Method and system for clinical data mining analysis of systemic lupus erythematosus in children

Abstract

The application provides a kind of child systemic lupus erythematosus clinical data mining analysis method and system, by determining the standardized clinical feature set based on multi-source clinical data, and then extracting the dynamic evolution characteristics of lupus erythematosus course from the clinical feature set;According to the dynamic evolution characteristics, the feature correlation atlas of lupus erythematosus is generated, and the key physiological influence factors of lupus erythematosus in different course stages are determined based on the feature correlation atlas;Cluster analysis is carried out on all key physiological influence factors, and the feature group with high risk evolution trend in lupus erythematosus patients is screened out, and the prediction curve of disease activity is generated based on the change gradient in the feature group;Disease fluctuation index is determined by the prediction curve, and then the clinical auxiliary decision result of child systemic lupus erythematosus disease is generated. By using the scheme of the application, the dynamic mining analysis of child systemic lupus erythematosus disease can be realized based on the interaction between different clinical characteristics and time sequence coupling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data mining analysis, more particularly, the present application relates to a kind of children systemic lupus erythematosus clinical data mining analysis method and system. BACKGROUND

[0002] Clinical data mining analysis refers to the use of statistical modeling, machine learning or deep learning algorithms to automatically extract and pattern discovery the implicit relationships in clinical data, such as identifying key biomarkers, constructing disease activity prediction models, analyzing disease recurrence risk factors, etc., to extract potential patterns from complex clinical data, support early disease identification, individualized treatment decisions and prognosis assessment.

[0003] Children systemic lupus erythematosus clinical data mining analysis refers to the use of statistical analysis and machine learning algorithms to extract, correlate and predict disease-related features based on multi-source clinical data of children with systemic lupus erythematosus, thus revealing the evolution of the disease, optimizing diagnosis and treatment decisions, and improving the accuracy of prognosis assessment; However, existing clinical data analysis methods are mostly limited to statistical analysis of single data type or cross-sectional comparison based on static time points, making it difficult to effectively capture the dynamic evolution of the disease, and the complex correlation in the clinical data of children with systemic lupus erythematosus is not fully mined, making it difficult to construct a feature network that fully reflects the internal mechanism of the disease, resulting in low disease activity prediction accuracy, and unable to intervene in the disease fluctuation in time, therefore, how to realize the dynamic mining analysis of children systemic lupus erythematosus based on the interaction between different clinical features and time sequence coupling has become a difficult problem in the industry. SUMMARY

[0004] The present application provides a kind of children systemic lupus erythematosus clinical data mining analysis method and system, which can realize the dynamic mining analysis of children systemic lupus erythematosus based on the interaction between different clinical features and time sequence coupling.

[0005] In the first aspect, the present application provides a kind of children systemic lupus erythematosus clinical data mining analysis method, comprising the following steps:

[0006] Obtaining multi-source clinical data of systemic lupus erythematosus patients throughout the diagnosis and treatment process;

[0007] Based on the structural differences of each data type in the multi-source clinical data, a standardized clinical feature set is determined, and then the dynamic evolution features of systemic lupus erythematosus disease course are extracted from the clinical feature set;

[0008] According to the extracted dynamic evolution features, a feature correlation map of systemic lupus erythematosus is generated, and then the key physiological impact factors of systemic lupus erythematosus at different stages of disease course are determined based on the feature correlation map;

[0009] performing clustering analysis on all the key physiological impact factors, screening out a characteristic group with a high-risk evolution trend in the lupus erythematosus patients, and generating a prediction curve of disease activity based on the change gradient of the key physiological impact factors in time in the characteristic group;

[0010] determining a disease fluctuation index from the prediction curve of disease activity, and further generating a clinical auxiliary decision result of the systemic lupus erythematosus disease of children through the disease fluctuation index.

[0011] In combination with the first aspect, in a possible implementation manner, the determining the standardized clinical feature set based on the structural differences of the various data types in the multi-source clinical data specifically includes:

[0012] identifying structural differences between unstructured text data and structured numerical value data in the multi-source clinical data;

[0013] performing uniform medical language system concept identifier mapping on the unstructured text data, and performing reference interval normalization based on age and gender on the structured numerical value data;

[0014] generating the standardized clinical feature set from the mapped concept identifiers and the normalized numerical value data.

[0015] In combination with the first aspect, in a possible implementation manner, the extracting dynamic evolution features of the lupus erythematosus disease course from the clinical feature set specifically includes:

[0016] selecting time series data of a preset core indicator list of lupus erythematosus from the clinical feature set;

[0017] determining dynamic evolution features of the lupus erythematosus disease course based on the time series data.

[0018] In combination with the first aspect, in a possible implementation manner, the generating a feature correlation graph of lupus erythematosus according to the extracted dynamic evolution features specifically includes:

[0019] respectively determining correlation coefficients between the extracted dynamic evolution features and each laboratory indicator node, immune response node and organ involvement node;

[0020] screening out strong correlation relationships with absolute values greater than a preset significance threshold from all the correlation coefficients;

[0021] generating a multi-dimensional feature correlation matrix through the screened strong correlation relationships;

[0022] constructing a feature correlation graph of lupus erythematosus based on the multi-dimensional feature correlation matrix.

[0023] In a possible implementation manner of the first aspect, determining the key physiological impact factors of SLE at different disease stages based on the feature correlation graph specifically includes:

[0024] determining the importance scores of each node in the feature correlation graph;

[0025] sorting all nodes in the feature correlation graph based on all the importance scores;

[0026] determining the key physiological impact factors of SLE at different disease stages through the sorting result.

[0027] In a possible implementation manner of the first aspect, performing cluster analysis on all the key physiological impact factors to screen a feature group with a high-risk evolution trend in SLE patients specifically includes:

[0028] organizing all the key physiological impact factors into a multi-dimensional time sequence of SLE patients in chronological order;

[0029] performing group cluster division on the multi-dimensional time sequence to obtain a clustering result;

[0030] identifying a feature group with a high-risk evolution trend in SLE patients from the clustering result according to a preset high-risk judgment rule.

[0031] In a possible implementation manner of the first aspect, generating a prediction curve of disease activity based on the change gradient of the key physiological impact factors in the feature group over time specifically includes:

[0032] extracting the change gradient of the key physiological impact factors in the feature group over time;

[0033] predicting a disease activity distribution based on the change gradient;

[0034] generating a prediction curve of disease activity according to the predicted disease activity distribution.

[0035] In a possible implementation manner of the first aspect, determining a disease condition fluctuation index based on the prediction curve of disease activity specifically includes:

[0036] determining an integral area of a reference curve representing a core development trend of disease activity in the prediction curve of disease activity;

[0037] determining an area of a region surrounded by upper and lower boundary curves representing fluctuation ranges of disease activity in the prediction curve;

[0038] determining a disease condition fluctuation index according to the integral area of the reference curve and the area of the region.

[0039] With reference to the first aspect, in a possible implementation manner, the multi-source clinical data comprises laboratory test indexes, immunology detection parameters, disease history record data, and medical image data.

[0040] In a second aspect, the present application provides a clinical data mining analysis system for childhood systemic lupus erythematosus, comprising:

[0041] An acquisition module is configured to acquire multi-source clinical data of a systemic lupus erythematosus patient in a whole diagnosis and treatment process;

[0042] A processing module is configured to determine a standardized clinical feature set based on structural differences of various data types in the multi-source clinical data, and further extract dynamic evolution features of a systemic lupus erythematosus disease course from the clinical feature set;

[0043] The processing module is further configured to generate a feature correlation graph of systemic lupus erythematosus according to the extracted dynamic evolution features, and further determine key physiological impact factors of systemic lupus erythematosus at different disease course stages based on the feature correlation graph;

[0044] The processing module is further configured to perform clustering analysis on all the key physiological impact factors, filter out a feature group with a high-risk evolution trend in the systemic lupus erythematosus patient, and generate a prediction curve of disease activity based on a change gradient of the key physiological impact factors in time in the feature group;

[0045] An execution module is configured to determine a disease condition fluctuation index from the prediction curve of disease activity, and further generate a clinical auxiliary decision result of a childhood systemic lupus erythematosus disease condition through the disease condition fluctuation index.

[0046] The technical scheme provided by the present application has the following beneficial effects:

[0047] In the present application, multi-source clinical data of a systemic lupus erythematosus patient in a whole diagnosis and treatment process is acquired; a standardized clinical feature set is determined based on structural differences of various data types in the multi-source clinical data, and further dynamic evolution features of a systemic lupus erythematosus disease course are extracted from the clinical feature set; a feature correlation graph of systemic lupus erythematosus is generated according to the extracted dynamic evolution features, and further key physiological impact factors of systemic lupus erythematosus at different disease course stages are determined based on the feature correlation graph; clustering analysis is performed on all the key physiological impact factors, a feature group with a high-risk evolution trend in the systemic lupus erythematosus patient is filtered out, and a prediction curve of disease activity is generated based on a change gradient of the key physiological impact factors in time in the feature group; a disease condition fluctuation index is determined from the prediction curve of disease activity, and further a clinical auxiliary decision result of a childhood systemic lupus erythematosus disease condition is generated through the disease condition fluctuation index.

[0048] Therefore, this application firstly extracts dynamic evolutionary features of the lupus disease course from the clinical feature set, transforming static point data into dynamic trend features. This process achieves temporal coupling analysis, quantifying the speed and stability of disease progression, thereby identifying key dynamic patterns and fundamentally improving the depth of understanding of the disease course evolution. Secondly, based on the feature association map, key physiological influencing factors of lupus at different disease stages are determined, intuitively revealing the complex interaction relationships between different clinical features. This allows for the dynamic location of the truly core driving factors at different disease stages, enabling doctors to focus on core targets from massive associations. Then, based on the temporal gradient of key physiological influencing factors in the feature group, a prediction curve of disease activity is generated. This process not only provides the most likely future but also quantifies optimism and pessimism. By defining the boundaries of the disease activity curve, a high-precision dynamic risk warning is achieved, which greatly enhances the information dimension and clinical reference value of the prediction results. Finally, the disease fluctuation index is determined from the disease activity prediction curve, which can extract the predicted disease severity and the range of prediction uncertainty into a comprehensive and quantitative risk assessment indicator. It can sensitively capture the highest risk situation with both high activity and high instability, providing an objective, unified and highly insightful quantitative basis for subsequent risk classification and resource priority allocation. Furthermore, the clinical auxiliary decision-making results of childhood systemic lupus erythematosus are generated through the disease fluctuation index, realizing a seamless connection from data mining to clinical action, which can significantly improve the timeliness of responding to disease fluctuations and the effectiveness of intervention. In summary, this scheme can realize dynamic mining and analysis of childhood systemic lupus erythematosus based on the interaction relationship and temporal coupling between different clinical features. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is an exemplary flowchart of a clinical data mining and analysis method for systemic lupus erythematosus in children, as shown in some embodiments of this application;

[0051] Figure 2 This is an exemplary flowchart illustrating the generation of feature association maps according to some embodiments of this application;

[0052] Figure 3 This is an exemplary flowchart illustrating the determination of key physiological influencing factors according to some embodiments of this application;

[0053] Figure 4 is a structural schematic diagram of a child systemic lupus erythematosus clinical data mining analysis system shown according to some embodiments of the present application;

[0054] Figure 5 is a structural schematic diagram of a computer device for implementing a child systemic lupus erythematosus clinical data mining analysis method shown according to some embodiments of the present application. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0056] Reference Figure 1 The figure is an exemplary flowchart of a child systemic lupus erythematosus clinical data mining analysis method shown according to some embodiments of the present application, which mainly includes the following steps:

[0057] In step 101, multi-source clinical data of a systemic lupus erythematosus patient in the whole process of diagnosis and treatment is acquired.

[0058] In specific implementation, the acquisition of the multi-source clinical data of the systemic lupus erythematosus patient in the whole process of diagnosis and treatment can be implemented in the following manner, that is, the multi-source clinical data can include laboratory test indexes, immunology detection parameters, disease course record data and medical image data, and the acquisition of the multi-source clinical data of the systemic lupus erythematosus patient in the whole process of diagnosis and treatment can be completed by an automatic data acquisition service deployed in a hospital data gateway. The automatic data acquisition service can asynchronously extract the disease course record data of the systemic lupus erythematosus patient from a hospital information system at a regular time through a pre-defined application programming interface (API) and a database query script, wherein each systemic lupus erythematosus patient has a unique identifier; the laboratory test indexes and the immunology detection parameters changing over time are acquired from a laboratory information system; the identifier and the structured report of the medical image data are extracted from an image archiving system; in addition, the acquired multi-source clinical data can be standardized in format and cleaned up, and then converted into a JSON format for subsequent analysis; in other embodiments, other methods can also be used for acquisition, which is not specifically limited here.

[0059] It should be noted that the multi-source clinical data in the present application refers to a set of heterogeneous clinical evidence generated by different medical information systems from multiple dimensions to record and reflect the physiological state, pathological changes and treatment intervention of a child systemic lupus erythematosus patient during the complete medical cycle from the first visit to subsequent follow-up.

[0060] In step 102, a standardized clinical feature set is determined based on the structural differences of each data type in the multi-source clinical data, and then a dynamic evolution feature of the lupus disease course is extracted from the clinical feature set.

[0061] In some embodiments, the determination of the standardized clinical feature set based on the structural differences of each data type in the multi-source clinical data can be implemented by the following steps:

[0062] Identifying the structural differences between unstructured text data and structured numerical data in the multi-source clinical data;

[0063] Performing uniform medical language system concept identifier mapping on the unstructured text data and performing age and gender reference interval normalization on the structured numerical data;

[0064] Generating a standardized clinical feature set from the mapped concept identifier and the normalized numerical data.

[0065] In a specific implementation, the identification of the structural differences between unstructured text data and structured numerical data in the multi-source clinical data can be implemented in the following manner, that is, a data mode parser can be used to traverse the multi-source clinical data; the parser identifies according to predefined rules, for example, the free text paragraphs and descriptive fields containing non-standard terms in the multi-source clinical data derived from the disease course record and the medical image report are determined as unstructured text data; at the same time, the test items in the multi-source clinical data derived from the laboratory information system, such as the fixed field name, the numerical result and the standard unit, such as "anti-dsDNA antibody", are determined as structured numerical data; the parser finally outputs a data classification list to provide a basis for subsequent differential standardization processing; in other embodiments, the data mode can also be automatically identified by a machine learning classifier, or the data type can be directly determined according to the database table structure definition of the data source, which is not limited here.

[0066] In a specific implementation, the unified medical language system concept identifier mapping for the unstructured text data and the age and gender based reference interval normalization for the structured numerical data can be implemented in the following manner: for the unstructured text data, an API of the UMLS (Unified Medical Language System) Metathesaurus can be invoked to map medical terms appearing in the text, such as “pulmonary inflammation” and “proteinuria”, to unique and standard concept identifiers to achieve semantic standardization; for the structured numerical data, a preconfigured age-gender specific reference value range database can be accessed to convert original detection values, such as the concentration of complement C3, to normalized ratios relative to the median of the reference range of the population through linear calculation; in this way, clinical data of different sources and structures are uniformly converted into machine readable and dimensionally consistent standard forms; in other embodiments, the mapping can also be completed using other medical terminology systems, such as SNOMED CT, and the normalization can also be implemented using Z-Score standardization or a quartile based scaling method, which are not limited in the present application.

[0067] In a specific implementation, the generation of the standardized clinical feature set from the mapped concept identifiers and the normalized numerical data can be implemented in the following manner: the mapped concept identifiers and the normalized numerical data can be integrated and structured; specifically, all UMLS concept identifiers obtained by mapping, normalized ratios of all laboratory and immune indicators, and other features can be combined according to a pre-defined field order and data structure for each patient, and finally a plurality of standardized clinical feature records in vector form indexed by patient-time point are generated, and the collection of all these records constitutes the standardized clinical feature set for subsequent analysis, which is not limited in the present application.

[0068] It should be noted that the structural differences of various data types in the present application refer to the inherent heterogeneity existing in multi-source clinical data; the mapped concept identifiers in the present application refer to machine readable unique identifiers after semantic standardization of medical concepts contained in the original unstructured text; the normalized numerical data in the present application refer to mutually comparable numerical indicators after eliminating the dimensional and reference differences caused by physiological factors such as age and gender; and the standardized clinical feature set in the present application refers to a structured data set integrating all semantic standardization and numerical normalization results.

[0069] In some embodiments, the extraction of dynamic evolution features of lupus disease course from the clinical feature set can be implemented in the following steps:

[0070] select time series data of a preset core indicator list of lupus erythematosus from the clinical feature set;

[0071] determine a dynamic evolution feature of the course of lupus erythematosus based on the time series data.

[0072] In specific implementation, the time series data of the preset core indicator list of lupus erythematosus selected from the clinical feature set can be implemented in the following manner, that is, the core indicator list pre-stored in the system configuration can be based on the clinical diagnosis and treatment guidelines, which includes but is not limited to anti-dsDNA antibody, complement C3, complement C4 and urine protein / cratin ratio, and then all normalized numerical sequences of the indicators corresponding to the core indicator list corresponding to the unique identifier of the patient can be extracted from the clinical feature set through database query operation, and used as time series data; this extraction process ensures that each selected indicator of each patient forms a longitudinal data track sorted by time stamp, to provide structured input for subsequent dynamic analysis; in other embodiments, the core indicator list can also be dynamically generated and updated from historical data by a feature importance scoring algorithm, which is not limited in the present application.

[0073] In specific implementation, the determination of the dynamic evolution feature of the course of lupus erythematosus based on the time series data can be implemented in the following manner, that is, a fixed-length sliding time window can be used to traverse the time series data of each indicator in the time series data, and a linear regression algorithm can be used to calculate the slope value of the data points in each window to quantify the change trend of the indicator, while the fluctuation variance of the data points in the window relative to the mean value is calculated to quantify the stability of the indicator; then the slope values and fluctuation variances of all preset indicators in the same time window are connected and combined in a preset order to form a feature vector representing the dynamic evolution of the multi-dimensional disease condition in the time period, and used as the dynamic evolution feature of the course of lupus erythematosus; in other embodiments, the dynamic evolution feature can also be constructed by dynamic time warping distance or a feature extraction method based on recurrent neural network, which is not limited here.

[0074] It should be noted that the preset core indicator list of childhood systemic lupus erythematosus in the present application refers to a subset of specific clinical indicators that has been determined before the analysis starts, contains a limited number of indicators that are highly related to the pathophysiological process of childhood systemic lupus erythematosus, and is used as a criterion and focus for data screening; the time series data in the present application refers to the standardized numerical records of the dynamic changes of clinical indicators in the clinical feature set; the dynamic evolution feature of the course in the present application refers to a composite indicator representing the behavior pattern of childhood systemic lupus erythematosus in a specific time period, which is a mathematical abstraction and integration of the change trend and fluctuation state of the disease to directly and quantitatively reveal the dynamics of the disease evolution.

[0075] In step 103, a feature correlation graph of lupus erythematosus is generated according to the extracted dynamic evolution features, and then key physiological influencing factors of lupus erythematosus in different disease stages are determined based on the feature correlation graph.

[0076] In some embodiments, the reference Figure 2 As shown in the figure, the figure is an exemplary flowchart for generating a feature correlation graph in some embodiments of the present application, and the generation of the feature correlation graph of lupus erythematosus according to the extracted dynamic evolution features can be implemented by the following steps:

[0077] The correlation coefficients between the extracted dynamic evolution features and each laboratory index node, immune response node and organ involvement node are determined respectively;

[0078] Strong correlation relationships with absolute values greater than a preset significance threshold are screened from all correlation coefficients;

[0079] A multidimensional feature correlation matrix is generated through the screened strong correlation relationships;

[0080] A feature correlation graph of lupus erythematosus is constructed based on the multidimensional feature correlation matrix.

[0081] In a specific implementation, the correlation coefficients between the extracted dynamic evolution features and each laboratory index node, immune response node and organ involvement node can be determined in the following manner, that is, the extracted dynamic evolution features and each defined laboratory index node, immune response node and organ involvement node can be traversed, and corresponding numerical pairs in all patient samples are collected within a preset time window; then a Spearman rank correlation coefficient algorithm in a statistical calculation library is called to calculate the numerical pairs of each feature-node combination to obtain a coefficient value reflecting its monotonic correlation as a correlation coefficient; finally, a complete set containing correlation coefficients between all dynamic evolution features and all laboratory index nodes, immune response nodes and organ involvement nodes is output; in other embodiments, Pearson correlation coefficient or Kendall rank correlation coefficient can also be used for similarity measurement, which is not limited in the present application.

[0082] In a specific implementation, the strong correlations with absolute values greater than a preset significance threshold can be selected from all the correlation coefficients in the following manner: all correlation coefficients with absolute values greater than the preset significance threshold can be selected as strong correlations by a conditional judgment statement, and the specific dynamic evolution characteristics corresponding to these strong correlations and the combinations of laboratory index nodes, immune response nodes, or organ involvement nodes are recorded; the laboratory index nodes include, but are not limited to, serum complement C3, C4, white blood cell count in blood routine, and other basic serological measurement entities; the immune response nodes include, but are not limited to, anti-dsDNA antibody, anti-Sm antibody, antinuclear antibody titer, and other disease-specific autoantibody entities; the organ involvement nodes include, but are not limited to, proteinuria (24-hour urine protein quantification), renal pathological activity index, nervous system symptom identification, and other key target organ damage clinical identification entities; the significance threshold can be preset based on clinical expert experience and statistical significance requirements, which can be set to 0.5 to retain correlations with moderate intensity or above, and in other embodiments, the significance threshold can also be adjusted based on specific clinical scenarios, which is not limited in the present application.

[0083] In a specific implementation, the multi-dimensional feature association matrix generated by the selected strong correlations can be implemented in the following manner: a three-dimensional matrix can be initialized, and the three dimensions correspond to the laboratory index node dimension, the immune response node dimension, and the organ involvement node dimension; then all selected strong correlations are traversed, and the coordinate position of each strong correlation in the three-dimensional matrix is determined according to the types of the laboratory index node, the immune response node, or the organ involvement node involved in the strong correlation, and the correlation coefficient is filled as the matrix element value in the corresponding position; the matrix positions not filled with values remain zero or null, and finally a multi-dimensional feature association matrix is generated; in other embodiments, the multi-dimensional feature association matrix can also be stored in the form of a key-value pair dictionary or a graph structure adjacency list, which is not limited here.

[0084] In a specific implementation, the feature correlation graph of lupus erythematosus can be constructed based on the multi-dimensional feature correlation matrix in the following manner: the multi-dimensional feature correlation matrix can be used as an adjacency matrix of a graph structure, and a batch import tool of a graph database such as Neo4j can be used to convert non-zero elements in the matrix into edges in the graph; each laboratory index node, immune response node, and organ involvement node is created as a graph node, and each non-zero matrix element, i.e., the correlation coefficient corresponding to the strong correlation relationship filtered out, is created as a directed edge connecting the corresponding nodes, and the value of the correlation coefficient is set as the weight attribute of the edge; finally, a visual correlation graph containing all nodes and weighted edges is constructed in the graph database, which is the feature correlation graph of lupus erythematosus; in other embodiments, the feature correlation graph can also be constructed in memory using a network analysis library such as NetworkX, or implemented using other graph computing frameworks, which are not limited here.

[0085] It should be noted that the laboratory index node in the present application refers to an objective measurement index entity representing the patient's basic physiological state and disease activity, which is used as a bottom benchmark anchor point for constructing complex correlation relationships; the immune response node in the present application refers to a functional index entity that collectively represents the abnormal activation state and specificity of the immune system, which is used to directly reflect the specific immunological disorder characteristics of the disease; the organ involvement node in the present application refers to a clinical endpoint entity that marks and quantifies the functional impact or structural damage caused by the disease to a specific target organ, which is used to directly reflect the pathological and physiological consequences caused by the disease; the correlation coefficient in the present application refers to a numerical index representing the linear or hierarchical correlation strength and direction between dynamic evolution characteristics and clinical nodes, including laboratory index nodes, immune response nodes, or organ involvement nodes; the strong correlation relationship in the present application refers to a significant feature-node pairing relationship, which is used to filter out a subset that is most likely to represent real and important pathological and physiological connections; the multi-dimensional feature correlation matrix in the present application refers to a sparse data container used to structure and integrate all strong correlation relationships, which is used to systematically organize and spatially map discrete, point-to-point strong correlation relationships according to their node type dimensions, thereby providing a complete correlation data basis that can be directly utilized for directly generating a feature correlation graph; the feature correlation graph in the present application refers to a structured network model that integrates and visualizes the complex correlation relationships in multi-dimensional clinical data of childhood systemic lupus erythematosus, which is used to integrate discrete laboratory indexes, immune responses, and organ involvement, and other clinical features, and the strong correlation relationships between them obtained by quantitative calculation in a unified graph structure framework, thereby directly revealing the potential, non-obvious interaction and synergistic evolution path between different dimensional clinical features, and the feature correlation graph serves as an intermediate representation connecting original data and high-level clinical insights.

[0086] In some embodiments, referenceFigure 3 As shown in the figure, the figure is an exemplary flowchart for determining key physiological impact factors in some embodiments of the present application, and the determination of key physiological impact factors of lupus erythematosus at different stages of disease progression in the present embodiment can be achieved by the following steps:

[0087] First, in step 1031, the importance score of each node in the feature correlation graph is determined;

[0088] Second, in step 1032, all nodes in the feature correlation graph are sorted based on all importance scores;

[0089] Finally, in step 1033, the key physiological impact factors of lupus erythematosus at different stages of disease progression are determined through the sorting result.

[0090] In specific implementation, the importance score of each node in the feature correlation graph can be achieved by the following way, that is, the centrality calculation module in the graph analysis library can be called, the relative importance of each node in the network is calculated based on the connection relationship of the nodes and the weight attribute of the edges in the feature correlation graph through the iterative propagation algorithm, and the calculated relative importance is taken as the importance score of each node; wherein the weight of the edge uses the correlation coefficient stored in the feature correlation graph to ensure that the node with higher connection strength obtains higher importance score; in other embodiments, the node importance can also be calculated by using graph theory algorithms such as feature vector centrality and betweenness centrality, which are not limited in the present application.

[0091] In specific implementation, the sorting of all nodes in the feature correlation graph based on all importance scores can be achieved by the following way, that is, the sorting algorithm can be used to arrange all nodes in the feature correlation graph in descending order according to the importance score, and a node sequence in descending order of importance is generated; a stable sorting algorithm can be used in the sorting process to ensure that the nodes with the same score maintain their relative order in the original feature correlation graph; in other embodiments, the node sorting can also be achieved by constructing a maximum heap data structure or using the sorting query function of the database, which is not limited in the present application.

[0092] In a specific implementation, the key physiological impact factors of lupus erythematosus at different stages of disease progression can be determined by the following method: according to the preset number threshold N, the first N nodes are selected from the sorted node sequence, the clinical features represented by the selected nodes are determined as the key physiological impact factors of the corresponding stage of disease progression, and the key physiological impact factors of lupus erythematosus at different stages of disease progression are obtained. The number threshold can be set based on the experience of clinical experts and actual analysis requirements, and it is usually an integer value between 5 and 10. In other embodiments, the number of key physiological impact factors can also be automatically determined according to the natural breakpoint or elbow rule of importance score, which is not limited here.

[0093] It should be noted that the importance score in the present application refers to a numerical indicator representing the relative influence and pivotal position of each node in the entire association network in the feature association graph; the sorting result in the present application refers to the node sequence formed by arranging all nodes in descending order according to their importance scores; the key physiological impact factor in the present application refers to the clinical feature that occupies the core regulatory position in each stage of disease progression, which is used to dynamically reveal the core pathological and physiological link driving the change of disease condition at different evolution periods of childhood systemic lupus erythematosus.

[0094] In step 104, all key physiological impact factors are subjected to cluster analysis to screen out a feature group with high risk evolution trend in lupus erythematosus patients, and a prediction curve of disease activity is generated based on the change gradient of key physiological impact factors in time in the feature group.

[0095] In some embodiments, the key physiological impact factors of all key physiological impact factors are subjected to cluster analysis to screen out a feature group with high risk evolution trend in lupus erythematosus patients can be achieved by the following steps:

[0096] All key physiological impact factors are organized in time sequence as a multi-dimensional time sequence of lupus erythematosus patients;

[0097] The multi-dimensional time sequence is subjected to group clustering division to obtain a clustering result;

[0098] According to a preset high risk judgment rule, a feature group with high risk evolution trend in lupus erythematosus patients is identified from the clustering result.

[0099] In a specific implementation, the following method can be used to organize all the key physiological impact factors in time sequence as a multi-dimensional time series of a lupus erythematosus patient: using the unique identifier of the lupus erythematosus patient from the hospital information system as the primary key, extracting the normalized values of the key physiological impact factors corresponding to each stage of the disease course from the standardized clinical feature set; constructing a multi-dimensional time series for each patient in chronological order, where each time point contains the values of all key physiological impact factors at that time, and the missing data can be filled by linear interpolation; and finally forming a complete patient trajectory data indexed by time and dimensioned by the values of the key physiological impact factors, i.e., a multi-dimensional time series of a lupus erythematosus patient. In other embodiments, spline interpolation or forward filling method can also be used to handle missing values, which are not limited in the present application.

[0100] In a specific implementation, the following method can be used to perform group clustering on the multi-dimensional time series to obtain a clustering result: first, the dynamic time warping algorithm can be used to calculate the similarity distance between the multi-dimensional time series of each patient to eliminate the influence of phase difference on the time axis on clustering; then, a clustering algorithm based on Gaussian mixture model is used to divide the similarity distance matrix into groups, which can automatically determine the optimal number of clusters and group patients with similar evolution patterns into the same group, and finally obtain multiple groups; in other embodiments, K-means clustering with dynamic time warping distance or a sequence encoding clustering method based on deep learning can also be used, which are not limited in the present application.

[0101] In a specific implementation, the following method can be used to identify a feature group with a high-risk evolution trend from the clustering result according to a predetermined high-risk judgment rule: first, the predetermined high-risk judgment rule can be set based on clinical expert knowledge, which requires that more than a predetermined proportion (such as 80%) of patients in the feature group continuously exceed the clinical danger threshold in the key organ involvement index, for example, more than 80% of patients in the clustering group continuously have a key organ involvement index (such as the normalized value of “24-hour urine protein quantification” greater than 1.2) in the abnormal high-risk state at the latest two consecutive observation time points; and then by traversing all groups in the clustering result, the group that meets the high-risk judgment rule is screened out as a feature group with a high-risk evolution trend in the lupus erythematosus patients; in other embodiments, a machine learning classifier can also be used to automatically identify a high-risk group, or a multi-index composite rule can be used for risk assessment, which are not limited in the present application.

[0102] It should be noted that the multi-dimensional time series in the present application refers to structured data records integrating all key physiological impact factor values of children with systemic lupus erythematosus at each disease stage, which is used to capture and present the dynamic evolution trajectory and multi-dimensional change pattern of the patient's condition; the clustering result in the present application refers to a collection of several patient groups with high similarity and obvious differences between each other in children with systemic lupus erythematosus, which is used to extract limited and representative typical evolution patterns from a large number of individual disease trajectories, thereby classifying patient groups and providing clear analysis targets for subsequent risk assessment; the high-risk judgment rule in the present application refers to a judgment basis for objectively screening a high-risk patient group from the clustering result based on pre-defined clinical professional knowledge; the feature group with high-risk evolution trend in the present application refers to a specific patient set or sets in the group whose typical disease trajectory points to adverse clinical outcomes, which is used to locate and identify a patient subgroup that needs high clinical attention and active intervention, thereby converting data analysis conclusions into clinical warning information with clear direction.

[0103] In some embodiments, generating a prediction curve of disease activity based on the change gradient of key physiological impact factors over time in the feature group can be achieved by the following steps:

[0104] Extracting the change gradient of key physiological impact factors over time in the feature group;

[0105] Predicting the distribution of disease activity based on the change gradient;

[0106] Generating a prediction curve of disease activity according to the predicted distribution of disease activity.

[0107] In specific implementation, extracting the change gradient of key physiological impact factors over time in the feature group can be achieved by the following way, that is, selecting key physiological impact factors within a preset time period from the multi-dimensional time series of the feature group, wherein the preset time period can be set to the last 6 months, and then using a sliding window to calculate the change slope of each key physiological impact factor within the time period using a linear regression algorithm, and combining the change slopes of all key physiological impact factors to form a change gradient representing the overall disease condition change trend; in other embodiments, the change gradient can also be calculated by using an exponential smoothing method or a first-order difference method, which is not limited in the present application.

[0108] In a specific implementation, predicting the disease activity distribution based on the change gradient can be achieved in the following manner: obtaining a trained quantile regression model, taking the change gradient as an input feature of the model, to predict the distribution of a disease activity index, such as a Systemic Lupus Erythematosus Disease Activity Index (SLEDAI score), in a future time period, such as the next three months, and then outputting a predicted value of a disease activity quantile, thereby obtaining distribution data reflecting the possible fluctuation range of the disease activity, i.e., the disease activity distribution. In other embodiments, a time series prediction algorithm or a machine learning regression model can also be used to predict the disease activity distribution, which is not limited in the present application.

[0109] It should be noted that the quantile regression model is trained based on historical patient cohort data, and the training method is, for example: taking the dynamic evolution features (including but not limited to the trend slope and fluctuation variance of the core indicators of systemic lupus erythematosus) extracted from the historical pediatric systemic lupus erythematosus patient data as input features, and taking the actual SLEDAI score in the corresponding future time period as the target variable; using grid search and cross-validation to optimize the model hyperparameters, such as regularization strength and quantile loss weight, and evaluating the model performance on an independent time validation set to ensure that the coverage probability of the prediction interval (such as 90% interval) is close to the nominal level, which is not described here.

[0110] In a specific implementation, generating a predicted curve of disease activity based on the predicted disease activity distribution can be achieved in the following manner: connecting the disease activity quantile data in the predicted disease activity distribution in chronological order to form three curves of the 10th, 50th, and 90th quantiles, respectively, and taking the prediction band formed by the three curves as the predicted curve of the disease activity, wherein the 50th quantile curve represents the most likely disease activity development trend, and the 10th and 90th quantile curves represent the optimistic and pessimistic prediction boundaries of the disease activity, respectively; in other embodiments, the prediction curve can also be displayed in the form of a confidence interval or a prediction interval, which is not limited here.

[0111] It should be noted that the change gradient in the present application refers to a composite trend index reflecting the rate and direction of the coordinated changes of multiple key physiological influencing factors; the disease activity distribution in the present application refers to a description of all possible values and their corresponding probabilities of the disease activity in a specific future time period, which is used to reveal all potential development possibilities and their uncertainties of the disease in the future; the prediction curve of the disease activity in the present application refers to a dynamic trajectory band obtained by visualizing and structuring the predicted disease activity distribution in the time dimension, which is used to convert the complex distribution data into a prediction atlas containing the most likely path, the optimistic boundary and the pessimistic boundary, so that the clinician can intuitively grasp the future development trend range of the disease.

[0112] In step 105, a disease fluctuation index is determined from the prediction curve of the disease activity, and then a clinical auxiliary decision result of the pediatric systemic lupus erythematosus disease is generated through the disease fluctuation index.

[0113] In some embodiments, the determination of the disease fluctuation index from the prediction curve of the disease activity can be achieved by the following steps:

[0114] determining the integral area of a reference curve representing the core development trend of the disease activity in the prediction curve of the disease activity;

[0115] determining the area of a region surrounded by upper and lower boundary curves representing the fluctuation range of the disease activity in the prediction curve;

[0116] determining the disease fluctuation index according to the integral area of the reference curve and the area of the region.

[0117] In specific implementation, the integral area of the reference curve representing the core development trend of the disease activity in the prediction curve of the disease activity can be achieved by the following way, i.e., the disease activity values of the reference curve (i.e., the 50th percentile curve) as the core trend line at each equal interval time point in the future preset time period can be extracted from the prediction curve of the disease activity, and then the trapezoidal rule in the numerical integration method can be used to calculate the trapezoidal area formed by the adjacent time point values and time intervals in time sequence and to accumulate the summation, so as to finally obtain the integral area of the reference curve representing the core development trend of the disease activity in the prediction curve of the disease activity; in other embodiments, Simpson's integral method or rectangular method can also be used for area calculation, which is not limited in the present application.

[0118] In a specific implementation, the area of the region enclosed by the upper and lower boundary curves of the predicted curve representing the fluctuation range of disease activity can be determined in the following manner: the values of the disease activity at the corresponding time points in the future for the upper boundary curve (i.e., the 90th percentile curve) and the lower boundary curve (i.e., the 10th percentile curve) in the predicted curve can be extracted simultaneously, the absolute value of the difference between the values of the two curves at each time point is calculated, and then the integral value of all the differences with respect to time is calculated using the same numerical integration method. The integral value is the area of the region enclosed by the upper and lower boundary curves. In other embodiments, the area of the region can also be obtained by calculating the difference between the integral areas of the two boundary curves, which is not limited in the present application.

[0119] In a specific implementation, the disease fluctuation index can be determined based on the integral area of the reference curve and the area of the region in the following manner: the integral area of the reference curve calculated is taken as the numerator, the area of the region enclosed by the upper and lower boundary curves is taken as the denominator, the ratio of the two is calculated, and the product of the ratio and a preset adjustment coefficient is calculated. The result of mapping the product value to the 0-1 interval using the Sigmoid function is used as the disease fluctuation index of the future disease course of the child systemic lupus erythematosus. The disease fluctuation index reflects the relative relationship between the severity of the disease activity and the uncertainty range during the prediction period. The value of the disease fluctuation index ranges between 0 and 1. The higher the index value, the greater the possibility of a high-risk state of high disease activity and clear development trend in the future. In addition, the preset adjustment coefficient is used to calibrate the scale and corresponding relationship between the ratio and the final disease fluctuation index. The adjustment coefficient can be determined through a fitting process based on a historical patient cohort. For example, first, the ratio of the integral area of the reference curve to the area of the region for each patient is calculated based on historical data. Then, according to the actual clinical course of these patients, a target clinical score that quantifies the severity and fluctuation of the disease is calculated. Finally, a linear regression method is used to establish the scale relationship between the historical ratio and the target clinical score. The proportional coefficient obtained from the regression analysis is determined as the adjustment coefficient. In other embodiments, the adjustment coefficient can be set based on clinical expert consensus or integrated into a lightweight machine learning model, which is not limited herein.

[0120] It should be noted that the integral area of the reference curve in the present application refers to the comprehensive quantification of the overall severity and duration of disease activity that is most likely to be reached in the future prediction period, which is used to measure the expected load of future disease; the area of the region in the present application refers to the quantitative measure of the maximum uncertainty range of the disease activity that may deviate from its core development trend in the future prediction period, which is used to measure the reliability and risk range of the prediction result; the disease fluctuation index in the present application refers to a composite risk assessment indicator that comprehensively considers the expected severity of future disease activity and prediction uncertainty, which is used to reflect the severity and uncertainty range of disease activity so that clinicians can pay more attention to and intervene.

[0121] In a specific implementation, the clinical auxiliary decision result of the pediatric systemic lupus erythematosus disease can be realized by the disease fluctuation index in the following manner: first, the threshold interval of the pediatric systemic lupus erythematosus disease clinical decision can be set based on the correlation analysis of the disease fluctuation index and adverse prognosis in the historical retrospective data, combined with multiple rounds of expert Delphi method evaluation, the threshold interval refers to a standard segmentation system that divides the disease fluctuation index into risk levels with different clinical significance, which is used to establish a clear mapping relationship between quantitative risk assessment results and qualitative clinical decision suggestions, the threshold interval corresponds to different decision suggestions, for example, the threshold interval is set as: low risk interval [0, 0.3), corresponding to routine follow-up; medium risk interval [0.3, 0.7), corresponding to enhanced monitoring; high risk interval [0.7, 1], corresponding to intensive intervention; further, the disease fluctuation index can be matched with the threshold interval, when the disease fluctuation index falls into a certain threshold interval, the decision text of the corresponding decision suggestion is automatically retrieved and output, for example: when the disease fluctuation index is in the high risk interval, the decision text "it is recommended to immediately admit to the hospital for intensive immunosuppressive treatment" is output, when the disease fluctuation index is in the medium risk interval, the decision text "it is recommended to visit the outpatient department within 2 weeks and adjust the drug regimen" is output, when the disease fluctuation index is in the low risk interval, the decision text "maintain the existing treatment regimen" is output; finally, the output decision text is pushed to the clinical physician workstation interface as the clinical auxiliary decision result of the pediatric systemic lupus erythematosus disease; in other embodiments, a decision generator based on a machine learning model or personalized decision recommendation combined with individual patient characteristics can also be used, which is not limited in the present application.

[0122] In addition, another aspect of the present application, in some embodiments, the present application provides a pediatric systemic lupus erythematosus clinical data mining analysis system, referring to Figure 4 The figure is a structural schematic diagram of a pediatric systemic lupus erythematosus clinical data mining analysis system according to some embodiments of the present application, which includes an acquisition module 401, a processing module 402 and an execution module 403, which are described as follows:

[0123] The acquisition module 401 is mainly used for acquiring multi-source clinical data of a lupus erythematosus patient in the whole diagnosis and treatment process in the present application.

[0124] The processing module 402 is mainly used for determining a standardized clinical feature set based on structural differences of each data type in the multi-source clinical data, and then extracting dynamic evolution features of a lupus erythematosus course from the clinical feature set in the present application.

[0125] The processing module 402 is also used for generating a feature correlation graph of lupus erythematosus according to the extracted dynamic evolution features, and then determining key physiological impact factors of lupus erythematosus at different course stages based on the feature correlation graph in the present application.

[0126] The processing module 402 is also used for performing cluster analysis on all the key physiological impact factors, screening out a feature group with a high-risk evolution trend in the lupus erythematosus patient, and generating a prediction curve of disease activity based on a change gradient of the key physiological impact factors in time in the feature group in the present application.

[0127] The execution module 403 is mainly used for determining a disease fluctuation index from the prediction curve of disease activity, and then generating a clinical auxiliary decision result of a child systemic lupus erythematosus disease condition through the disease fluctuation index in the present application.

[0128] The above describes an example of a child systemic lupus erythematosus clinical data mining and analysis method and system provided by the present application in detail. It can be understood that a corresponding device includes a hardware structure and / or a software module corresponding to each function to achieve the above functions. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed in the present application, the present application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed by hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0129] In some embodiments, the present application also provides a computer device including a memory and a processor, the memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the computer device executes the child systemic lupus erythematosus clinical data mining and analysis method described above.

[0130] In some embodiments, reference is made to Figure 5, the dashed line in the figure indicates that the unit or the module is optional, and the figure is a structural schematic diagram of a computer device for implementing the child systemic lupus erythematosus clinical data mining analysis method. The child systemic lupus erythematosus clinical data mining analysis method in the above embodiment can be implemented by the computer device shown in the figure, and the computer device 500 includes at least one processor 501, a memory 502, and at least one communication unit 505. The computer device 500 can be a terminal device or a server or a chip. Figure 5 The processor 501 can be a general-purpose processor or a special-purpose processor. For example, the processor 501 can be a central processing unit (CPU), which can be used to control the computer device 500, execute a software program, and process data of the software program. The computer device 500 can further include a communication unit 505 to realize input (reception) and output (transmission) of signals.

[0131] The processor 501 can be a general-purpose processor or a special-purpose processor. For example, the processor 501 can be a central processing unit (CPU), which can be used to control the computer device 500, execute a software program, and process data of the software program. The computer device 500 can further include a communication unit 505 to realize input (reception) and output (transmission) of signals.

[0132] For example, the computer device 500 can be a chip, and the communication unit 505 can be an input and / or output circuit of the chip, or the communication unit 505 can be a communication interface of the chip. The chip can be a component of a terminal device or a network device or other device.

[0133] For another example, the computer device 500 can be a terminal device or a server, and the communication unit 505 can be a transceiver of the terminal device or the server, or the communication unit 505 can be a transceiver circuit of the terminal device or the server.

[0134] The computer device 500 can include one or more memories 502, which store programs 504 that can be run by the processor 501 to generate instructions 503, so that the processor 501 executes the method described in the above method embodiments according to the instructions 503. Optionally, the memory 502 can also store data (such as a target audit model). Optionally, the processor 501 can also read the data stored in the memory 502. The data can be stored in the same storage address as the program 504, or the data can be stored in a different storage address from the program 504.

[0135] The processor 501 and the memory 502 can be separately arranged or integrated together, for example, integrated on a system on chip (SOC) of a terminal device.

[0136] It should be understood that each step of the above method embodiments can be accomplished by logic circuits in the form of hardware or instructions in the form of software in the processor 501, which can be a CPU, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof.

[0137] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0138] For example, in some embodiments, the present application also provides a computer readable storage medium, which stores instructions or codes, when the instructions or codes are run on a computer, cause the computer to perform the above-mentioned clinical data mining analysis method for children with systemic lupus erythematosus.

[0139] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to encompass within their scope all such variations and modifications as are included within the scope of the application.

[0140] Obviously, various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and scope of the present application. Accordingly, the present application intends to include all such modifications and changes as fall within the scope of the claims and their equivalents.

Claims

1. A method for clinical data mining and analysis of systemic lupus erythematosus in children, characterized in that, Includes the following steps: Acquire multi-source clinical data of lupus patients throughout the entire diagnosis and treatment process; Based on the structural differences of various data types in the multi-source clinical data, a standardized set of clinical features is determined, and then the dynamic evolution characteristics of the lupus erythematosus course are extracted from the set of clinical features. Based on the extracted dynamic evolution features, a feature association map of lupus erythematosus is generated, and then the key physiological influencing factors of lupus erythematosus at different stages of disease are determined based on the feature association map. Cluster analysis was performed on all key physiological influencing factors to screen out a characteristic group of lupus patients with a high risk of evolution. Based on the temporal change gradient of key physiological influencing factors in the characteristic group, a prediction curve of disease activity was generated. The disease activity prediction curve is used to determine the disease fluctuation index, and then the disease fluctuation index is used to generate clinical auxiliary decision-making results for the disease of systemic lupus erythematosus in children. Specifically, extracting the dynamic evolutionary features of the lupus disease course from the set of clinical features includes: Time-series data of a preset list of core indicators for lupus erythematosus are selected from the set of clinical features; Based on the aforementioned time-series data, the dynamic evolution characteristics of the lupus erythematosus disease course were determined; Specifically, generating a feature association map for lupus based on the extracted dynamic evolution features includes: The correlation coefficients between the extracted dynamic evolution characteristics and each laboratory indicator node, immune response node, and organ involvement node were determined. Select strong correlations from all correlation coefficients whose absolute values ​​are greater than a preset significance threshold; A multidimensional feature association matrix is ​​generated by selecting strong correlations; A feature association map of lupus erythematosus is constructed based on the multidimensional feature association matrix; Specifically, the generation of a disease activity prediction curve based on the temporal change gradient of key physiological influencing factors in the aforementioned characteristic population includes: Extract the temporal gradient of key physiological influencing factors in the characteristic population; Predict the distribution of disease activity based on the aforementioned gradient changes; A prediction curve for disease activity is generated based on the predicted distribution of disease activity. Specifically, determining the disease fluctuation index from the disease activity prediction curve includes: Determine the area of ​​integration of the baseline curve representing the core development trend of disease activity in the predicted curve of disease activity; Determine the area of ​​the region enclosed by the upper and lower boundary curves that characterize the range of fluctuations in disease activity in the prediction curve; The disease fluctuation index is determined based on the integral area of ​​the baseline curve and the area of ​​the region.

2. The method as described in claim 1, characterized in that, Determining a standardized set of clinical features based on the structural differences in various data types within the multi-source clinical data specifically includes: Identify the structural differences between unstructured text data and structured numerical data in the multi-source clinical data; Perform a Uniform Medical Language System (UMLS) concept identifier mapping on the unstructured text data, and perform age and gender-based reference interval normalization on the structured numerical data; A standardized set of clinical features is generated from the mapped conceptual identifiers and normalized numerical data.

3. The method as described in claim 1, characterized in that, Based on the aforementioned feature association map, the key physiological influencing factors of lupus erythematosus at different disease stages are identified as follows: Determine the importance score of each node in the feature association graph; All nodes in the feature association graph are sorted based on all importance scores; The key physiological factors affecting lupus erythematosus at different stages of the disease were identified by the ranking results.

4. The method as described in claim 1, characterized in that, Cluster analysis of all key physiological influencing factors identified a high-risk evolutionary trend among lupus patients, specifically including: All key physiological influencing factors were organized chronologically into a multidimensional time series of lupus patients. The multidimensional time series is subjected to population clustering to obtain clustering results; Based on preset high-risk determination rules, the clustering results are used to identify characteristic groups among lupus patients with a high-risk evolutionary trend.

5. The method as described in claim 1, characterized in that, The multi-source clinical data includes laboratory test indicators, immunological test parameters, medical record data, and medical imaging data.

6. A clinical data mining and analysis system for childhood systemic lupus erythematosus, comprising performing clinical data mining and analysis of childhood systemic lupus erythematosus using the method described in any one of claims 1 to 5, characterized in that, The system includes: The acquisition module is used to acquire multi-source clinical data of lupus patients throughout the entire diagnosis and treatment process; The processing module is used to determine a standardized set of clinical features based on the structural differences of various data types in the multi-source clinical data, and then extract the dynamic evolution features of the lupus erythematosus course from the set of clinical features. The processing module is also used to generate a feature association map of lupus based on the extracted dynamic evolution features, and then determine the key physiological influencing factors of lupus at different disease stages based on the feature association map. The processing module is also used to perform cluster analysis on all key physiological influencing factors, screen out characteristic groups with high-risk evolutionary trends among lupus patients, and generate a prediction curve of disease activity based on the temporal change gradient of key physiological influencing factors in the characteristic group. The execution module is used to determine the disease fluctuation index from the prediction curve of the disease activity, and then generate clinical auxiliary decision-making results for the disease status of systemic lupus erythematosus in children through the disease fluctuation index.

Images