data to obtain emergency field research risk data;

[0051] Step S42: performing emergency field knowledge reasoning on the symptom research associated data and the symptom manifestation associated data according to the emergency field research risk data to obtain emergency field knowledge reasoning data;

[0052] Step S43: Optimize the knowledge graph data in the emergency field according to the emergency field research risk data and the emergency field knowledge reasoning data to obtain the emergency field knowledge graph optimized data.

[0053] The present invention provides risk identification and assessment for research activities in the emergency field by conducting research risk assessment on research program data in the emergency field, thereby revealing potential research bottlenecks and risk factors. This process can help researchers deeply understand the complexity of current emergency field research and the risks it faces, and provide more rational and systematic guidance for the formulation and implementation of the next research program. With the analysis of the risk data of emergency field research, knowledge reasoning can be performed on the disease research association data and the disease manifestation association data, and then the correlation and potential influence mechanism between diseases and symptoms can be revealed through reasoning, providing a new perspective and a deeper understanding for the research in the emergency field. Based on the risk data and knowledge reasoning data of emergency field research, the knowledge graph of emergency field is optimized to ensure that the knowledge graph is more accurate, comprehensive and meets the actual research needs. The optimized knowledge graph not only reflects the relationship between symptoms more accurately, but also combines risk assessment and reasoning data to provide newer and more forward-looking data support for further research in the emergency field, thereby promoting the in-depth development of the field of emergency medicine. This series of analysis and optimization operations effectively improves the research efficiency and quality in the emergency field and maximizes the use of existing multi-source data. According to the output emergency field knowledge graph, the data is optimized for deep data extraction, and comprehensive and targeted related information is obtained by extracting key entities and relationship data related to emergency. In the data extraction process, multi-level data analysis methods, including but not limited to statistical analysis, correlation analysis, and trend analysis, are applied to comprehensively interpret the data to reveal potential patterns and key influencing factors. In addition, by analyzing the weaknesses or weak links in the data, the focus of the training module can be optimized to make the training more effective. The results of data analysis can also be used to improve work quality. For example, based on the related data displayed in the knowledge graph, a more scientific and reasonable work process can be formulated to improve the efficiency of information flow and ensure that the execution of emergency-related tasks is more organized and standardized. Through this data-driven improvement process, not only can the training effect be improved, but also the overall work quality can be effectively improved.

[0054] Preferably, step S41 includes the following steps:

[0055] Step S411: performing a central venous catheter usage status analysis based on the emergency field research program data to obtain central venous catheter usage status data;

[0056] Step S412: performing statistics on the usage time of the central venous catheter according to the central venous catheter usage status data to obtain the central venous catheter usage time data;

[0057] Step S413: performing a central venous catheter thrombosis prediction on the central venous catheter usage time data to obtain central venous catheter thrombosis data;

[0058] Step S414: Calculate the probability of bacterial infection of the central venous catheter according to the central venous catheter usage time data and the central venous catheter thrombosis generation data to obtain the central venous catheter bacterial infection probability data;

[0059] Step S415: performing emergency field complication risk inference based on the central venous catheter bacterial infection probability data and the central venous catheter thrombosis generation data to obtain emergency field complication risk data;

[0060] Step S416: Conduct emergency field research risk assessment based on emergency field complication risk data and central venous catheter bacterial infection probability data to obtain emergency field research risk data.

[0061] The present invention analyzes the use of central venous catheters in emergency research program data, comprehensively understands the application of central venous catheters in different emergency situations, identifies its usage pattern, frequency and its changing trend, and provides detailed data support for further clinical and research activities. This analysis helps to build a basic database on the use of central venous catheters, so that the subsequent statistics and analysis of its usage time and related risks are more scientific and accurate. After the usage time statistics are performed based on the central venous catheter usage data, the specific usage time data of the device is obtained, which helps to identify the potential association between the usage time and complications (such as thrombosis and bacterial infection). This time statistics provides a quantitative basis for evaluating the safety of central venous catheter use and the health risks faced by patients, which helps to better conduct subsequent risk assessments. Thrombosis is estimated based on the central venous catheter usage time data, revealing the relationship between the usage time and thrombosis, and then providing important data support for the prediction of related complications in the emergency field. This estimate can identify potential high-risk situations in advance through in-depth analysis of historical data, thereby providing a basis for further intervention and research programs. The probability of bacterial infection is calculated by combining the duration of central venous catheter use and thrombosis data, further refining the assessment of central venous catheter-related complications. This calculation can more accurately assess the risk of bacterial infection, provide more information on the probability of bacterial infection, and provide detailed numerical references for subsequent research and risk control.

[0062] The present invention also provides an emergency field knowledge extraction system based on multi-source data, which is used to execute the emergency field knowledge extraction method based on multi-source data as described above. The emergency field knowledge extraction system based on multi-source data includes:

[0063] Patient historical ultrasound image acquisition module: obtain patient historical electronic health record data; perform record desensitization processing based on patient historical electronic health record data to obtain patient historical electronic health record desensitized data; collect patient vital signs from medical records based on patient historical electronic health record desensitized data to obtain patient vital signs data from medical records; obtain patient historical ultrasound images based on patient historical electronic health record desensitized data to obtain patient historical ultrasound image data;

[0064] Symptom treatment association analysis module: based on the patient's historical ultrasound image data and the patient's physical sign data in the medical record, the symptom manifestation association is estimated to obtain the symptom manifestation association data; based on the symptom manifestation association data and the patient's historical electronic health record desensitized data, the historical symptom treatment association analysis is performed to obtain the historical symptom treatment association data;

[0065] Research plan estimation module: construct the knowledge graph of the emergency field based on the historical disease treatment related data and the disease manifestation related data to obtain the knowledge graph data of the emergency field; estimate the research plan of the emergency field based on the knowledge graph data of the emergency field to obtain the research plan data of the emergency field;

[0066] Knowledge graph optimization module: conduct research risk assessment in the emergency field based on the research plan data in the emergency field to obtain research risk data in the emergency field; optimize the knowledge graph data in the emergency field based on the research risk data in the emergency field to obtain optimized knowledge graph data in the emergency field.

[0067] The present invention is to ensure the privacy of data and make it suitable for further analysis by obtaining and desensitizing the patient's historical electronic health record data. By extracting the patient's physical sign data and ultrasound image data from the medical record, a comprehensive patient health record can be constructed, thereby providing high-quality input for subsequent data analysis. In this process, the combined analysis of medical record data and image data not only reveals the patient's clinical characteristics, but also provides a key basis for in-depth understanding of symptom manifestations and associations. In the process of symptom manifestation association estimation, through in-depth analysis of the patient's historical data, the potential association between different symptom manifestations can be revealed, which provides accurate support for subsequent disease pattern recognition and derivation of emergency research programs. The association analysis of symptom manifestations is further combined with the historical symptom processing association data to provide comprehensive knowledge support for research in the emergency field. By constructing a knowledge graph in the emergency field, these data are integrated in a structured manner to form a knowledge base for in-depth discussion, which helps identify key trends and potential research directions in the field of emergency medicine. The estimation of emergency field research programs is supported by knowledge graphs, so that research work can be carried out under a more accurate framework, thereby effectively identifying research hotspots and potential risk points in the emergency field. Based on these research plan data, a risk assessment of emergency research is conducted, so that potential risks in the research process can be discovered and controlled in a timely manner, the research direction and methods can be optimized, and the effectiveness and credibility of the research can be further improved. By optimizing the knowledge graph in the emergency field, the knowledge system in the emergency field can be continuously updated and improved, and the continuous progress of emergency medicine can be promoted. This series of steps enhances the depth and breadth of emergency research through multi-dimensional data analysis, and improves the ability to understand and explore emergency problems. The present invention is an optimization of the traditional emergency field knowledge extraction method based on multi-source data, which solves the problem of inaccurate central venous catheter risk analysis and inaccurate disease research-related data in the traditional emergency field knowledge extraction method based on multi-source data, and improves the accuracy of central venous catheter risk analysis and the accuracy of disease research-related data. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flowchart of the steps of a method for extracting knowledge in the emergency field based on multi-source data;

[0069] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0070] Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0071] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0072] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0073] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0074] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0075] To achieve this, please refer to Figures 1 to 3 , a method for extracting emergency field knowledge based on multi-source data, comprising the following steps:

[0076] Step S1: Acquire the patient's historical electronic health record data; perform record desensitization processing according to the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitized data; collect the patient's physical signs according to the patient's historical electronic health record desensitized data to obtain the patient's physical signs data; acquire the patient's historical ultrasound image according to the patient's historical electronic health record desensitized data to obtain the patient's historical ultrasound image data;

[0077] Step S2: performing a symptom manifestation correlation estimation based on the patient's historical ultrasound image data and the patient's medical record vital sign data to obtain symptom manifestation correlation data; performing a historical symptom treatment correlation analysis based on the symptom manifestation correlation data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment correlation data;

[0078] Step S3: constructing an emergency field knowledge graph based on historical disease treatment related data and disease manifestation related data to obtain emergency field knowledge graph data; estimating emergency field research plans based on the emergency field knowledge graph data to obtain emergency field research plan data;

[0079] Step S4: Conduct emergency field research risk assessment based on the emergency field research plan data to obtain emergency field research risk data; optimize the emergency field knowledge graph data based on the emergency field research risk data to obtain emergency field knowledge graph optimized data.

[0080] In the embodiment of the present invention, reference Figure 1 As shown, it is a schematic diagram of the steps of a method for extracting emergency field knowledge based on multi-source data of the present invention. In this example, the method for extracting emergency field knowledge based on multi-source data includes the following steps:

[0081] Step S1: Acquire the patient's historical electronic health record data; perform record desensitization processing according to the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitized data; collect the patient's physical signs according to the patient's historical electronic health record desensitized data to obtain the patient's physical signs data; acquire the patient's historical ultrasound image according to the patient's historical electronic health record desensitized data to obtain the patient's historical ultrasound image data;

[0082] In an embodiment of the present invention, the patient's historical electronic health record data is obtained from the hospital's electronic health record system. This data contains the patient's basic information, medical history, physical sign information, test results, and known disease data. In order to ensure patient privacy, all personal identity information needs to be desensitized, for example, through encryption, deletion or blurring technology to ensure that the patient's identity information cannot be directly restored. The desensitized data is used for subsequent analysis processes. Next, the medical record patient physical sign data is extracted from the desensitized historical electronic health record data, including but not limited to vital signs, laboratory test results, etc. By consulting the patient's historical medical record files and combining the patient's past ultrasound examination records, historical ultrasound image data is extracted.

[0083] Step S2: performing a symptom manifestation correlation estimation based on the patient's historical ultrasound image data and the patient's medical record vital sign data to obtain symptom manifestation correlation data; performing a historical symptom treatment correlation analysis based on the symptom manifestation correlation data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment correlation data;

[0084] In an embodiment of the present invention, when the correlation between symptom manifestations is estimated based on the patient's historical ultrasound image data and the patient's physical sign data in the medical record, a grayscale histogram analysis is performed on the ultrasound image data to extract the grayscale distribution characteristics of the image; and a texture analysis algorithm is used to calculate the texture complexity parameters of the image, such as the contrast and entropy value of the co-occurrence matrix. Next, a cluster analysis is performed on the patient's physical sign data in the medical record to identify abnormal combination patterns of different physical sign indicators. Based on the joint analysis of ultrasound image features and physical sign indicators, association rule mining technology is used to generate symptom manifestation association rules to form symptom manifestation association data. Then, the associated data is matched with the symptom information recorded in the desensitized electronic health record, and research data related to similar symptoms in the history are extracted. The correlation between the symptom manifestation association data and the historical data is obtained through statistical analysis to form a historical symptom association analysis framework, and historical symptom processing association data is generated.

[0085] Step S3: constructing an emergency field knowledge graph based on historical disease treatment related data and disease manifestation related data to obtain emergency field knowledge graph data; estimating emergency field research plans based on the emergency field knowledge graph data to obtain emergency field research plan data;

[0086] In an embodiment of the present invention, when constructing a knowledge graph in the emergency field based on historical disease processing related data and disease manifestation related data, entities and relationships are extracted from the data. Key entities in the data, such as disease names, physical signs, imaging indicators, etc., are identified through natural language processing technology, and mapped to nodes in the knowledge graph. Subsequently, semantic analysis technology is used to extract relationships between entities from the data, such as "symptoms-related diseases", etc., to construct edges in the knowledge graph. Nodes and edges are organized into structured knowledge graph data through graph database technology (such as Neo4j). Next, based on the constructed knowledge graph, typical case data in the emergency field are statistically analyzed, and potential research topics are inferred based on the entity relationships of historical data, and research program data in the emergency field is formed, including research directions, target variables and background information.

[0087] Step S4: Conduct emergency field research risk assessment based on the emergency field research plan data to obtain emergency field research risk data; optimize the emergency field knowledge graph data based on the emergency field research risk data to obtain emergency field knowledge graph optimized data.

[0088] In an embodiment of the present invention, when conducting an emergency field research risk assessment based on the emergency field research program data, a comprehensive quality review is conducted on the variable design and data source of the research program, focusing on evaluating the deviation rate, missing rate and importance of key variables in the research. Then, based on the experimental conditions in the research program, the risk level of each condition is calculated by the risk matrix method, such as the tolerance range of low-risk conditions and sensitive parameters of high-risk conditions. Next, the research risk data is compared and analyzed with the existing data in the emergency field knowledge graph to identify the nodes and relationships that need to be optimized. The optimized area of ​​the knowledge graph is reconstructed using a graph algorithm, including deleting redundant nodes and strengthening the weights of associated edges, and finally the optimized emergency field knowledge graph optimization data is output.

[0089] Preferably, step S1 comprises the following steps:

[0090] Step S11: Obtaining patient historical electronic health record data;

[0091] Step S12: performing record desensitization processing according to the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitization data;

[0092] Step S13: Collecting medical record patient vital signs based on the patient's historical electronic health record desensitized data to obtain medical record patient vital sign data;

[0093] Step S14: Acquire the patient's historical ultrasound image based on the patient's historical electronic health record desensitized data to obtain the patient's historical ultrasound image data.

[0094] In an embodiment of the present invention, the patient's historical electronic health record data is obtained from the hospital's electronic health record system. The data usually includes the patient's basic personal information, past medical history, physical examination results, laboratory test data information, medication records, allergy history, surgical records, etc. During the data extraction process, the hospital information management system (HIS) and the electronic health record (EHR) database interface are used to access the data, and the data of relevant patients are obtained from different systems through SQL queries or API calls to ensure that the acquired information is complete and accurate. The acquired patient's historical electronic health record data is desensitized to ensure that it meets the requirements of personal privacy protection. Desensitization mainly includes technical means such as data encryption, anonymization, and data masking. Specifically, a hash algorithm (such as SHA-256) is used to encrypt the patient's name, ID number and other personal identity information or directly replace it with a randomly generated identifier to eliminate the relevance of the patient's identity. At the same time, for sensitive data related to the patient's medical history, such as disease name and historical processing information, data de-identification technology is used to desensitize to ensure that the patient's true identity cannot be restored through this information during the data processing process. The desensitized data will be stored in the data storage system as input data for subsequent analysis. According to the desensitized patient historical electronic health record data, the patient's vital sign data is collected from the medical record. This process requires extracting relevant vital sign information from the patient's electronic medical record, such as blood pressure, heart rate, body temperature, respiratory rate, blood sugar level, etc. The structured data in the medical record is extracted using a data analysis tool, and the vital sign information therein is summarized and standardized. For unstructured medical records (such as handwritten medical record texts), natural language processing technology (NLP) is used to perform text analysis on them to extract useful vital sign data. The patient's historical ultrasound examination records are extracted, and the ultrasound image data is usually saved in image files (such as DICOM format), which records the ultrasound examination results that the patient has received in the past. By connecting with the hospital's picture information system (PACS), the image acquisition interface is used to call the corresponding ultrasound image data. These image data are standardized and format converted by image processing software to ensure that all image data meet the requirements of subsequent processing. During the processing process, the image metadata (such as image acquisition time, ultrasound equipment model, image resolution, etc.) will also be collected and stored together with the image data.

[0095] Preferably, step S2 comprises the following steps:

[0096] Step S21: Analyze the patient's symptom manifestations according to the patient's historical ultrasound image data and the patient's physical sign data in the medical history to obtain the patient's symptom manifestation data;

[0097] Step S22: collecting historical patient conditions based on the desensitized data of the patient's historical electronic health records to obtain historical patient condition data;

[0098] Step S23: estimating the correlation between symptom manifestations based on historical patient condition data and patient symptom manifestation data to obtain symptom manifestation correlation data;

[0099] Step S24: Perform historical symptom treatment association analysis based on the symptom manifestation association data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment association data.

[0100] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0101] Step S21: Analyze the patient's symptom manifestations according to the patient's historical ultrasound image data and the patient's physical sign data in the medical history to obtain the patient's symptom manifestation data;

[0102] In an embodiment of the present invention, the patient's symptom performance analysis is performed based on the patient's historical ultrasound image data and the patient's medical record vital sign data. The analysis is based on the texture features of the patient's ultrasound image and the changes in vital sign data, combined with statistical analysis and image processing methods, to perform a quantitative assessment of symptoms, and use image processing technology to process the patient's historical ultrasound image to extract key features such as grayscale co-occurrence matrix, texture features, and image contrast. The image's contrast, entropy, texture complexity and other information are extracted through the image grayscale co-occurrence matrix (GLCM), and these features help reveal the performance of the lesion area. Then, combined with the vital sign information recorded in the medical record data, such as heart rate, blood pressure, body temperature, etc., it is standardized and analyzed synchronously with the ultrasound image data. This step generates the patient's symptom performance data, including the relationship between vital sign changes and imaging changes, through image analysis software combined with the vital sign data processing module.

[0103] Step S22: collecting historical patient conditions based on the desensitized data of the patient's historical electronic health records to obtain historical patient condition data;

[0104] In the embodiment of the present invention, the desensitized data is cleaned and structured, and the scattered text records, image descriptions and numerical indicators are organized into a directly readable data set through a unified data format. Then, based on the automatic extraction technology of disease condition keywords, such as part-of-speech tagging and semantic analysis methods in natural language processing technology, the core information in the disease condition description, such as the course of the disease, the duration of symptoms, etc., is extracted. By aggregating the data points of multiple records, a comprehensive historical patient condition data set is formed.

[0105] Step S23: estimating the correlation between symptom manifestations based on historical patient condition data and patient symptom manifestation data to obtain symptom manifestation correlation data;

[0106] In an embodiment of the present invention, the association between symptom manifestations is estimated based on historical patient condition data and patient symptom manifestation data. In this step, the correlation between symptoms and symptoms is analyzed by statistical models and data mining methods in combination with historical patient condition and symptom manifestation data, and the disease information in the historical condition data is used to extract key information such as the patient's disease type and course of disease. Then, in combination with the symptom manifestation data obtained in step S21, the correlation between symptoms and the condition is analyzed by establishing association rules, logistic regression analysis and other statistical methods. For example, if a patient has symptoms of hypertension accompanied by abnormal signals in the ultrasound image, the model will calculate the probabilistic relationship between these symptoms and the disease. This step relies on a powerful data processing platform and data mining technology to obtain symptom manifestation association data through big data analysis.

[0107] Step S24: Perform historical symptom treatment association analysis based on the symptom manifestation association data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment association data.

[0108] In an embodiment of the present invention, historical symptom treatment association analysis is performed based on symptom manifestation association data and patient historical electronic health record desensitized data. The symptom manifestation association data needs to be used as input, and the treatment path and related data corresponding to the current symptom manifestation are identified by comparing it with the symptoms and their treatment history in the desensitized electronic health record. For example, data mapping and knowledge mining technology are used to construct an association link from symptom manifestation to symptom characteristics and then to historical processing information, and the historical occurrence frequency, degree of association and corresponding subsequent development data of different treatment schemes are analyzed. By constructing a logical path between data, historical symptom treatment association data is finally generated.

[0109] Preferably, step S21 includes the following steps:

[0110] Step S211: constructing a gray-level co-occurrence matrix according to the patient's historical ultrasound image data to obtain ultrasound image gray-level co-occurrence matrix data;

[0111] Step S212: performing ultrasonic image texture feature recognition on the ultrasonic image gray level co-occurrence matrix data to obtain ultrasonic image texture feature data;

[0112] Step S213: performing a physical sign detection on the patient in the medical record based on the physical sign data of the patient in the medical record to obtain the physical sign data of the patient in the medical record;

[0113] Step S214: Analyze the patient's symptom manifestation based on the ultrasound image texture feature data and the patient's internal physical sign data in the medical record to obtain the patient's symptom manifestation data.

[0114] In an embodiment of the present invention, a grayscale co-occurrence matrix is ​​constructed according to the patient's historical ultrasound image data to obtain ultrasound image grayscale co-occurrence matrix data, obtain the patient's historical ultrasound image data, and convert the image into a grayscale image. For each ultrasound image, a grayscale co-occurrence matrix (GLCM) algorithm is used to process the image. The specific operation is to select a pixel point and calculate its grayscale value relationship with the pixel point in different directions and distances to form a co-occurrence matrix. By calculating the matrix, the texture information of the image, such as key parameters such as contrast, homogeneity, and entropy value, is obtained. These parameters can reflect the uniformity and complexity of the tissue structure in the ultrasound image. The obtained ultrasound image grayscale co-occurrence matrix data will include co-occurrence matrices in multiple directions (such as 0°, 45°, 90°, 135°) and at different distances. The ultrasound image grayscale co-occurrence matrix data is subjected to ultrasound image texture feature recognition to obtain ultrasound image texture feature data. In this step, the information in the grayscale co-occurrence matrix data is used to further extract the texture features of the ultrasound image. The specific operation is to extract multiple texture features from the constructed gray-level co-occurrence matrix, such as contrast, correlation, energy, uniformity, entropy, etc. Each feature represents the performance of different tissue structures or lesion areas in the ultrasound image. For example, the contrast feature reflects the difference in brightness in the image, while the energy feature reflects the detail richness of the image. Through the analysis of these features, the texture difference between the lesion tissue and the healthy tissue is revealed. Through professional image processing software or programming environment (such as Python's OpenCV or Matlab), the gray-level co-occurrence matrix is ​​calculated and analyzed to obtain the texture feature data of the ultrasound image. The in-vivo physical sign data of the medical record patient is detected based on the physical sign data of the medical record patient to obtain the in-vivo physical sign data of the medical record patient, and the physical sign data contained in the patient's historical electronic health record is obtained, including physiological parameters (such as blood pressure, heart rate, body temperature, respiratory rate, etc.), laboratory test data (such as blood routine, urine routine, imaging test results, etc.). Then, the physical sign data is standardized using a specific data processing method, for example, the physical sign data of different units are converted into a unified unit, or the physical sign data of different patients are normalized to eliminate the influence of individual differences on the analysis results. For different vital sign data, data smoothing, denoising and other technologies are used for preprocessing to ensure the accuracy and consistency of the data. The in vivo vital sign data obtained will include various physiological indicators of the patient at different time points. The patient's symptom performance analysis is performed based on the ultrasound image texture feature data and the in vivo vital sign data of the medical record patient to obtain the patient's symptom performance data. In this step, the ultrasound image texture feature data obtained in step S212 is combined with the in vivo vital sign data obtained in step S213 for comprehensive analysis, and the two types of data are aligned in chronological order to ensure that the vital sign data of the same patient in the same time period matches the ultrasound image feature data.Then, multidimensional data analysis methods (such as multivariate linear regression analysis or principal component analysis) are used to explore the relationship between the texture features of ultrasound images and the physiological signs of patients. For example, when the patient's physical sign data changes, the texture features of the corresponding area in the ultrasound image change, thereby revealing the patient's pathological manifestations. Through this analysis, the patient's symptom manifestation data obtained will include the lesion manifestations in the ultrasound image and the change trends in the physical sign data, further forming the patient's comprehensive symptom manifestation information.

[0115] Preferably, step S212 includes the following steps:

[0116] Performing image grayscale value change intensity detection on the ultrasound image grayscale co-occurrence matrix data, thereby obtaining ultrasound image grayscale value change intensity data;

[0117] Performing ultrasound image contrast analysis according to the ultrasound image gray value change intensity data, thereby obtaining ultrasound image contrast data;

[0118] Acquiring the ultrasonic image entropy value according to the ultrasonic image gray level co-occurrence matrix data, thereby obtaining the ultrasonic image entropy value data;

[0119] Performing ultrasound image texture complexity analysis based on ultrasound image entropy value data and ultrasound image gray-level co-occurrence matrix data, thereby obtaining ultrasound image texture complexity data;

[0120] Ultrasonic image texture feature analysis is performed according to the ultrasonic image contrast data and the ultrasonic image texture complexity data, so as to obtain ultrasonic image texture feature data.

[0121] In an embodiment of the present invention, the grayscale co-occurrence matrix data of the ultrasound image provides the grayscale relationship information between each pixel and its neighboring pixels. In order to detect the intensity of the grayscale value change of the image, the grayscale co-occurrence matrix data is imported into a data processing platform (such as NumPy or Matlab in Python), and the intensity of the grayscale value change is calculated according to the grayscale difference between each pixel in the image and its neighboring pixels. The specific method is to scan the neighborhood of each grayscale value, calculate the grayscale difference between the neighboring pixels, and construct a grayscale difference map. According to the change of the grayscale value in the image, the change intensity index, such as the grayscale change amplitude, standard deviation, etc., is calculated to obtain the intensity data of the grayscale value change of the ultrasound image. This data reflects the severity of the grayscale value change in the image, and the ultrasound image contrast analysis is performed according to the grayscale value change intensity data obtained in step 1. In this step, the grayscale value change intensity data of the ultrasound image is obtained, and the image contrast is calculated based on this data. Image contrast is usually defined as the difference in brightness in the image. Commonly used calculation methods include calculating the difference between the maximum and minimum values ​​of the grayscale value, or calculating the standard deviation of the image based on statistics. In the specific operation, the grayscale difference of each area in the image is quantified, and the contrast of the image is calculated by statistical methods. For example, the energy value of the grayscale co-occurrence matrix is ​​used to represent the contrast of the image area. The higher the energy value, the stronger the contrast of the area. The obtained ultrasound image contrast data can describe the intensity of the change in brightness in the image. The ultrasound image entropy value is a measure of the amount of image information, reflecting the degree of detail and complexity in the image. According to the grayscale co-occurrence matrix data of the ultrasound image, by calculating the entropy value of the image, the uncertainty and complexity of the information in the image can be revealed. The grayscale co-occurrence matrix is ​​converted into a probability distribution, and the probability of each pixel changing the grayscale of its neighboring pixels is calculated. The entropy value of the image is calculated using the Shannon entropy formula, which is: H = -∑p(i,j)log2 p(i,j), where p(i,j) is the probability of each pair of pixel grayscale values ​​in the grayscale co-occurrence matrix. A high entropy value indicates that the image contains more information and details, and a low entropy value indicates that the image is relatively simple or uniform. After this calculation step, the ultrasound image entropy data obtained reflects the complexity and detail information of the image. Texture complexity analysis is performed based on the ultrasound image entropy data obtained in step 3 and the ultrasound image grayscale co-occurrence matrix data obtained in step 1. The specific operation is to evaluate the texture complexity of the image by combining the grayscale co-occurrence matrix data and entropy data of the image, and calculating the grayscale value difference and information entropy of different regions. Texture complexity usually refers to the diversity and detail of the texture in the image, and involves the changes in various local patterns in the image. By analyzing the changes between different gray levels, the complexity of the details and structures in the image is identified, and an evaluation of the texture complexity of the image is formed.In this process, image processing tools (such as Matlab, Python's OpenCV library) are used to divide the image area, and the grayscale value change and entropy value of each area are further calculated to obtain the ultrasound image texture complexity data. Based on the ultrasound image contrast data obtained in step 2 and the ultrasound image texture complexity data obtained in step 4, the ultrasound image texture feature analysis is performed. In this step, the texture features of the image are further analyzed by combining contrast and texture complexity. The specific method is to take contrast and texture complexity as two independent variables, merge and analyze them through mathematical modeling methods (such as weighted average, principal component analysis, etc.) to obtain the comprehensive texture feature data of the ultrasound image. These texture feature data can reflect the overall texture pattern and local changes of the ultrasound image, and provide a richer image description. In actual operation, statistical methods are used to analyze the correlation between contrast and complexity, and comprehensive features, such as the product or weighted sum of contrast and complexity, are calculated to obtain ultrasound image texture feature data.

[0122] Preferably, step S214 includes the following steps:

[0123] Calculate the gradient change of texture feature direction according to the texture feature data of the ultrasonic image to obtain the gradient change data of the texture feature direction of the image;

[0124] Estimating the edge feature of the ultrasound image based on the directional gradient change data of the image texture feature, thereby obtaining the edge feature data of the ultrasound image;

[0125] Acquiring boundary features of the lesion region according to the texture feature data of the ultrasound image and the edge feature data of the ultrasound image to obtain boundary feature data of the lesion region of the ultrasound image;

[0126] Performing geometric feature analysis on the lesion region boundary feature data and the ultrasound image texture feature data to obtain geometric shape feature data of the lesion region;

[0127] The pain level of internal body signs is assessed according to the internal body sign data of the patient in the medical record, thereby obtaining the pain level data of the internal body signs of the patient in the medical record;

[0128] Based on the patient's physical sign pain level data and the lesion area geometric shape feature data in the medical record, the patient's symptom manifestation analysis is performed to obtain the patient's symptom manifestation data.

[0129] In an embodiment of the present invention, the calculation of the directional gradient change of the image texture feature is based on the texture feature data of the ultrasound image, and is intended to evaluate the direction and intensity of the texture change in different regions of the image. In this step, the texture feature data of the ultrasound image, such as the grayscale co-occurrence matrix, entropy value and contrast of the image, are extracted, and then the grayscale change of each pixel in the image is calculated using an image gradient operator (such as a Sobel operator). Next, the change amplitude of the texture feature in different directions is calculated to obtain the directional gradient change data of the image texture feature. This process can reveal the directionality and spatial distribution of the texture change in the image by combining the texture feature data with the gradient calculation of the image, thereby providing basic data for further image analysis. Based on the directional gradient change data of the image texture feature, the edge feature of the ultrasound image is estimated. The edge feature refers to the area in the image where the grayscale or texture changes most dramatically. These areas usually represent the boundaries and structures in the image. The directional gradient change data of the image texture feature obtained in step 1 is processed to identify the area where the texture feature changes dramatically. Commonly used methods are edge detection algorithms, such as the Canny edge detection algorithm or the Laplace operator. By calculating the amplitude of the grayscale or texture change in each area of ​​the image, the boundary of the edge feature is determined, and the edge feature data of the ultrasound image is obtained. This data reflects the edge information of each area in the image. The boundary feature acquisition of the lesion area is carried out by combining the ultrasound image texture feature data and the ultrasound image edge feature data. In this step, based on the texture feature data of the ultrasound image, such as grayscale co-occurrence matrix, contrast, texture complexity, etc., the potential lesion area is identified. Then, the specific boundary of the lesion area is determined in combination with the edge feature data of the ultrasound image. In this process, image segmentation technology is used. Common methods include threshold segmentation method, region growing method or edge-based segmentation algorithm. By analyzing the texture features and edge information in the image, the boundary of the lesion area is accurately extracted. The obtained lesion area boundary feature data describes the morphology and position of the lesion area in detail. The geometric feature analysis of the lesion area is a process of further analyzing the lesion area boundary feature data extracted in step 3. In this step, based on the boundary feature data of the lesion area, the geometric shape features of the lesion area, such as area, perimeter, shape index, shape complexity, etc., are calculated. The specific operation is to obtain shape-related geometric data by calculating the contour length and area of ​​the boundary of the lesion area. Then, combined with the image texture feature data, the morphological changes of the lesion area are further analyzed to evaluate the structural complexity and heterogeneity of the lesion area. For example, by calculating the aspect ratio, circularity and other indicators of the lesion area, the geometric morphology of the lesion area is judged, and the geometric shape feature data of the lesion area is obtained. The in vivo physical sign pain level assessment is to judge the patient's pain level by analyzing the physical sign data of the patient in the medical record. In this step, the in vivo physical sign data of the patient in the medical record is collected, including physiological parameters (such as heart rate, blood pressure, etc.) and related signs (such as respiratory rate, body temperature, etc.).By analyzing these physical sign data, the patient's pain level is assessed using medical clinical standardized scoring methods (such as visual analog scale, digital scoring method, etc.). For some patients, the characteristics of the lesion area in the ultrasound image are combined to analyze whether the lesion is associated with pain, thereby providing a more comprehensive basis for pain assessment. After the assessment is completed, the in vivo physical sign pain level data obtained is combined with the in vivo physical sign pain level data of the medical history patient obtained in step 5 and the lesion area geometric shape feature data obtained in step 4 to analyze the patient's symptom manifestation. By analyzing the patient's in vivo physical sign pain level data and combining the geometric characteristics of the lesion area, such as morphology, size, complexity, etc., the association between the lesion area and the patient's pain level is identified. For example, if the lesion area is large or complex in shape, more severe pain symptoms are associated. By quantitatively analyzing the geometric data of the lesion area and the in vivo physical sign pain data, the symptom manifestation data of historical patients is obtained.

[0130] Preferably, step S31 includes the following steps:

[0131] Step S31: extracting entity relationships based on historical disease treatment related data and disease manifestation related data to obtain disease entity relationship data;

[0132] Step S32: constructing an emergency field knowledge graph based on the disease entity relationship data to obtain emergency field knowledge graph data;

[0133] Step S33: Estimate the disease type in the emergency field based on the patient's historical electronic health record desensitized data according to the emergency field knowledge graph data to obtain emergency field disease type estimation data;

[0134] Step S34: Estimate the research plan in the emergency field based on the disease type estimation data in the emergency field and the knowledge graph data in the emergency field to obtain the research plan data in the emergency field.

[0135] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0136] Step S31: extracting entity relationships based on historical disease treatment related data and disease manifestation related data to obtain disease entity relationship data;

[0137] In an embodiment of the present invention, when entity relationships are extracted based on historical disease treatment related data and disease manifestation related data, the two types of related data are structured preprocessed and the data is represented in a standardized format, for example, in the form of triples (entity 1, relationship, entity 2). For entities in the historical disease treatment related data, core elements including disease name, lesion location, and related behavior description are extracted through a named entity recognition algorithm (NER); for disease manifestation related data, symptoms, occurrence sites, and descriptive words related to the disease are extracted. Subsequently, a rule-based method or a pattern matching method is used to analyze the relationships between different entities, for example, by analyzing the context and co-occurrence word frequency to determine the causal relationship or correlation between diseases. Finally, all extracted entities and their relationships are stored as structured disease entity relationship data.

[0138] Step S32: constructing an emergency field knowledge graph based on the disease entity relationship data to obtain emergency field knowledge graph data;

[0139] In an embodiment of the present invention, based on the disease entity relationship data obtained in step S31, a graph database technology is used to construct an emergency field knowledge graph. In a specific implementation, the ontology of the knowledge graph is defined, including entity types (such as diseases, lesion sites, symptoms) and relationship types (such as causal relationships, concomitant relationships). Then, based on the defined ontology architecture, the disease entity relationship data is entered into the graph database one by one, for example, using the Cypher query language of the Neo4j database to enter triple data to form a knowledge graph structure represented by a directed graph. During the construction process, it is necessary to deduplicate the duplicate items of the data and verify the uniqueness of the entities in the graph. After completion, the knowledge graph is displayed and verified through visualization tools to ensure data integrity and relationship correctness, thereby generating emergency field knowledge graph data.

[0140] Step S33: Estimate the disease type in the emergency field based on the patient's historical electronic health record desensitized data according to the emergency field knowledge graph data to obtain emergency field disease type estimation data;

[0141] In an embodiment of the present invention, when estimating the disease type in the emergency field based on the patient's historical electronic health record desensitized data according to the emergency field knowledge graph data, the patient's historical electronic health record desensitized data is segmented to extract key information related to the disease, such as symptom descriptions, medical history records, laboratory test results, etc. Then, the extracted information is matched with the entities and relationships in the knowledge graph, and the disease type related to the patient's symptoms and medical history is searched based on the path analysis function of the graph. For example, if the patient's symptoms have a high degree of match with a disease path in the knowledge graph, the disease is marked as a potential disease type. At the same time, the estimation priority is calculated based on the frequency and association strength of the symptoms in the graph, and finally the disease type estimation data in the emergency field is obtained.

[0142] Step S34: Estimate the research plan in the emergency field based on the disease type estimation data in the emergency field and the knowledge graph data in the emergency field to obtain the research plan data in the emergency field.

[0143] In an embodiment of the present invention, when estimating the research plan in the emergency field based on the disease type estimation data in the emergency field and the knowledge graph data in the emergency field, a knowledge graph subgraph related to the disease type estimation data is extracted, which contains all the path and node information related to the disease, such as common symptoms, susceptible populations and related behavioral characteristics. Then, the research plan parameters are designed based on the information in the knowledge graph subgraph, such as selecting relevant experimental variables and analysis objects. In specific implementation, the causal relationship and associated data in the graph are analyzed through the logical reasoning module to generate a research plan related to the target disease type. Taking the research plan of acute symptoms as an example, the key factors affecting the onset of the disease and their priorities are inferred according to the graph, so as to output a research plan including research objectives, variable selection and research data requirements, and form the research plan data for the emergency field.

[0144] Preferably, step S4 comprises the following steps:

[0145] Step S41: conducting emergency field research risk assessment according to emergency field research program data to obtain emergency field research risk data;

[0146] Step S42: performing emergency field knowledge reasoning on the symptom research associated data and the symptom manifestation associated data according to the emergency field research risk data to obtain emergency field knowledge reasoning data;

[0147] Step S43: Optimize the knowledge graph data in the emergency field according to the emergency field research risk data and the emergency field knowledge reasoning data to obtain the emergency field knowledge graph optimized data.

[0148] In an embodiment of the present invention, when conducting an emergency field research risk assessment based on the emergency field research program data, the emergency field research program data is parsed, key elements such as research objectives, variable settings, and expected experimental conditions are extracted, and these elements are stored in a structured manner. Subsequently, data traceability analysis is performed on each variable and condition to evaluate the uncertainty brought by the variable. For example, for an experimental variable involved in the program, the integrity, accuracy, and consistency of its source data are analyzed, and the data missing rate or deviation is quantitatively calculated. Next, the correlation and conflict between multiple variables are analyzed, and a variable relationship diagram is generated using a graph theory method to determine the risk node. For different experimental conditions, a risk matrix is ​​used for evaluation, and the risk level is divided into three levels: low, medium, and high, and the range of experimental results affected by high-risk conditions is marked. Finally, the emergency field research risk data is output based on the above evaluation, including a list of risk factors and their potential impact analysis results on the research program. When performing emergency field knowledge reasoning on the symptom research association data and symptom manifestation association data according to the emergency field research risk data, the risk factors in the research risk data are cross-mapped with the symptom research association data and symptom manifestation association data to screen the research topics and symptom characteristic data related to the risk factors. Subsequently, a reasoning logic rule base is constructed, such as analyzing the potential impact path between risk factors and symptom characteristics based on causal reasoning rules. During the reasoning process, the input data is iteratively deduced using a rule-based reasoning engine to output the knowledge path most relevant to the target risk factor. For example, for a certain high-risk experimental condition, the symptom error range and the corresponding symptom association path caused by it are analyzed. The accuracy of the reasoning output is verified by comparing the consistency of the reasoning results with the original data. Finally, the reasoning results are integrated into structured emergency field knowledge reasoning data, including the impact path of risk factors and the update information of symptom association data involved in the reasoning. When optimizing the knowledge graph of the emergency field knowledge graph data according to the emergency field research risk data and the emergency field knowledge reasoning data, the key nodes and edges involved in the research risk data and knowledge reasoning data are extracted to construct the optimization candidate subgraph. For high-risk nodes and redundant edges in the knowledge graph, a graph structure analysis algorithm is used to calculate their centrality and importance, and to identify nodes and edges that need to be adjusted or removed. For example, for redundant nodes introduced in the knowledge graph due to redundant reasoning, the strength of their connection with the nodes is analyzed, and irrelevant nodes with weak connections are deleted. Next, the newly added associations in the reasoning data are verified and merged with the existing knowledge graph through similarity analysis to ensure that the optimized graph has logical coherence and data consistency. After the optimization is completed, the updated knowledge graph is displayed using visualization tools to confirm the rationality of its structure. Finally, the optimized data of the knowledge graph in the emergency field is output, including detailed information on the optimized nodes, edges and their attributes.According to the output emergency field knowledge graph optimization data for deep data extraction, by extracting key entities and relationship data related to emergency, comprehensive and targeted related information is obtained. In the data extraction process, multi-level data analysis methods are applied, including but not limited to statistical analysis, correlation analysis and trend analysis, to comprehensively interpret the data. After completing the data analysis, the emergency medicine professional medical quality control indicators are obtained, and the emergency medicine professional medical quality control indicators are combined with specific training content to design more targeted and practical training materials. For example, according to the information in the knowledge graph optimization data, the processing procedures for common diseases in the emergency field or important knowledge points in related research fields can be refined, and a more scientific and reasonable workflow can be formulated based on the related data displayed in the knowledge graph.

[0149] Preferably, step S41 includes the following steps:

[0150] Step S411: performing a central venous catheter usage status analysis based on the emergency field research program data to obtain central venous catheter usage status data;

[0151] Step S412: performing statistics on the usage time of the central venous catheter according to the central venous catheter usage status data to obtain the central venous catheter usage time data;

[0152] Step S413: performing a central venous catheter thrombosis prediction on the central venous catheter usage time data to obtain central venous catheter thrombosis data;

[0153] Step S414: Calculate the probability of bacterial infection of the central venous catheter according to the central venous catheter usage time data and the central venous catheter thrombosis generation data to obtain the central venous catheter bacterial infection probability data;

[0154] Step S415: performing emergency field complication risk inference based on the central venous catheter bacterial infection probability data and the central venous catheter thrombosis generation data to obtain emergency field complication risk data;

[0155] Step S416: Conduct emergency field research risk assessment based on emergency field complication risk data and central venous catheter bacterial infection probability data to obtain emergency field research risk data.

[0156] In an embodiment of the present invention, emergency research program data is obtained, which contains specific information on the use of central venous catheters, including basic information of patients, frequency of catheter use, physical signs during catheter use, and clinical observation data. By analyzing these data, the use of central venous catheters of various types of patients is specifically analyzed. This analysis process uses data mining techniques, such as statistical analysis and pattern recognition, to extract risk points and abnormal situations with high frequency of use. Tools such as Pandas in Python are used to clean and organize data, remove irrelevant data or missing values, and obtain central venous catheter use status data. The data includes the use of catheters for each patient, records the time period, duration, and any abnormal manifestations of each use, such as catheter blockage or displacement, to form a structured data set. By statistically analyzing the use time information involved in the central venous catheter use status data, the use time data of the central venous catheter is obtained, the start time and end time of each central venous catheter use are obtained, and the specific duration of each use is calculated. Then, by summarizing the data of all patients, the cumulative use time of the central venous catheter of each patient is counted, and the use time data of each patient is obtained. In this process, tools such as Numpy and Pandas in Excel or Python are used for data processing and analysis, and the time difference between each use is calculated and summed or averaged to form structured usage time data. The central venous catheter usage time data is used to estimate the thrombosis risk of the central venous catheter. The association pattern between the catheter usage time and thrombosis is extracted by retrospective analysis of historical data. The usage time is matched with the probability of thrombosis using statistical methods (such as regression analysis) or correlations obtained in medical research. For each patient, the probability of thrombosis is calculated based on the usage time. This process does not require complex machine learning models, but relies on statistical data and empirical formulas in existing literature, combined with historical patient data for calculation. Through this process, central venous catheter thrombosis data is obtained, and the probability of bacterial infection is calculated based on the central venous catheter usage time data and thrombosis data. Combined with medical research and literature, the relationship between central venous catheter usage time and thrombosis is clarified, and whether thrombosis increases the risk of bacterial infection is further studied. Using a logistic regression model or a simple linear relationship, the duration of central venous catheter use and thrombosis data are used as input to calculate the probability of bacterial infection. In this process, referring to existing clinical studies, a computational model based on statistical analysis is established to clarify the probability of bacterial infection under different degrees of thrombosis. These steps are modeled and analyzed by Scikit-learn in Python to obtain the bacterial infection probability data for each patient. The central venous catheter bacterial infection probability data and thrombosis data are used to infer the risk of complications.By combining the probabilities of thrombosis and bacterial infection through statistical methods (such as Bayesian reasoning or conditional probability calculation), the risk of complications in the emergency field can be inferred. For example, if the probability of bacterial infection is high, it will lead to the occurrence of serious complications such as septic shock. By integrating these data, a complication risk inference model is constructed to evaluate the probability of complications in patients during the emergency process. During the reasoning process, database management tools (such as SQL) are used to organize and query data, and the complication risk data of each patient is obtained by merging information from different data sources. This step conducts emergency research risk assessment based on emergency field complication risk data and bacterial infection probability data, evaluates the influencing factors of complication risk data, and calculates the overall emergency research risk in combination with bacterial infection probability data. This process uses a variety of methods such as weighted average method and hierarchical analysis method (AHP) to quantify different types of risks and conduct a comprehensive assessment in combination with historical emergency data. By analyzing the probability of bacterial infection and complication risk of each emergency case, the risks that occur during the research process are evaluated. For example, by matching different complication risks with the characteristics of the research subjects, emergency research risk data is obtained.

[0157] The present invention also provides an emergency field knowledge extraction system based on multi-source data, which is used to execute the emergency field knowledge extraction method based on multi-source data as described above. The emergency field knowledge extraction system based on multi-source data includes:

[0158] Patient historical ultrasound image acquisition module: obtain patient historical electronic health record data; perform record desensitization processing based on patient historical electronic health record data to obtain patient historical electronic health record desensitized data; collect patient vital signs from medical records based on patient historical electronic health record desensitized data to obtain patient vital signs data from medical records; obtain patient historical ultrasound images based on patient historical electronic health record desensitized data to obtain patient historical ultrasound image data;

[0159] Symptom treatment association analysis module: based on the patient's historical ultrasound image data and the patient's physical sign data in the medical record, the symptom manifestation association is estimated to obtain the symptom manifestation association data; based on the symptom manifestation association data and the patient's historical electronic health record desensitized data, the historical symptom treatment association analysis is performed to obtain the historical symptom treatment association data;

[0160] Research plan estimation module: construct the knowledge graph of the emergency field based on the historical disease treatment related data and the disease manifestation related data to obtain the knowledge graph data of the emergency field; estimate the research plan of the emergency field based on the knowledge graph data of the emergency field to obtain the research plan data of the emergency field;

[0161] Knowledge graph optimization module: conduct research risk assessment in the emergency field based on the research plan data in the emergency field to obtain research risk data in the emergency field; optimize the knowledge graph data in the emergency field based on the research risk data in the emergency field to obtain optimized knowledge graph data in the emergency field.

[0162] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for extracting emergency field knowledge based on multi-source data, characterized in that: The following steps are involved: Step S1: Obtain patient historical electronic health record data; Perform record desensitization processing based on the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitized data; According to the patient's historical electronic health record desensitized data, the patient's physical signs are collected in the medical record to obtain the patient's physical signs data; according to the patient's historical electronic health record desensitized data, the patient's historical ultrasound image is acquired to obtain the patient's historical ultrasound image data; Step S2: performing a symptom manifestation correlation estimation based on the patient's historical ultrasound image data and the patient's medical record vital sign data to obtain symptom manifestation correlation data; performing a historical symptom treatment correlation analysis based on the symptom manifestation correlation data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment correlation data; Step S3: constructing an emergency field knowledge graph based on historical disease treatment related data and disease manifestation related data to obtain emergency field knowledge graph data; estimating emergency field research plans based on the emergency field knowledge graph data to obtain emergency field research plan data; Step S4: conducting emergency field research risk assessment according to emergency field research plan data to obtain emergency field research risk data; The knowledge graph data in the emergency field is optimized according to the research risk data in the emergency field to obtain the optimized knowledge graph data in the emergency field.

2. The method for extracting emergency field knowledge based on multi-source data according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Obtaining patient historical electronic health record data; Step S12: performing record desensitization processing according to the patient's historical electronic health record data to obtain the patient's historical electronic health record desensitization data; Step S13: Collecting medical record patient vital signs based on the patient's historical electronic health record desensitized data to obtain medical record patient vital sign data; Step S14: Acquire the patient's historical ultrasound image based on the patient's historical electronic health record desensitized data to obtain the patient's historical ultrasound image data.

3. The method for extracting emergency field knowledge based on multi-source data according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: Analyze the patient's symptom manifestations according to the patient's historical ultrasound image data and the patient's physical sign data in the medical history to obtain the patient's symptom manifestation data; Step S22: collecting historical patient conditions based on the desensitized data of the patient's historical electronic health records to obtain historical patient condition data; Step S23: estimating the correlation between symptom manifestations based on historical patient condition data and patient symptom manifestation data to obtain symptom manifestation correlation data; Step S24: Perform historical symptom treatment association analysis based on the symptom manifestation association data and the patient's historical electronic health record desensitized data to obtain historical symptom treatment association data.

4. The method for extracting emergency field knowledge based on multi-source data according to claim 3 is characterized in that: Step S21 includes the following steps: Step S211: constructing a gray-level co-occurrence matrix according to the patient's historical ultrasound image data to obtain ultrasound image gray-level co-occurrence matrix data; Step S212: performing ultrasonic image texture feature recognition on the ultrasonic image gray level co-occurrence matrix data to obtain ultrasonic image texture feature data; Step S213: performing a physical sign detection on the patient in the medical record based on the physical sign data of the patient in the medical record to obtain the physical sign data of the patient in the medical record; Step S214: Analyze the patient's symptom manifestation based on the ultrasound image texture feature data and the patient's internal physical sign data in the medical record to obtain the patient's symptom manifestation data.

5. The method for extracting emergency field knowledge based on multi-source data according to claim 4 is characterized in that: Step S212 includes the following steps: Performing image grayscale value change intensity detection on the ultrasound image grayscale co-occurrence matrix data, thereby obtaining ultrasound image grayscale value change intensity data; Performing ultrasound image contrast analysis according to the ultrasound image gray value change intensity data, thereby obtaining ultrasound image contrast data; Acquiring the ultrasonic image entropy value according to the ultrasonic image gray level co-occurrence matrix data, thereby obtaining the ultrasonic image entropy value data; Performing ultrasound image texture complexity analysis based on ultrasound image entropy value data and ultrasound image gray-level co-occurrence matrix data, thereby obtaining ultrasound image texture complexity data; Ultrasonic image texture feature analysis is performed according to the ultrasonic image contrast data and the ultrasonic image texture complexity data, so as to obtain ultrasonic image texture feature data.

6. The method for extracting emergency field knowledge based on multi-source data according to claim 4 is characterized in that: Step S214 includes the following steps: Calculate the gradient change of texture feature direction according to the texture feature data of the ultrasound image to obtain the gradient change data of the texture feature direction of the image; Estimating the edge feature of the ultrasound image based on the directional gradient change data of the image texture feature, thereby obtaining the edge feature data of the ultrasound image; Acquiring boundary features of the lesion region according to the texture feature data of the ultrasound image and the edge feature data of the ultrasound image to obtain boundary feature data of the lesion region of the ultrasound image; Performing geometric feature analysis on the lesion region boundary feature data and the ultrasound image texture feature data to obtain geometric shape feature data of the lesion region; The pain level of internal body signs is assessed according to the internal body sign data of the patient in the medical record, thereby obtaining the pain level data of the internal body signs of the patient in the medical record; Based on the patient's physical sign pain level data and the lesion area geometric shape feature data in the medical record, the patient's symptom manifestation analysis is performed to obtain the patient's symptom manifestation data.

7. The method for extracting emergency field knowledge based on multi-source data according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: extracting entity relationships based on historical disease treatment related data and disease manifestation related data to obtain disease entity relationship data; Step S32: constructing an emergency field knowledge graph based on the disease entity relationship data to obtain emergency field knowledge graph data; Step S33: Estimate the disease type in the emergency field based on the patient's historical electronic health record desensitized data according to the emergency field knowledge graph data to obtain emergency field disease type estimation data; Step S34: Estimate the research plan in the emergency field based on the disease type estimation data in the emergency field and the knowledge graph data in the emergency field to obtain the research plan data in the emergency field.

8. The method for extracting emergency field knowledge based on multi-source data according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: conducting emergency field research risk assessment according to emergency field research program data to obtain emergency field research risk data; Step S42: performing emergency field knowledge reasoning on the symptom research associated data and the symptom manifestation associated data according to the emergency field research risk data to obtain emergency field knowledge reasoning data; Step S43: Optimize the knowledge graph data in the emergency field according to the emergency field research risk data and the emergency field knowledge reasoning data to obtain the emergency field knowledge graph optimized data.

9. The method for extracting emergency field knowledge based on multi-source data according to claim 8, characterized in that: Step S41 includes the following steps: Step S411: performing a central venous catheter usage status analysis based on the emergency field research program data to obtain central venous catheter usage status data; Step S412: performing statistics on the usage time of the central venous catheter according to the central venous catheter usage status data to obtain the central venous catheter usage time data; Step S413: performing a central venous catheter thrombosis prediction on the central venous catheter usage time data to obtain central venous catheter thrombosis data; Step S414: Calculate the probability of bacterial infection of the central venous catheter according to the central venous catheter usage time data and the central venous catheter thrombosis generation data to obtain the central venous catheter bacterial infection probability data; Step S415: performing emergency field complication risk inference based on the central venous catheter bacterial infection probability data and the central venous catheter thrombosis generation data to obtain emergency field complication risk data; Step S416: Conduct emergency field research risk assessment based on emergency field complication risk data and central venous catheter bacterial infection probability data to obtain emergency field research risk data.

10. An emergency field knowledge extraction system based on multi-source data, characterized in that: Used to execute the emergency field knowledge extraction method based on multi-source data as claimed in claim 1, the emergency field knowledge extraction system based on multi-source data includes: Patient historical ultrasound image acquisition module: obtain patient historical electronic health record data; perform record desensitization processing based on patient historical electronic health record data to obtain patient historical electronic health record desensitized data; collect patient vital signs from medical records based on patient historical electronic health record desensitized data to obtain patient vital signs data from medical records; obtain patient historical ultrasound images based on patient historical electronic health record desensitized data to obtain patient historical ultrasound image data; Symptom treatment association analysis module: based on the patient's historical ultrasound image data and the patient's physical sign data in the medical record, the symptom manifestation association is estimated to obtain the symptom manifestation association data; based on the symptom manifestation association data and the patient's historical electronic health record desensitized data, the historical symptom treatment association analysis is performed to obtain the historical symptom treatment association data; Research plan estimation module: construct the knowledge graph of the emergency field based on the historical disease treatment related data and the disease manifestation related data to obtain the knowledge graph data of the emergency field; estimate the research plan of the emergency field based on the knowledge graph data of the emergency field to obtain the research plan data of the emergency field; Knowledge graph optimization module: conduct research risk assessment in the emergency field based on the research plan data in the emergency field to obtain research risk data in the emergency field; optimize the knowledge graph data in the emergency field based on the research risk data in the emergency field to obtain optimized knowledge graph data in the emergency field.

Images