Medical etiological analysis method and system based on natural language processing and storage medium

Abstract

The invention discloses a medical cause analysis method and system based on natural language processing and a storage medium. The method comprises the steps of obtaining target medical data to analyze and determine medical record data; data processing is conducted on the medical record data to obtain basic judgment data, and different types of medical record data are different in processing process; the basic judgment data is input into a cause analytic hierarchy analysis model for analysis to obtain a cause identification result, and the cause analytic hierarchy analysis model comprises a data mining layer, a clustering division layer, a map auxiliary layer and a verification evaluation layer; and performing visual display based on the cause identification result, the display modes at least including graphic, image and report display. According to the method, the dependence on a dictionary in the traditional technology is reduced, the recognition capability of new terms is enhanced, and the method is easier to popularize and apply in clinical practice by efficiently processing various data sources and improving the integration capability.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to a medical etiology analysis method, system and storage medium based on natural language processing. Background Art

[0002] In the field of modern medical etiology analysis, natural language processing technology is widely used to improve the processing efficiency of diagnostic information. However, despite certain technological progress, existing methods still face many challenges in practical applications, which affects their effectiveness and widespread application.

[0003] Current technologies mainly rely on methods such as dictionary construction, information matching strategies, and deep learning models, but they still have the following significant shortcomings:

[0004] (1) Strong dependence on dictionaries. The excessive dependence on dictionaries is a significant defect of current technology. Existing methods are highly dependent on pre-established dictionaries. The quality and comprehensiveness of dictionaries directly affect the accuracy of recognition. Due to the continuous updating of medical terminology, dictionaries are difficult to update in a timely manner and cover comprehensively, resulting in the inability to recognize newly emerging disease terms or common names, affecting the accuracy and completeness of the analysis results. In addition, dictionaries are unable to cope with complex contexts or polysemous words, which easily leads to a decrease in the accuracy of information understanding and cannot meet the needs of precise analysis in real medical scenarios.

[0005] (2) Inadequate processing of complex data. Existing technologies often face challenges of difficult integration and complex processing when processing multiple data sources (such as paper medical records, recordings, images, etc.). In particular, in the identification and conversion of unstructured data, existing methods may require a lot of manual intervention, affecting overall efficiency and accuracy. Traditional solutions usually rely on manual data entry and simple rule matching. This method is inefficient when faced with diverse data sources, is prone to errors, and has difficulty in effectively processing unstructured data.

[0006] (3) Insufficient model interpretability. Although many network models in existing technologies have shown strong capabilities in etiology identification, their “black box” characteristics make the results difficult to interpret and trust. In addition, in a dynamically changing medical data environment, traditional hierarchical clustering methods may lack sufficient adaptability and are difficult to quickly adjust to reflect the latest medical knowledge, further exacerbating the problem of difficult-to-interpret results. Summary of the invention

[0007] The purpose of the present invention is to provide a medical etiology analysis method, system and storage medium based on natural language processing, which reduces the dependence on dictionaries in traditional technologies, enhances the ability to recognize new terms, and is easier to promote and apply in clinical practice through efficient processing of various data sources and improved integration capabilities.

[0008] The first aspect of the present invention provides a medical etiology analysis method based on natural language processing, comprising the following steps:

[0009] Obtain target medical data to parse and determine medical record data;

[0010] Processing the medical record data to obtain basic judgment data, wherein different types of medical record data have different processing processes;

[0011] Inputting the basic judgment data into a cause-of-disease hierarchical analysis model for analysis to obtain a cause-of-disease identification result, wherein the cause-of-disease hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer, and a verification and evaluation layer;

[0012] Visual display is performed based on the cause identification result, wherein the display method at least includes graphics, images and report display.

[0013] In this solution, the target medical data is obtained, including:

[0014] Acquiring patient data based on a user interaction interface, wherein the patient data includes at least a patient identifier and a case number;

[0015] Acquire image data and perform optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records;

[0016] Obtaining the medical consultation recording data for speech-to-text processing, and adding the text processing result to the text data;

[0017] The target medical data is obtained based on the patient data and the text data.

[0018] In this solution, parsing the target medical data to determine the medical record data specifically includes:

[0019] The types of the medical record data include time series attributes and structured text attributes, where:

[0020] Parsing text keywords and time information in the target medical data to obtain the time intervals between different medical services to obtain a consultation cycle, wherein the consultation cycle is a time series attribute;

[0021] Based on the consultation cycle, the minimum data analysis unit in a single consultation process is extracted and the medical entity is obtained by using the named entity recognition technology, wherein the medical entity is a structured text attribute.

[0022] In this solution, the data processing of the medical record data to obtain basic judgment data specifically includes:

[0023] Use regular expressions to perform location-oriented and keyword-oriented data processing on the medical record data, wherein:

[0024] For the data with time series attributes, time window alignment is applied for normalization;

[0025] Perform outlier detection and missing value filling on structured text attribute data, including filling with mean or median.

[0026] In this solution, the etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer, wherein the data mining layer is used to mine potential factors related to the disease; the clustering layer is used to obtain the degree of correlation between different etiology characteristics and potential factors; the graph auxiliary layer is used to construct a professional knowledge graph using knowledge graph technology combined with the degree of correlation, and the verification and evaluation layer is used to verify and evaluate the reliability and effectiveness of the identification results.

[0027] In this solution, the user interaction interface can be used for display based on the visualization of the cause identification results, wherein differentiated display is performed based on user attributes. When the user attributes are defined as professional medical personnel, report visualization is performed, and when the user attributes are defined as patients and their families, graphic visualization is performed, specifically including bar charts and line graphs.

[0028] The second aspect of the present invention further provides a medical etiology analysis system based on natural language processing, comprising a memory and a processor, wherein the memory comprises a medical etiology analysis method program based on natural language processing, and when the medical etiology analysis method program based on natural language processing is executed by the processor, the following steps are implemented:

[0029] Obtain target medical data to parse and determine medical record data;

[0030] Processing the medical record data to obtain basic judgment data, wherein different types of medical record data have different processing processes;

[0031] Inputting the basic judgment data into a cause-of-disease hierarchical analysis model for analysis to obtain a cause-of-disease identification result, wherein the cause-of-disease hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer, and a verification and evaluation layer;

[0032] Visual display is performed based on the cause identification result, wherein the display method at least includes graphics, images and report display.

[0033] In this solution, the target medical data is obtained, including:

[0034] Acquiring patient data based on a user interaction interface, wherein the patient data includes at least a patient identifier and a case number;

[0035] Acquire image data and perform optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records;

[0036] Obtaining the medical consultation recording data for speech-to-text processing, and adding the text processing result to the text data;

[0037] The target medical data is obtained based on the patient data and the text data.

[0038] In this solution, parsing the target medical data to determine the medical record data specifically includes:

[0039] The types of the medical record data include time series attributes and structured text attributes, where:

[0040] Parsing text keywords and time information in the target medical data to obtain the time intervals between different medical services to obtain a consultation cycle, wherein the consultation cycle is a time series attribute;

[0041] Based on the consultation cycle, the minimum data analysis unit in a single consultation process is extracted and the medical entity is obtained by using the named entity recognition technology, wherein the medical entity is a structured text attribute.

[0042] In this solution, the data processing of the medical record data to obtain basic judgment data specifically includes:

[0043] Use regular expressions to perform location-oriented and keyword-oriented data processing on the medical record data, wherein:

[0044] For the data with time series attributes, time window alignment is applied for normalization;

[0045] Perform outlier detection and missing value filling on structured text attribute data, including filling with mean or median.

[0046] In this solution, the etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer, wherein the data mining layer is used to mine potential factors related to the disease; the clustering layer is used to obtain the degree of correlation between different etiology characteristics and potential factors; the graph auxiliary layer is used to construct a professional knowledge graph using knowledge graph technology combined with the degree of correlation, and the verification and evaluation layer is used to verify and evaluate the reliability and effectiveness of the identification results.

[0047] In this solution, the user interaction interface can be used for display based on the visualization of the cause identification results, wherein differentiated display is performed based on user attributes. When the user attributes are defined as professional medical personnel, report visualization is performed, and when the user attributes are defined as patients and their families, graphic visualization is performed, specifically including bar charts and line graphs.

[0048] The third aspect of the present invention provides a computer-readable storage medium, which includes a machine-readable program for a medical etiology analysis method based on natural language processing. When the program for the medical etiology analysis method based on natural language processing is executed by a processor, the steps of a medical etiology analysis method based on natural language processing as described in any one of the above items are implemented.

[0049] The present invention discloses a medical etiology analysis method, system and storage medium based on natural language processing, which provides an efficient, accurate and intelligent innovative solution for etiology identification and analysis in the field of medical diagnosis, and has the following beneficial effects:

[0050] (1) Reduce dictionary dependence and enhance the ability to identify new terms: The present invention uses natural language processing technologies, such as named entity recognition and regular expression technology, combined with deep learning models in the data regularization module and the data preprocessing module. Through these technical means, it is possible to automatically parse and understand keywords, time information and newly emerging medical terms in medical texts without over-reliance on pre-built dictionaries. This method reduces the dependence on dictionaries, making the system more adaptable, able to timely identify and understand new disease terms and polysemous words, and ensure the accuracy and completeness of the analysis results. It can also better handle complex contexts, improve the ability to understand and analyze information, and better meet the needs of accurate analysis in real medical scenarios.

[0051] (2) Efficiently process multiple data sources and improve data integration capabilities: The data acquisition module and data preprocessing module of the present invention use advanced image recognition technology, optical character recognition algorithm based on deep learning, and speech-to-text technology. Through these technologies, seamless integration and efficient processing of multiple data sources such as paper medical records, photo screenshots, and medical interview recordings are achieved. Among them, it can automatically extract information from various forms of medical data quickly and accurately, greatly reducing manual intervention and improving overall efficiency and accuracy. In particular, when processing unstructured data, it can efficiently identify and convert it to ensure the integrity and reliability of the data. This capability meets the diverse data processing needs in modern medicine and improves the efficiency and accuracy of data processing.

[0052] (3) Improve model interpretability and enhance model adaptability and trust: The potential cause hierarchical identification and analysis module of the present invention introduces attention mechanism, hierarchical clustering algorithm and knowledge graph technology, and constructs data deep mining layer, hierarchical clustering division layer, knowledge graph auxiliary layer and verification evaluation optimization layer. Through the combination of these technologies, the interpretability and adaptability of the model are improved. Among them, through the attention mechanism, the model can focus on key medical information and improve the accuracy of cause analysis. The introduction of knowledge graph enables the model to combine rich medical knowledge for cause analysis, enhances the credibility and interpretability of the results, and the verification evaluation optimization layer ensures the adaptability of the model in a dynamically changing medical data environment, allowing the model to be quickly adjusted to reflect the latest medical knowledge. These improvements make the model easier for medical professionals to understand and trust, which helps to promote its application in clinical practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flowchart of a medical etiology analysis method based on natural language processing according to the present invention is shown;

[0054] Figure 2 A flowchart of obtaining target medical data in a medical etiology analysis method based on natural language processing according to the present invention is shown;

[0055] Figure 3 A schematic diagram showing the steps of parsing medical text in a medical etiology analysis method based on natural language processing of the present invention is shown;

[0056] Figure 4 A schematic diagram showing the steps of parsing data in a medical etiology analysis method based on natural language processing according to the present invention is shown;

[0057] Figure 5 A schematic diagram showing the composition of a cause-of-medical hierarchical analysis model in a medical cause-of-medical analysis method based on natural language processing according to the present invention is shown;

[0058] Figure 6 A block diagram of a medical etiology analysis system based on natural language processing according to the present invention is shown. DETAILED DESCRIPTION

[0059] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0060] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0061] Figure 1 A flowchart of a medical etiology analysis method based on natural language processing of the present application is shown.

[0062] like Figure 1 As shown, the present application discloses a medical etiology analysis method based on natural language processing, comprising the following steps:

[0063] S102, obtaining target medical data to parse and determine medical record data;

[0064] S104, processing the medical record data to obtain basic judgment data, wherein different types of medical record data have different processing processes;

[0065] S106, inputting the basic judgment data into the etiology hierarchical analysis model for analysis to obtain the etiology identification result;

[0066] S108, performing visual display based on the cause identification result, wherein the display method at least includes graphics, images and report display.

[0067] It should be noted that in this embodiment, when analyzing the medical causes of diseases, it is first necessary to clarify the medical record data of the medical treatment before targeted data processing can be performed to analyze the cause identification results. Therefore, the medical record data must first be determined. Specifically, the target medical data is obtained to analyze and determine the medical record data. Accordingly, the target medical data includes patient data and text data, which will be described in detail in the subsequent instructions.

[0068] Furthermore, after obtaining the medical record data, the medical record data can be processed to obtain basic judgment data, wherein different types of medical record data have different processing procedures, wherein the types of medical record data include, for example, time series attribute data and structured text attribute data, and therefore the processing procedures corresponding to data with different attributes are different, which will also be described in detail in the subsequent instructions.

[0069] Furthermore, after obtaining the basic judgment data, it can be input into the etiology hierarchical analysis model, and processed by the model to obtain the etiology identification result. Accordingly, the etiology hierarchical analysis model includes a data mining layer, a clustering division layer, a graph auxiliary layer and a verification and evaluation layer. The processing process of each layer is also different, and the details will also be described in detail in the subsequent instructions. Among them, in order to facilitate the use of medical staff and patients, after obtaining the etiology identification result, it can be visualized based on the etiology identification result, and the display method includes at least graphics, images and report display.

[0070] According to an embodiment of the present invention, obtaining target medical data specifically includes:

[0071] Acquiring patient data based on a user interaction interface, wherein the patient data includes at least a patient identifier and a case number;

[0072] Acquire image data and perform optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records;

[0073] Obtaining the medical consultation recording data for speech-to-text processing, and adding the text processing result to the text data;

[0074] The target medical data is obtained based on the patient data and the text data.

[0075] It should be noted that, in this embodiment, Figure 2As shown, it is shown as a corresponding process step schematic diagram, wherein when applied, the identified target medical data specifically includes patient data and text data, wherein patient data is obtained based on a user interaction interface, and the patient data includes at least a patient identification and a case number. Specifically, the most cutting-edge natural language processing technology is used to achieve seamless integration with a variety of different types of data sources. Through a highly flexible and user-friendly user interaction interface, the required target medical data range, such as patient identification and case number, can be clearly defined, and the source data of the text data includes image data and consultation recording data. Specifically, image data is obtained for optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records; and consultation recording data is obtained for speech-to-text processing, and the text processing results are added to the text data. Specifically, when processing data types such as hospital paper medical records, photographed screenshots and consultation recordings, specific image recognition technology, such as optical character recognition algorithm, is used to efficiently recognize text; at the same time, for consultation recordings, advanced speech-to-text technology is used to accurately convert speech into text information, thereby ensuring that data from all sources can be fully incorporated into the analysis system.

[0076] According to an embodiment of the present invention, parsing the target medical data to determine the medical record data specifically includes:

[0077] The types of the medical record data include time series attributes and structured text attributes, where:

[0078] Parsing text keywords and time information in the target medical data to obtain the time intervals between different medical services to obtain a consultation cycle, wherein the consultation cycle is a time series attribute;

[0079] Based on the consultation cycle, the minimum data analysis unit in a single consultation process is extracted and the medical entity is obtained by using the named entity recognition technology, wherein the medical entity is a structured text attribute.

[0080] It should be noted that, in this embodiment, Figure 3 and Figure 4As shown, it is a schematic diagram of the parsing steps, wherein the medical record data includes data with time series attributes and data with structured text attributes, wherein the text keywords and time information in the target medical data are parsed to obtain the time intervals between different medical services to obtain the medical cycle, and the corresponding medical cycle is a time series attribute; and based on the medical cycle, the minimum data analysis unit in a single medical visit is extracted, and the named entity recognition technology is used to obtain the medical entity, wherein the medical entity is a structured text attribute, specifically, for the irregular interval data generated by multiple outpatient visits and hospitalizations, information reorganization and integration are performed to shape the minimum data analysis unit of a single medical visit process, and the most important medical record records are extracted, so that in terms of information fusion, the named entity recognition technology is used to identify the medical entity, and the time information extraction technology is used to determine the medical visit time sequence, to ensure data coherence and consistency, and to improve accuracy and availability.

[0081] According to an embodiment of the present invention, the data processing of the medical record data to obtain basic judgment data specifically includes:

[0082] Use regular expressions to perform location-oriented and keyword-oriented data processing on the medical record data, wherein:

[0083] For the data with time series attributes, time window alignment is applied for normalization;

[0084] Perform outlier detection and missing value filling on structured text attribute data, including filling with mean or median.

[0085] It should be noted that, in this embodiment, natural language processing technology is used to perform structured processing on medical record texts. Among them, regular expression technology is used, and according to the structural characteristics of the medical record text, position-oriented and keyword-oriented modes are selected to achieve accurate extraction and organization of information. Specifically, for data with time series attributes, time window alignment technology is specifically applied to perform normalization processing to ensure data consistency. In addition, for data with structured text attributes, outlier detection and processing are performed, and missing values ​​are filled. Specifically, mean or median filling and standardized coding processing can be used. Among them, when processing paper medical records, photo screenshots and consultation recordings, the module ensures the stability and reliability of the data, and through multiple verification and optimization algorithms, it guarantees data quality and applicability, providing a solid foundation for subsequent medical analysis and decision-making.

[0086] According to an embodiment of the present invention, the etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer, wherein the data mining layer is used to mine potential factors related to the disease; the clustering layer is used to obtain the degree of correlation between different etiology characteristics and potential factors; the graph auxiliary layer is used to construct a professional knowledge graph using knowledge graph technology combined with the degree of correlation, and the verification and evaluation layer is used to verify and evaluate the reliability and validity of the identification results.

[0087] It should be noted that, in this embodiment, the etiology hierarchical analysis model described has been trained, and the specific training steps are known technical options for those skilled in the art and will not be described in detail herein. Figure 5 As shown, the etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer. In this embodiment, the purpose of each layer will be emphasized. Specifically, the data mining layer is used to mine potential factors related to the disease. Among them, this layer, as the basis of the potential etiology hierarchical identification and analysis module, comprehensively uses advanced deep learning algorithms to conduct in-depth analysis of the regularized and pre-processed medical data. Through the constructed complex neural network model, it can efficiently and automatically learn the implicit patterns and features in the data, so as to accurately identify potential factors that may be related to the disease. The introduction of the attention mechanism focuses on key medical information, such as symptoms, test results, family medical history, etc., which not only helps to focus on key information, but also greatly improves the accuracy of etiology analysis, and lays a solid data foundation for subsequent hierarchical identification work.

[0088] Furthermore, the clustering division layer is used to obtain the degree of association between different etiological characteristics and potential factors. Based on the results of deep data mining, this layer adopts a hierarchical clustering algorithm to scientifically divide the potential causes into different levels according to different etiological characteristics and the degree of association, and accurately classifies the causes that directly lead to the disease into the first level. These causes often have a direct causal relationship and are the core factors of the disease, while those that may indirectly affect the occurrence of the disease are classified as the second level or higher. For example, some bad living habits may indirectly cause the disease by affecting the body's immune system, etc. These factors will be classified into a higher level. This hierarchical division method can provide a more comprehensive understanding of the causes of the disease and provide a strong decision-making basis for formulating personalized treatment plans.

[0089] Furthermore, the graph auxiliary layer is used to construct a professional knowledge graph using knowledge graph technology combined with the degree of association. In particular, the professional knowledge graph in the medical field is constructed by combining knowledge graph technology to clearly visualize various medical entities such as diseases, symptoms, examination items, treatment methods, and the relationships between them. In the process of identifying potential causes, by efficiently querying the knowledge graph, rich medical knowledge and valuable experience related to the current case can be quickly obtained. The knowledge graph can not only provide a wider range of medical background information, but also assist in verifying and optimizing the identification results of potential causes, further improving the accuracy and reliability of cause analysis.

[0090] Furthermore, the verification and evaluation layer is used to verify and evaluate the reliability and effectiveness of the identification results. In order to ensure the reliability and effectiveness of the potential cause identification results, a strict verification and evaluation mechanism is set up at this level. When applied, various scientific methods such as cross-validation and independent sample testing are used to repeatedly verify and optimize the identification results. At the same time, in actual application, the professional opinions and feedback of clinical experts can be combined to continuously adjust and improve the algorithm. Through a continuous optimization process, the application value of the algorithm in actual medical diagnosis is improved, providing a more efficient, accurate and intelligent innovative solution for cause identification and analysis in the field of medical diagnosis.

[0091] According to an embodiment of the present invention, the user interaction interface can be used for display based on the visualization display of the cause identification result, wherein differentiated display is performed based on user attributes. When the user attributes are defined as professional medical personnel, report visualization is performed, and when the user attributes are defined as patients and their families, graphic visualization is performed, specifically including bar charts and line graphs.

[0092] It should be noted that in this embodiment, after obtaining the result of etiology identification, it can be visualized, and can be specifically displayed using the user interaction interface. In order to meet the needs of different users, a variety of display modes are provided. For professional medical personnel, report visualization is performed, and detailed data analysis reports are specifically provided, including specific descriptions of the cause of the disease, relevant evidence support, risk assessment and other information, to help doctors deeply understand the process and results of etiology analysis, so as to formulate more accurate treatment plans. For patients and their families, graphic visualization is performed, including bar graphs and line graphs, providing concise and clear popular science displays, explaining the possible causes of the disease, treatment recommendations and precautions in plain language, and enhancing patients' understanding of the disease and self-management ability. For example, a bar graph is used to show the possible distribution of different causes, a line graph is used to show the changing trend of key indicators in the development of the disease, and a pie chart is used to show the proportion of various potential causes in the overall causes.

[0093] It is worth mentioning that the user interaction interface provided in this embodiment is also interactive, wherein the user can obtain more detailed information by clicking on a specific area in a chart or graph when using it. For example, by clicking on a column corresponding to a cause of disease in a bar chart, you can view a detailed analysis report of the cause of disease; by clicking on a specific data point in a line chart, you can understand the specific indicator value and related explanation at that time point. Furthermore, in the display process, attention is paid to the security and privacy protection of data. In the process of displaying the results, medical data confidentiality regulations are strictly observed, and encryption, authorized access and other measures are taken to ensure the patient's personal privacy and the security of medical data. At the same time, in order to facilitate the storage and sharing of data, the user interaction interface also provides a data export function, and users can export the analysis results in a specific format for viewing and further analysis on different devices and platforms.

[0094] Figure 6 A block diagram of a medical etiology analysis system based on natural language processing according to the present invention is shown.

[0095] like Figure 6 As shown, the present invention discloses a medical etiology analysis system based on natural language processing, including a memory and a processor, wherein the memory includes a medical etiology analysis method program based on natural language processing, and when the medical etiology analysis method program based on natural language processing is executed by the processor, the following steps are implemented:

[0096] Obtain target medical data to parse and determine medical record data;

[0097] Processing the medical record data to obtain basic judgment data, wherein different types of medical record data have different processing processes;

[0098] Inputting the basic judgment data into a cause-of-disease hierarchical analysis model for analysis to obtain a cause-of-disease identification result, wherein the cause-of-disease hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer, and a verification and evaluation layer;

[0099] Visual display is performed based on the cause identification result, wherein the display method at least includes graphics, images and report display.

[0100] It should be noted that in this embodiment, when analyzing the medical causes of diseases, it is first necessary to clarify the medical record data of the medical treatment before targeted data processing can be performed to analyze the cause identification results. Therefore, the medical record data must first be determined. Specifically, the target medical data is obtained to analyze and determine the medical record data. Accordingly, the target medical data includes patient data and text data, which will be described in detail in the subsequent instructions.

[0101] Furthermore, after obtaining the medical record data, the medical record data can be processed to obtain basic judgment data, wherein different types of medical record data have different processing procedures, wherein the types of medical record data include, for example, time series attribute data and structured text attribute data, and therefore the processing procedures corresponding to data with different attributes are different, which will also be described in detail in the subsequent instructions.

[0102] Furthermore, after obtaining the basic judgment data, it can be input into the etiology hierarchical analysis model, and processed by the model to obtain the etiology identification result. Accordingly, the etiology hierarchical analysis model includes a data mining layer, a clustering division layer, a graph auxiliary layer and a verification and evaluation layer. The processing process of each layer is also different, and the details will also be described in detail in the subsequent instructions. Among them, in order to facilitate the use of medical staff and patients, after obtaining the etiology identification result, it can be visualized based on the etiology identification result, and the display method includes at least graphics, images and report display.

[0103] According to an embodiment of the present invention, obtaining target medical data specifically includes:

[0104] Acquiring patient data based on a user interaction interface, wherein the patient data includes at least a patient identifier and a case number;

[0105] Acquire image data and perform optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records;

[0106] Obtaining the medical consultation recording data for speech-to-text processing, and adding the text processing result to the text data;

[0107] The target medical data is obtained based on the patient data and the text data.

[0108] It should be noted that, in this embodiment, Figure 2As shown, it is shown as a corresponding process step schematic diagram, wherein when applied, the identified target medical data specifically includes patient data and text data, wherein patient data is obtained based on a user interaction interface, and the patient data includes at least a patient identification and a case number. Specifically, the most cutting-edge natural language processing technology is used to achieve seamless integration with a variety of different types of data sources. Through a highly flexible and user-friendly user interaction interface, the required target medical data range, such as patient identification and case number, can be clearly defined, and the source data of the text data includes image data and consultation recording data. Specifically, image data is obtained for optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records; and consultation recording data is obtained for speech-to-text processing, and the text processing results are added to the text data. Specifically, when processing data types such as hospital paper medical records, photographed screenshots and consultation recordings, specific image recognition technology, such as optical character recognition algorithm, is used to efficiently recognize text; at the same time, for consultation recordings, advanced speech-to-text technology is used to accurately convert speech into text information, thereby ensuring that data from all sources can be fully incorporated into the analysis system.

[0109] According to an embodiment of the present invention, parsing the target medical data to determine the medical record data specifically includes:

[0110] The types of the medical record data include time series attributes and structured text attributes, where:

[0111] Parsing text keywords and time information in the target medical data to obtain the time intervals between different medical services to obtain a consultation cycle, wherein the consultation cycle is a time series attribute;

[0112] Based on the consultation cycle, the minimum data analysis unit in a single consultation process is extracted and the medical entity is obtained by using the named entity recognition technology, wherein the medical entity is a structured text attribute.

[0113] It should be noted that, in this embodiment, Figure 3 and Figure 4As shown, it is a schematic diagram of the parsing steps, wherein the medical record data includes data with time series attributes and data with structured text attributes, wherein the text keywords and time information in the target medical data are parsed to obtain the time intervals between different medical services to obtain the medical cycle, and the corresponding medical cycle is a time series attribute; and based on the medical cycle, the minimum data analysis unit in a single medical visit is extracted, and the named entity recognition technology is used to obtain the medical entity, wherein the medical entity is a structured text attribute, specifically, for the irregular interval data generated by multiple outpatient visits and hospitalizations, information reorganization and integration are performed to shape the minimum data analysis unit of a single medical visit process, and the most important medical record records are extracted, so that in terms of information fusion, the named entity recognition technology is used to identify the medical entity, and the time information extraction technology is used to determine the medical visit time sequence, to ensure data coherence and consistency, and to improve accuracy and availability.

[0114] According to an embodiment of the present invention, the data processing of the medical record data to obtain basic judgment data specifically includes:

[0115] Use regular expressions to perform location-oriented and keyword-oriented data processing on the medical record data, wherein:

[0116] For the data with time series attributes, time window alignment is applied for normalization;

[0117] Perform outlier detection and missing value filling on structured text attribute data, including filling with mean or median.

[0118] It should be noted that, in this embodiment, natural language processing technology is used to perform structured processing on medical record texts. Among them, regular expression technology is used, and according to the structural characteristics of the medical record text, position-oriented and keyword-oriented modes are selected to achieve accurate extraction and organization of information. Specifically, for data with time series attributes, time window alignment technology is specifically applied to perform normalization processing to ensure data consistency. In addition, for data with structured text attributes, outlier detection and processing are performed, and missing values ​​are filled. Specifically, mean or median filling and standardized coding processing can be used. Among them, when processing paper medical records, photo screenshots and consultation recordings, the module ensures the stability and reliability of the data, and through multiple verification and optimization algorithms, it guarantees data quality and applicability, providing a solid foundation for subsequent medical analysis and decision-making.

[0119] According to an embodiment of the present invention, the etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer, wherein the data mining layer is used to mine potential factors related to the disease; the clustering layer is used to obtain the degree of correlation between different etiology characteristics and potential factors; the graph auxiliary layer is used to construct a professional knowledge graph using knowledge graph technology combined with the degree of correlation, and the verification and evaluation layer is used to verify and evaluate the reliability and validity of the identification results.

[0120] It should be noted that, in this embodiment, the etiology hierarchical analysis model described has been trained, and the specific training steps are known technical options for those skilled in the art and will not be described in detail herein. Figure 5 As shown, the etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer. In this embodiment, the purpose of each layer will be emphasized. Specifically, the data mining layer is used to mine potential factors related to the disease. Among them, this layer, as the basis of the potential etiology hierarchical identification and analysis module, comprehensively uses advanced deep learning algorithms to conduct in-depth analysis of the regularized and pre-processed medical data. Through the constructed complex neural network model, it can efficiently and automatically learn the implicit patterns and features in the data, so as to accurately identify potential factors that may be related to the disease. The introduction of the attention mechanism focuses on key medical information, such as symptoms, test results, family medical history, etc., which not only helps to focus on key information, but also greatly improves the accuracy of etiology analysis, and lays a solid data foundation for subsequent hierarchical identification work.

[0121] Furthermore, the clustering division layer is used to obtain the degree of association between different etiological characteristics and potential factors. Based on the results of deep data mining, this layer adopts a hierarchical clustering algorithm to scientifically divide the potential causes into different levels according to different etiological characteristics and the degree of association, and accurately classifies the causes that directly lead to the disease into the first level. These causes often have a direct causal relationship and are the core factors of the disease, while those that may indirectly affect the occurrence of the disease are classified as the second level or higher. For example, some bad living habits may indirectly cause the disease by affecting the body's immune system, etc. These factors will be classified into a higher level. This hierarchical division method can provide a more comprehensive understanding of the causes of the disease and provide a strong decision-making basis for formulating personalized treatment plans.

[0122] Furthermore, the graph auxiliary layer is used to construct a professional knowledge graph using knowledge graph technology combined with the degree of association. In particular, the professional knowledge graph in the medical field is constructed by combining knowledge graph technology to clearly visualize various medical entities such as diseases, symptoms, examination items, treatment methods, and the relationships between them. In the process of identifying potential causes, by efficiently querying the knowledge graph, rich medical knowledge and valuable experience related to the current case can be quickly obtained. The knowledge graph can not only provide a wider range of medical background information, but also assist in verifying and optimizing the identification results of potential causes, further improving the accuracy and reliability of cause analysis.

[0123] Furthermore, the verification and evaluation layer is used to verify and evaluate the reliability and effectiveness of the identification results. In order to ensure the reliability and effectiveness of the potential cause identification results, a strict verification and evaluation mechanism is set up at this level. When applied, various scientific methods such as cross-validation and independent sample testing are used to repeatedly verify and optimize the identification results. At the same time, in actual application, the professional opinions and feedback of clinical experts can be combined to continuously adjust and improve the algorithm. Through a continuous optimization process, the application value of the algorithm in actual medical diagnosis is improved, providing a more efficient, accurate and intelligent innovative solution for cause identification and analysis in the field of medical diagnosis.

[0124] According to an embodiment of the present invention, the user interaction interface can be used for display based on the visualization display of the cause identification result, wherein differentiated display is performed based on user attributes. When the user attributes are defined as professional medical personnel, report visualization is performed, and when the user attributes are defined as patients and their families, graphic visualization is performed, specifically including bar charts and line graphs.

[0125] It should be noted that in this embodiment, after obtaining the result of etiology identification, it can be visualized, and can be specifically displayed using the user interaction interface. In order to meet the needs of different users, a variety of display modes are provided. For professional medical personnel, report visualization is performed, and detailed data analysis reports are specifically provided, including specific descriptions of the cause of the disease, relevant evidence support, risk assessment and other information, to help doctors deeply understand the process and results of etiology analysis, so as to formulate more accurate treatment plans. For patients and their families, graphic visualization is performed, including bar graphs and line graphs, providing concise and clear popular science displays, explaining the possible causes of the disease, treatment recommendations and precautions in plain language, and enhancing patients' understanding of the disease and self-management ability. For example, a bar graph is used to show the possible distribution of different causes, a line graph is used to show the changing trend of key indicators in the development of the disease, and a pie chart is used to show the proportion of various potential causes in the overall causes.

[0126] It is worth mentioning that the user interaction interface provided in this embodiment is also interactive, wherein the user can obtain more detailed information by clicking on a specific area in a chart or graph when using it. For example, by clicking on a column corresponding to a cause of disease in a bar chart, you can view a detailed analysis report of the cause of disease; by clicking on a specific data point in a line chart, you can understand the specific indicator value and related explanation at that time point. Furthermore, in the display process, attention is paid to the security and privacy protection of data. In the process of displaying the results, medical data confidentiality regulations are strictly observed, and encryption, authorized access and other measures are taken to ensure the patient's personal privacy and the security of medical data. At the same time, in order to facilitate the storage and sharing of data, the user interaction interface also provides a data export function, and users can export the analysis results in a specific format for viewing and further analysis on different devices and platforms.

[0127] The third aspect of the present invention provides a computer-readable storage medium, which includes a medical etiology analysis method program based on natural language processing. When the medical etiology analysis method program based on natural language processing is executed by a processor, it implements the steps of a medical etiology analysis method based on natural language processing as described in any one of the above items.

[0128] The present invention discloses a medical etiology analysis method, system and storage medium based on natural language processing, which reduces the reliance on dictionaries in traditional technologies, enhances the ability to recognize new terms, and is easier to promote and apply in clinical practice through efficient processing of various data sources and improved integration capabilities.

[0129] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0130] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0131] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0132] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks or optical disks, and other media that can store program codes.

[0133] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

Claims

1. A medical etiology analysis method based on natural language processing, characterized in that: The following steps are involved: Obtain target medical data to parse and determine medical record data; Processing the medical record data to obtain basic judgment data, wherein different types of medical record data have different processing processes; Inputting the basic judgment data into a cause-of-disease hierarchical analysis model for analysis to obtain a cause-of-disease identification result, wherein the cause-of-disease hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer, and a verification and evaluation layer; Visual display is performed based on the cause identification result, wherein the display method at least includes graphics, images and report display.

2. A method for medical etiology analysis based on natural language processing according to claim 1, characterized in that: Obtain target medical data, including: Acquiring patient data based on a user interaction interface, wherein the patient data includes at least a patient identifier and a case number; Acquire image data and perform optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records; Obtaining the medical consultation recording data for speech-to-text processing, and adding the text processing result to the text data; The target medical data is obtained based on the patient data and the text data.

3. A method for medical etiology analysis based on natural language processing according to claim 2, characterized in that: Parsing the target medical data to determine the medical record data includes: The types of the medical record data include time series attributes and structured text attributes, where: Parsing text keywords and time information in the target medical data to obtain the time intervals between different medical services to obtain a consultation cycle, wherein the consultation cycle is a time series attribute; Based on the consultation cycle, the minimum data analysis unit in a single consultation process is extracted and the medical entity is obtained by using the named entity recognition technology, wherein the medical entity is a structured text attribute.

4. The method for medical etiology analysis based on natural language processing according to claim 3, characterized in that: The data processing of the medical record data to obtain basic judgment data specifically includes: Use regular expressions to perform location-oriented and keyword-oriented data processing on the medical record data, wherein: For the data with time series attributes, time window alignment is applied for normalization; Perform outlier detection and missing value filling on structured text attribute data, including filling with mean or median.

5. A method for medical etiology analysis based on natural language processing according to claim 4, characterized in that: The etiology hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer and a verification and evaluation layer, wherein the data mining layer is used to mine potential factors related to the disease; the clustering layer is used to obtain the degree of association between different etiology characteristics and potential factors; the graph auxiliary layer is used to use knowledge graph technology combined with the degree of association to construct a professional knowledge graph; and the verification and evaluation layer is used to verify and evaluate the reliability and effectiveness of the identification results.

6. A method for medical etiology analysis based on natural language processing according to claim 5, characterized in that: The user interaction interface can be used for displaying the visualization based on the etiology identification result, wherein differentiated display is performed based on user attributes. When the user attributes are defined as professional medical personnel, report visualization is performed. When the user attributes are defined as patients and their families, graphic visualization is performed, specifically including bar charts and line graphs.

7. A medical etiology analysis system based on natural language processing, characterized in that: The invention comprises a memory and a processor, wherein the memory comprises a medical etiology analysis method program based on natural language processing, and the medical etiology analysis method program based on natural language processing is executed by the processor to implement the following steps: Obtain target medical data to parse and determine medical record data; Processing the medical record data to obtain basic judgment data, wherein different types of medical record data have different processing processes; Inputting the basic judgment data into a cause-of-disease hierarchical analysis model for analysis to obtain a cause-of-disease identification result, wherein the cause-of-disease hierarchical analysis model includes a data mining layer, a clustering layer, a graph auxiliary layer, and a verification and evaluation layer; Visual display is performed based on the cause identification result, wherein the display method at least includes graphics, images and report display.

8. The medical etiology analysis system based on natural language processing according to claim 7, characterized in that: Obtain target medical data, including: Acquiring patient data based on a user interaction interface, wherein the patient data includes at least a patient identifier and a case number; Acquire image data and perform optical character recognition to obtain text data, wherein the image data includes photographed images and / or screenshots corresponding to paper medical records; Obtaining the medical consultation recording data for speech-to-text processing, and adding the text processing result to the text data; The target medical data is obtained based on the patient data and the text data.

9. A medical etiology analysis system based on natural language processing according to claim 8, characterized in that: Parsing the target medical data to determine the medical record data includes: The types of the medical record data include time series attributes and structured text attributes, where: Parsing text keywords and time information in the target medical data to obtain the time intervals between different medical services to obtain a consultation cycle, wherein the consultation cycle is a time series attribute; Based on the consultation cycle, the minimum data analysis unit in a single consultation process is extracted and the medical entity is obtained by using the named entity recognition technology, wherein the medical entity is a structured text attribute.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a medical etiology analysis method program based on natural language processing. When the medical etiology analysis method program based on natural language processing is executed by a processor, the steps of a medical etiology analysis method based on natural language processing as described in any one of claims 1 to 6 are implemented.

Images