An intelligent special disease data association and integration method and system

Abstract

The application relates to the technical field of medical data, and discloses an intelligent special disease data correlation integration method and system.The method comprises the following steps: performing data cleaning on original diagnosis and treatment data of a patient to obtain standardized data; performing real-time screening and judgment on the standardized data to obtain a special disease patient list; extracting special disease data of special disease patients in the special disease patient list from the standardized data, performing index correlation analysis on the special disease data, and obtaining a combined index set; performing multi-dimensional dynamic matching on the combined index set to obtain early warning grading information; performing strategy fusion on the combined index set to generate a personalized intervention scheme for the special disease patients; when the special disease patients are diagnosed and treated according to the personalized intervention scheme, real-time special disease data of the special disease patients is synchronously and continuously collected, and the real-time special disease data is fed back to a special disease screening rule base and the early warning rule base; and the application can improve the efficiency of intelligent special disease data correlation integration.

Description

Technical Field

[0001] This invention relates to the field of medical data technology, and in particular to an intelligent method and system for linking and integrating disease-specific data. Background Technology

[0002] In the disease-specific data processing workflow, existing technologies have significant limitations in cleaning and standardizing patients' original medical data. They cannot accurately correct medical terminology, perform structuring, and filter low-quality data based on the different characteristics of clinical text data, medical imaging data, and molecular sequencing data within the original medical data. This results in standardized output data often exhibiting inconsistent formats, missing or distorted key information, directly creating fundamental obstacles for subsequent disease-specific screening and data extraction, thus affecting the overall effectiveness of data processing.

[0003] Existing technologies also have significant shortcomings in the deep integration and clinical application of disease-specific data. On the one hand, they struggle to construct effective indicator association networks based on the medical correlations between clinical indicators in disease-specific data, and cannot accurately identify indicator nodes and strongly correlated indicator combinations that match the current disease stage of a patient, resulting in a lack of specificity in the extraction of combined indicator sets. On the other hand, in the process of matching early warning rules, they cannot dynamically allocate clinical decision weights for feature components based on the current disease stage of the patient, leading to low accuracy of early warning grading information. Consequently, it is difficult to generate personalized intervention plans through strategy fusion, and the lack of a real-time data feedback mechanism to optimize the disease screening rule base and early warning rule base severely restricts the efficiency and clinical application value of disease-specific data integration. Therefore, how to improve the accuracy and clinical service capabilities of disease-specific data integration has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides an intelligent method and system for linking and integrating disease-specific data to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an intelligent method for linking and integrating disease-specific data, comprising:

[0006] S1. Perform data cleaning on the patient's original medical data to obtain standardized data of the original medical data;

[0007] S2. Based on a preset disease screening rule base, the standardized data is screened and judged in real time to obtain a list of patients with specific diseases.

[0008] S3. Extract the disease-specific data of the disease-specific patients from the disease-specific patient list from the standardized data, and perform indicator correlation analysis on the disease-specific data to obtain the combined indicator set of the disease-specific data;

[0009] S4. Based on a predefined early warning rule base, perform multi-dimensional dynamic matching on the combined indicator set to obtain the early warning classification information of the combined indicator set;

[0010] S5. Based on the early warning classification information, perform strategy fusion on the combined indicator set to generate a personalized intervention plan for the patient with the specific disease.

[0011] S6. When the patient with the specific disease is receiving treatment according to the personalized intervention plan, real-time disease data of the patient with the specific disease is collected synchronously and continuously, and the real-time disease data is fed back to the disease screening rule base and the early warning rule base.

[0012] In a preferred embodiment, the step of cleaning the patient's original medical data to obtain standardized data from the original medical data includes:

[0013] Based on the data source, the patient's original diagnosis and treatment data are divided into clinical text data, medical imaging data, and molecular sequencing data;

[0014] The clinical text data is corrected using medical terminology to obtain standardized text data of the original diagnosis and treatment data.

[0015] The medical image data is structured to obtain unified image data of the original diagnostic and treatment data;

[0016] Low-quality data is filtered from the molecular sequencing data to obtain high-quality molecular data from the original diagnostic and treatment data.

[0017] The standardized text data, the unified image data, and the high-quality molecular data are collected in chronological order to obtain the standardized data of the original diagnostic and treatment data.

[0018] In a preferred embodiment, the standardized data is screened and judged in real time based on a preset disease-specific screening rule base to obtain a list of patients with specific diseases, including:

[0019] The diagnostic conditions, test indicator conditions, examination result conditions, and medication conditions in the preset disease screening rule base are integrated into the screening condition set for the patient;

[0020] Based on the set of screening criteria, patient records in the standardized data that meet the set of screening criteria are selected;

[0021] Clinical phenotype clustering is performed on the patient records to obtain the disease classification results of the patient records;

[0022] The patient records are arranged hierarchically according to the disease classification results to obtain a list of patients with specific diseases.

[0023] In a preferred embodiment, the step of extracting disease-specific data from the disease-specific patient list from the standardized data, and performing indicator correlation analysis on the disease-specific data to obtain a combined indicator set of the disease-specific data, includes:

[0024] Based on the medical correlations between clinical indicators in the disease-specific data, an indicator correlation network for the clinical indicators is constructed.

[0025] Identify indicator nodes in the indicator association network that are related to the current disease stage of the patient with the specific disease;

[0026] Starting from the indicator node, traverse the associated edges in the indicator association network to obtain the indicator combination with strong association in the indicator association network.

[0027] The aforementioned combination of indicators is compiled into a combined indicator set for the specific disease data.

[0028] In a preferred embodiment, constructing an indicator association network among the clinical indicators based on the medical correlations among the clinical indicators in the disease-specific data includes:

[0029] Medical semantic analysis was performed on the clinical indicators to obtain their interaction relationships.

[0030] Based on the aforementioned interaction relationships, the nature and strength of the association between the clinical indicators are determined;

[0031] Using the clinical indicators as nodes, the correlation properties as connecting edges, and the correlation strength as the weight of the connecting edges, an initial network structure is constructed among the clinical indicators.

[0032] By removing spurious associations from the initial network structure, the association network between the clinical indicators is obtained.

[0033] In a preferred embodiment, the step of performing multi-dimensional dynamic matching on the combined indicator set based on a predefined early warning rule base to obtain early warning classification information for the combined indicator set includes:

[0034] The combined index set is arrayed in multiple dimensions to obtain the combined index matrix of the combined index set;

[0035] Eigenvalue decomposition is performed on the combined index matrix to obtain the multidimensional feature vector of the combined index set;

[0036] The difference between the feature components in the multidimensional feature vector and the multidimensional reference standard of the early warning rule is compared to obtain the single-dimensional matching degree of the feature component.

[0037] Based on the current stage of the disease in the patients with the specific disease, assign clinical decision weights to the characteristic components;

[0038] Based on the clinical decision weights, calculate the weighted comprehensive matching degree of the single-dimensional matching degree;

[0039] The weighted comprehensive matching degree is mapped to the early warning threshold range to confirm the early warning classification information of the combined indicator set.

[0040] In a preferred embodiment, the weighted overall matching degree is calculated using the following formula:

[0041] ;

[0042] In the formula, The overall matching degree, For the first Weighting of clinical decision-making in each dimension. For the first Single-dimensional matching degree of each dimension The total number of dimensions, This is the weighted average of the matching degree of the single dimension. This is the preset matching degree adjustment coefficient. For summation operations, This is for square root operations.

[0043] In a preferred embodiment, the step of performing strategy fusion on the combined indicator set based on the early warning classification information to generate a personalized intervention plan for the specific disease patient includes:

[0044] The target elements of the combined indicator set are separated to obtain the core intervention targets and auxiliary regulation indicators of the combined indicator set.

[0045] Based on the aforementioned early warning classification information, select the strategy elements corresponding to the core intervention target from the preset disease-specific intervention plan;

[0046] Based on the clinical importance of the core intervention targets and the regulation coefficients of the auxiliary regulatory indicators, the intervention logic of the strategy elements is reconstructed to obtain the basic intervention plan for the patients with the specific disease.

[0047] The basic intervention plan is clinically validated, and based on the validation results, a personalized intervention plan is generated for the patients with the specific disease.

[0048] In a preferred embodiment, the step of reconstructing the intervention logic of the strategy elements based on the clinical importance of the core intervention target and the adjustment coefficient of the auxiliary regulatory index to obtain the basic intervention plan for the disease-specific patient includes:

[0049] Based on the clinical importance of the core intervention targets and the adjustment coefficients of the auxiliary adjustment indicators, parameter adjustment data for the strategy elements are generated;

[0050] Path simulation is performed on the parameter adjustment data to obtain the path deviation result of the parameter adjustment data;

[0051] Based on the path deviation results, the parameter adjustment data is dynamically corrected to obtain the optimized parameter configuration of the strategy elements;

[0052] The optimized parameters are configured and combined to form the basic intervention plan for the patients with the specific disease.

[0053] To address the above problems, the present invention also provides an intelligent disease-specific data association and integration system, the system comprising:

[0054] The data standardization module is used to clean the patient's original medical data to obtain standardized data from the original medical data.

[0055] The real-time screening module is used to perform real-time screening and judgment on the standardized data based on a preset disease-specific screening rule base to obtain a list of patients with specific diseases.

[0056] The indicator analysis module is used to extract disease-specific data of patients in the disease-specific patient list from the standardized data, and to perform indicator correlation analysis on the disease-specific data to obtain a set of combined indicators for the disease-specific data.

[0057] The early warning matching module is used to perform multi-dimensional dynamic matching of the combined indicator set based on a predefined early warning rule base to obtain the early warning classification information of the combined indicator set.

[0058] The strategy fusion module is used to perform strategy fusion on the combined indicator set based on the early warning classification information to generate a personalized intervention plan for the patient with the specific disease.

[0059] The feedback optimization module is used to synchronously and continuously collect real-time disease data of the patients with the specific disease when they receive diagnosis and treatment according to the personalized intervention plan, and to feed the real-time disease data back to the disease screening rule base and the early warning rule base.

[0060] Compared with the prior art, the present invention has the following beneficial effects:

[0061] 1. This invention effectively generates standardized data with uniform format and accurate information by performing medical terminology correction, structuring processing, and low-quality data filtering on original medical data from multiple sources of patients according to their types, providing a high-quality foundation for subsequent disease-specific data processing. Furthermore, based on a preset disease-specific screening rule base, screening conditions are integrated and clinical phenotype clustering is combined to obtain an accurate list of disease-specific patients. By constructing a clinical indicator association network to extract a set of strongly correlated combined indicators that match the disease course, the accuracy of disease-specific data extraction and the correlation of indicators are significantly improved.

[0062] 2. This invention performs multidimensional dynamic matching of a set of combined indicators, assigns clinical decision weights based on the patient's current disease course, and accurately calculates the weighted comprehensive matching degree to determine early warning classification information, significantly improving the accuracy of early warning judgment. At the same time, it separates core intervention targets and auxiliary regulatory indicators, reconstructs intervention logic to generate personalized plans, and collects data in real time to optimize the rule base, which not only enhances the pertinence of intervention plans, but also continuously improves the efficiency of disease-specific data association and integration and clinical application value. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating an intelligent method for linking and integrating disease-specific data, provided in an embodiment of the present invention.

[0064] Figure 2 A functional module diagram of an intelligent disease-specific data association and integration system provided in an embodiment of the present invention;

[0065] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0066] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0067] This application provides an intelligent method for linking and integrating disease-specific data. The executing entity of this intelligent method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the intelligent method for linking and integrating disease-specific data can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0068] Reference Figure 1 The diagram shown is a flowchart illustrating an intelligent method for linking and integrating disease-specific data according to an embodiment of the present invention. In this embodiment, the intelligent method for linking and integrating disease-specific data includes:

[0069] S1. Perform data cleaning on the patient's original medical data to obtain standardized data of the original medical data;

[0070] In this embodiment of the invention, the step of cleaning the patient's original medical data to obtain standardized data from the original medical data includes:

[0071] Based on the data source, the patient's original diagnosis and treatment data are divided into clinical text data, medical imaging data, and molecular sequencing data;

[0072] The clinical text data is corrected using medical terminology to obtain standardized text data of the original diagnosis and treatment data.

[0073] The medical image data is structured to obtain unified image data of the original diagnostic and treatment data;

[0074] Low-quality data is filtered from the molecular sequencing data to obtain high-quality molecular data from the original diagnostic and treatment data.

[0075] The standardized text data, the unified image data, and the high-quality molecular data are collected in chronological order to obtain the standardized data of the original diagnostic and treatment data.

[0076] First, we identified the sources and characteristics of the original medical data collection, and clarified the data classification rules. Data obtained from written medical records such as outpatient medical records, inpatient medical records, doctor's orders, progress notes, and examination request forms, containing information such as patient symptom descriptions, disease diagnoses, medication regimens, and treatment operation records, were classified as clinical text data. Data exported from medical imaging equipment such as spiral CT, 1.5T MRI, color Doppler ultrasound, digital X-ray machines (DR), and PET-CT, stored in the form of image sequences and DICOM files, reflecting the patient's organ function, was classified as clinical text data. Data containing information on structure, tissue lesions, and physiological functions are classified as medical imaging data. Data generated from molecular detection instruments such as second-generation sequencers (NGS), third-generation sequencers (such as PacBio sequencers), and gene chip detection equipment, which are presented in the form of base sequence files, gene expression matrices, and variant detection results, and record information such as patient gene sequences, protein structures, and molecular markers, are classified as molecular sequencing data. The entire classification process requires checking the collection channel and data format of each piece of raw data one by one to ensure that the classification of each type of data is accurate and there is no cross-mixing or omission in the classification.

[0077] First, the clinical text data was scanned sentence by sentence to extract all medical terms, covering disease names such as coronary atherosclerotic heart disease and type 2 diabetes, symptom terms such as chest pain and polydipsia / polyuria, drug names such as enteric-coated aspirin and insulin injection, diagnostic and treatment terms such as percutaneous coronary intervention and blood glucose monitoring, and examination and testing terms such as complete blood count and myocardial enzyme profile. Then, based on authoritative sources such as the 11th revision of the International Classification of Diseases (ICD-11), the WHO Collaborating Centre for International Drug Monitoring (WHO-ART), and the National Standard for Clinical Laboratory Procedure Terminology, the extracted medical terms were verified one by one. Any non-standard expressions were corrected, such as correcting myocardial infarction to acute myocardial infarction, penicillin injection to injectable penicillin sodium, and blood Rt expanded to complete blood count. After correction, the clinical text data was organized and archived according to a unified format of patient ID-data collection time-term type-standard term content-term corresponding diagnosis and treatment scenario to form standardized text data, ensuring that all medical terms in the text conform to authoritative standards, are uniformly expressed, and have complete related information.

[0078] First, the raw storage file of the medical image data is read, and the file format is identified, such as DICOM 3.0, NIFT1-1, JPEG2000, etc. Basic patient information, including patient ID, name, gender, and age, and examination-related information, including the department, examination items, examination date and time, imaging equipment model, scanning parameters such as tube voltage, tube current, slice thickness, and matrix size, are extracted. Imaging technical information, including image resolution, pixel depth, number of images, and tomographic location, is also extracted. Next, the image data is preprocessed. A Gaussian filtering algorithm is used to remove high-frequency noise caused by equipment circuit noise and signal interference. A median filtering algorithm is used to eliminate salt-and-pepper artifacts in the images. Finally, histogram equalization is used to adjust the image grayscale. Contrast and brightness are adjusted to ensure that images of the same area acquired by different devices and in different scanning batches maintain consistency in visual grayscale range and detail clarity. Then, the data is integrated according to a preset structured template, which includes fixed fields such as patient information area, examination parameter area, image storage area, and quality verification area. The extracted basic information is filled into the corresponding patient information area and examination parameter area. The preprocessed image data is stored in the image storage area in tomographic or sequential order. The quality verification area marks the quality assessment results of the image preprocessing, such as noise removal and contrast meeting the standard. Finally, unified image data with uniform format, complete basic information, and stable image quality is formed, which meets the consistency requirements of image data format and quality for subsequent disease-specific data association analysis.

[0079] First, based on molecular sequencing technical specifications and industry quality standards, specific criteria for judging low-quality molecular sequencing data are established. These include a base quality value below Q30 (corresponding to a base identification error rate higher than 0.1%), a sequence length shorter than the preset effective length for the sequencing project (e.g., an effective sequence length of no less than 100 bp for whole exome sequencing and no less than 18 bp for small RNA sequencing), fuzzy bases (bases that cannot be clearly identified, represented by N) accounting for more than 5%, the presence of repetitive sequences with 10 or more consecutive identical bases, and a sequence alignment rate with the reference genome below 80%. Then, a sequence-by-sequence detection method is used to verify each original sequence in the molecular sequencing data item by item. Each sequence is first read using base quality scoring software. The sequence was screened based on its base quality value, selecting sequences with a base quality value of Q30 or higher. The actual length of each sequence was then calculated, and sequences shorter than the preset effective length were removed. Next, the percentage of ambiguous bases in each sequence was calculated, and sequences exceeding this percentage were excluded. Then, the presence of continuous repetitive base fragments was detected, and sequences meeting the repetitive sequence criteria were removed. Finally, the remaining sequences were compared with the corresponding reference genome, retaining sequences with an alignment rate of 80% or higher. After all detection steps were completed, the retained qualified sequences were organized into a high-quality molecular data set in the format of sequencing sample ID-sequence number-sequence region-base sequence-quality score-alignment result, ensuring a reliable data foundation for subsequent disease-specific index analysis based on molecular data.

[0080] First, time information was extracted from standardized text data, unified imaging data, and high-quality molecular data respectively. Time information from standardized text data was extracted from the text content, such as the visit time in outpatient medical records, the recording time in inpatient progress notes, the issuance time of medical orders, and the report generation time of examination reports. Time information from unified imaging data was extracted from the examination date and time fields in its basic information, accurate to the minute. Time information from high-quality molecular data was obtained from sequencing experiment records, including sample receipt time, sequencing start time, and sequencing completion time, with the sequencing completion time used as the core time identifier for this type of data. Then, a timeline index table was established based on the chronological order of the data, and the three types of data were sorted according to their respective... The core time information is mapped to specific nodes on the timeline. For different types of data collected at the same time node, such as the standardized text data corresponding to the blood routine examination and the unified image data corresponding to the chest CT examination of the same patient at 9:30 on May 10, 2024, a connection and binding are established through dual identification of patient ID and time node. This ensures that the diagnosis and treatment information at the same time is completely corresponding. Finally, in the order from morning to evening on the timeline, all the standardized text data, unified image data and high-quality molecular data that have been associated and sorted are integrated into a continuous data set with a clear timeline. This data set is the standardized data of the original diagnosis and treatment data, which fully presents the various data information of the patient at different stages of diagnosis and treatment.

[0081] The beneficial effects are as follows: by accurately classifying and specifically processing clinical text data, medical imaging data, and molecular sequencing data according to their sources, standardized text data with standardized terminology, unified imaging data with unified format, and high-quality molecular data that meets quality standards are formed respectively, ensuring the standardization and reliability of various types of data from the source. Then, the three types of data are systematically aggregated into standardized data of original diagnosis and treatment data in chronological order, clearly sorting out the timeline of the patient's diagnosis and treatment process. This provides a consistent, complete, and reliable basic data for subsequent steps such as disease screening and judgment, disease data extraction, and indicator correlation analysis, effectively avoiding subsequent processing deviations caused by chaotic or distorted original data, and significantly improving the accuracy and efficiency of disease-specific data correlation and integration.

[0082] S2. Based on a preset disease screening rule base, the standardized data is screened and judged in real time to obtain a list of patients with specific diseases.

[0083] In this embodiment of the invention, the step of performing real-time screening and judgment on the standardized data based on a preset disease-specific screening rule base to obtain a list of patients with specific diseases includes:

[0084] The diagnostic conditions, test indicator conditions, examination result conditions, and medication conditions in the preset disease screening rule base are integrated into the screening condition set for the patient;

[0085] Based on the set of screening criteria, patient records in the standardized data that meet the set of screening criteria are selected;

[0086] Clinical phenotype clustering is performed on the patient records to obtain the disease classification results of the patient records;

[0087] The patient records are arranged hierarchically according to the disease classification results to obtain a list of patients with specific diseases.

[0088] First, the specific content of the pre-defined disease screening rule base, including diagnostic conditions, laboratory indicator conditions, examination result conditions, and medication conditions, is clearly defined. The diagnostic conditions cover the International Classification of Diseases (ICD) code for the corresponding disease, specific symptom combinations, and symptom duration requirements. The laboratory indicator conditions include the normal reference range and abnormal thresholds of blood biochemical and immune indicators related to the disease. The examination result conditions involve the judgment criteria such as characteristic imaging manifestations and pathological examination results of the disease. The medication conditions specify the types of specific drugs that the patient has used related to the treatment or diagnosis of the disease and the required duration of medication. Then, the four types of conditions are linked and integrated according to clinical diagnosis and treatment logic. For conditions that need to be met simultaneously, logical and relational connections are used. For conditions that only require the satisfaction of one, logical and relational connections are used to ensure that all conditions in the rule base are included and that the logical relationships are clear and complete. Finally, a set of screening conditions for screening patients is formed.

[0089] First, complete data for each patient is read one by one from the standardized data of the acquired original diagnostic and treatment data. This standardized data includes standardized text data, unified imaging data, and high-quality molecular data. Then, based on the integrated set of screening criteria, the standardized data of each patient is checked item by item. The diagnostic criteria check whether the disease diagnosis record in the standardized text data meets the coding or symptom requirements in the rule base. The test indicator criteria check whether the test values ​​recorded in the standardized text data are within the abnormal threshold range set by the rule base. The examination result criteria check whether the unified imaging data presents the disease-specific characteristic manifestations specified in the rule base. The medication criteria check whether the medication record in the standardized text data contains the drugs specified in the rule base and meets the medication duration. Only patient data that passes all the checks is retained, ultimately forming patient records that meet the set of screening criteria.

[0090] First, clinical phenotype information for each patient is extracted from the screened patient records. This information includes symptom descriptions, laboratory test values, and medication history from standardized text data, lesion size and morphological characteristics from unified imaging data, and disease-specific molecular marker performance from high-quality molecular data. Then, all patient records are grouped according to the similarity of clinical phenotypes. Patient records with highly similar symptom presentations, consistent trends in laboratory test values, similar imaging features, and similar molecular marker performances are grouped into the same group. During the grouping process, the differences in clinical phenotypes within each group are minimized, while the differences in clinical phenotypes between different groups are maximized. Finally, the disease classification result corresponding to each patient record is obtained.

[0091] First, based on the disease classification results obtained from clinical phenotype clustering, patient records belonging to the same disease category are grouped together. Then, for patient records within the same disease category, combined with disease severity-related indicators recorded in standardized text data, such as the degree of abnormality of test indicators and the frequency and duration of symptom onset, the patient records are further ordered. Patient records with more severe conditions are placed at the beginning of the category, and those with milder conditions are placed at the end. Finally, all patient records grouped by disease category and ordered within the group are compiled into a structured list. The list contains key clinical phenotype information that uniquely identifies the patient's disease category and disease severity-related data, ultimately yielding a disease-specific patient list.

[0092] The beneficial effects include: integrating multiple conditions from the disease-specific screening rule base to form a logically clear set of screening conditions, providing a comprehensive and accurate basis for patient screening; accurately selecting patient records that meet the requirements using the set of screening conditions, avoiding interference from irrelevant data; improving the scientific nature of classification by making disease classification more consistent with the actual clinical characteristics of patients through clinical phenotype clustering; and forming a disease-specific patient list by stratifying the disease classification results, clearly presenting the distribution and severity of patient symptoms, facilitating subsequent disease-specific data extraction and treatment management. Overall, it significantly improves the efficiency and accuracy of disease-specific patient screening and classification, providing a high-quality patient data foundation for subsequent disease-specific data association and integration.

[0093] S3. Extract the disease-specific data of the disease-specific patients from the disease-specific patient list from the standardized data, and perform indicator correlation analysis on the disease-specific data to obtain the combined indicator set of the disease-specific data;

[0094] In this embodiment of the invention, the step of extracting disease-specific data from the disease-specific patient list from the standardized data, and performing indicator correlation analysis on the disease-specific data to obtain a combined indicator set of the disease-specific data, includes:

[0095] Based on the medical correlations between clinical indicators in the disease-specific data, an indicator correlation network for the clinical indicators is constructed.

[0096] Identify indicator nodes in the indicator association network that are related to the current disease stage of the patient with the specific disease;

[0097] Starting from the indicator node, traverse the associated edges in the indicator association network to obtain the indicator combination with strong association in the indicator association network.

[0098] The aforementioned combination of indicators is compiled into a combined indicator set for the specific disease data.

[0099] The step of constructing an indicator association network among clinical indicators based on the medical correlations among clinical indicators in the disease-specific data includes:

[0100] Medical semantic analysis was performed on the clinical indicators to obtain their interaction relationships.

[0101] Based on the aforementioned interaction relationships, the nature and strength of the association between the clinical indicators are determined;

[0102] Using the clinical indicators as nodes, the correlation properties as connecting edges, and the correlation strength as the weight of the connecting edges, an initial network structure is constructed among the clinical indicators.

[0103] By removing spurious associations from the initial network structure, the association network between the clinical indicators is obtained.

[0104] First, from the standardized data of the acquired original medical records, for each patient in the disease-specific patient list, information directly related to the target disease is extracted: Symptom descriptions corresponding to the disease are extracted from standardized text data, such as the frequency and duration of dizziness and headaches for hypertension; laboratory indicator values, such as systolic blood pressure, diastolic blood pressure, lipid profile, blood glucose; medication records showing the names, dosages, and frequencies of drugs used for disease treatment; and medical records showing examinations related to the disease, such as 24-hour ambulatory blood pressure monitoring or treatment, such as adjustments to antihypertensive medications. Characteristic imaging information of the disease is extracted from standardized imaging data, such as the location, extent, and density characteristics of ischemic lesions shown on brain CT or MRI for stroke. Molecular marker data related to the disease is extracted from high-quality molecular data, such as BRCA12 gene mutation status and HER2 gene expression levels for breast cancer. All this information closely related to the disease extracted from these three types of standardized data is summarized and organized to form disease-specific data for each patient, ensuring that each set of disease-specific data fully covers the key information at the text, imaging, and molecular levels without any irrelevant data being mixed in.

[0105] First, identify the types of clinical indicators extracted from the disease-specific data. These include laboratory indicators such as white blood cell count in a complete blood count, liver function indicators in biochemical tests, thyroid function indicators in endocrine tests, symptom indicators such as pain severity rating, fever duration, and frequency of dyspnea episodes, imaging indicators such as the size and number of tumor lesions, the clarity of their boundaries, and the degree of displacement at the fracture site, and molecular indicators such as gene mutation type, gene expression level, and protein concentration. Then, refer to authoritative medical guidelines in the disease-specific field, such as the hypertension prevention and treatment guidelines and diabetes diagnosis and treatment guidelines published by the Chinese Medical Association, as well as core medical literature such as disease-related clinical research papers and standardized clinical diagnosis and treatment pathways included in PubMedCNKI. Perform medical semantic analysis on the relationship between each pair of clinical indicators. For example, when analyzing fasting blood glucose and glycated hemoglobin, the guidelines clearly state that both are used to assess blood glucose control. Fasting blood glucose reflects the immediate blood glucose level, while glycated hemoglobin reflects the average blood glucose level over the past 23 months. This establishes an interaction between the two, reflecting the state of blood glucose control. When analyzing tumor size and carcinoembryonic antigen (CEA) concentration, clinical research literature clearly shows an interaction between the two, where tumor volume increase is often accompanied by an increase in CEA concentration. Through this step-by-step analysis, the interaction between all clinical indicators is obtained.

[0106] Based on the aforementioned interactions between clinical indicators, the nature and strength of the association between each pair of clinical indicators are further determined: the nature of the association is categorized as positive, negative, and neutral. A positive association refers to an abnormal change in one indicator accompanied by a similar abnormal change in another indicator, such as elevated blood pressure accompanied by an increased incidence of hypertension or heart disease. A negative association refers to an abnormal change in one indicator accompanied by a similar abnormal change in another indicator, such as elevated insulin levels accompanied by decreased blood glucose levels. A neutral association refers to no significant similar or similar changes between the two indicators, but they need to be considered together in the diagnosis and treatment of specific diseases, such as two independent molecular markers for certain tumors. The strength of the association is determined based on the level of clinical evidence supporting the interaction. Associations supported by randomized controlled trials are classified as strong associations, those supported by cohort studies are classified as moderate associations, and those supported by case-control studies are classified as weak associations. Furthermore, the frequency of the association in clinical practice is considered. If the association is observed in more than 80% of patients with a specific disease, the strength is increased by one level; if it is observed only in less than 20% of patients, the strength is decreased by one level. Through this comprehensive assessment, the nature and strength of the association between each pair of clinical indicators are clarified.

[0107] Each clinical indicator is treated as an independent node in the indicator association network. Node information must include indicator name, indicator type, test results, symptoms, imaging, molecular patterns, normal reference range, and disease-related abnormal thresholds. The determined association nature is used as the connecting edge between two nodes: positive association is represented by a solid line, negative association by a dashed line, and neutral association by a dotted line. The association strength is used as the weight of the connecting edge: strong association is weighted at 3, medium association at 2, and weak association at 1. First, all nodes are initially grouped and arranged according to indicator type, such as clustering indicator nodes and clustering symptom indicator nodes. Then, based on the association relationship between each pair of indicators, the relevant nodes are connected with connecting edges of the corresponding type, and the corresponding weight values ​​are marked on the connecting edges. For example, the fasting blood glucose node and the glycated hemoglobin node are connected by a solid line with a weight of 3, and the insulin node and the blood glucose node are connected by a dashed line with a weight of 3. This constructs the initial network structure between clinical indicators, ensuring that each node can establish an association with relevant nodes through connecting edges, and that the type and weight of the connecting edges accurately reflect the association nature and strength.

[0108] First, a large-scale clinical database corresponding to the specific disease is retrieved. This database must contain complete diagnostic and treatment data of at least 1000 confirmed patients with the specific disease, including laboratory, symptom, imaging, and molecular data from multiple years of follow-up. Each pair of correlation indicators in the initial network structure is validated in this database, and the proportion of patients showing a corresponding correlation between these indicators is calculated. If the proportion is less than 30% and there is no other authoritative guidelines or high-level research evidence to support it, it is considered a spurious correlation. Next, three or more clinical experts with the title of associate chief physician or above in the field of this specific disease are invited to conduct a clinical rationality assessment of each correlation in the initial network structure. The experts review the literature supporting the correlations. Based on their own clinical experience, if a certain correlation is unanimously considered to have no actual diagnostic or therapeutic significance—for example, a correlation between two indicators with no pathophysiological relationship due to accidental data fluctuations—it is judged as a spurious correlation. All edges judged as spurious correlations are removed from the initial network structure. At the same time, it is checked whether there are any isolated nodes after deletion, i.e., nodes without any edges. If the isolated node has no independent reference value in the diagnosis and treatment of specific diseases—for example, a symptom indicator that only appears in a very few patients and has no diagnostic significance—it is also deleted. Finally, an indicator correlation network between clinical indicators is obtained, ensuring that all correlations in the network are supported by sufficient clinical evidence and have actual diagnostic and therapeutic reference significance.

[0109] First, determine the current stage of the patient's disease. The division of disease stages is based on multi-dimensional information in the disease data: duration of symptoms, such as a cough lasting more than 2 weeks, which may indicate that the lung disease has entered the chronic stage; trends in laboratory indicators, such as blood sugar changing from occasional elevation to continuous elevation, which indicates that diabetes has entered the acute stage; progression of imaging lesions, such as the development of tumor lesions from being confined to the primary site to distant metastasis, which indicates that it has entered the late stage; treatment response, such as blood pressure remaining above 140 / 90 mmHg after using antihypertensive drugs, which indicates that hypertension is poorly controlled. For example, diabetes can be divided into the pre-acute stage and the chronic complication stage and the remission stage. Based on the patient's specific data, the stage of the disease is determined. Then, in the constructed indicator association network, indicator nodes related to the key points of diagnosis and treatment at that stage are screened out. For example, key indicator nodes for the chronic complications stage of diabetes include glycated hemoglobin reflecting long-term blood glucose control, urinary microalbumin reflecting early kidney damage, serum creatinine reflecting kidney function, and retinopathy imaging indicators reflecting eye complications. Key indicator nodes for stage IV lung cancer include tumor marker carcinoembryonic antigen CYFRA211, systemic metastatic lesion imaging indicator system status score, and related symptom indicators. Through this targeted screening, indicator nodes related to the current stage of the disease in patients with specific diseases are identified.

[0110] Starting with each identified indicator node related to the current disease stage, the system traverses outwards along the connection edges in the indicator association network. During the traversal, nodes corresponding to strong association edges with a weight of 3 are prioritized, followed by nodes corresponding to medium association edges with a weight of 2. Nodes corresponding to weak association edges with a weight of 1 are not included in the traversal. During the traversal, the system records in detail the strong association nodes associated with each starting node, forming a preliminary indicator combination. For example, starting with the glycated hemoglobin node, the traversal extends to fasting blood glucose (strong association weight 3), 2-hour postprandial blood glucose (strong association weight 3), and urinary microbiota (strong association weight 3). If the correlation weight in albumin is 2, the initial indicator combination is glycated hemoglobin plus fasting blood glucose plus 2-hour postprandial blood glucose. Then, the clinical significance of the initial indicator combination is verified. Specialty clinical experts are invited to evaluate whether the combination can jointly reflect a key pathophysiological process or disease characteristic of the specialty. If the above combination can jointly reflect the patient's long-term blood glucose control and early kidney damage risk, the verified combination is determined to be an indicator combination with strong correlation. If only a combination of 2 indicators can be formed after traversing a certain starting node, one additional correlational and clinically significant indicator is added to ensure that the combination has sufficient diagnostic and treatment reference value.

[0111] All strongly correlated indicator combinations are summarized, and each combination is checked for duplicates, such as identical combinations obtained after traversing different starting nodes. If duplicates exist, only one is retained. Each indicator combination is labeled with its corresponding clinical significance. For example, glycated hemoglobin plus fasting blood glucose plus 2-hour postprandial blood glucose is labeled as a combination for assessing blood glucose control and early kidney damage risk; tumor size plus carcinoembryonic antigen plus CYFRA211 is labeled as a combination for monitoring lung cancer lesion progression and treatment efficacy; systolic blood pressure plus diastolic blood pressure plus ambulatory blood pressure variability is labeled as a combination for assessing the severity of hypertension. Then, all indicator combinations are sorted according to the pathophysiological process of the specific disease, from etiology-related indicator combinations to symptom-related indicator combinations to complication-related indicator combinations, forming a clearly structured and logically coherent set of specific disease data combination indicators. This ensures that each combination in the set accurately corresponds to a key assessment dimension in the diagnosis and treatment of the specific disease, providing comprehensive and accurate indicator basis for subsequent early warning rule matching and personalized intervention plan development.

[0112] The beneficial effects are as follows: by accurately extracting disease-specific data from standardized data, the data is highly correlated with the target disease and free of redundant information; when constructing the indicator association network, the scientific nature and strength of the association are judged by medical semantic analysis and spurious associations are removed, ensuring the scientificity and reliability of the association relationships within the network; the indicator nodes related to the current disease stage are identified and strongly correlated indicator combinations are obtained by traversing them, so that the indicator combinations fit the actual stage of the patient's condition and have clear clinical significance; the final aggregated combined indicator set comprehensively covers the key dimensions of disease diagnosis and treatment, providing accurate data support for subsequent early warning grading and personalized intervention, significantly improving the accuracy and practicality of disease-specific data indicator association analysis, and promoting the efficient transformation of disease-specific diagnosis and treatment data into clinical decision-making basis.

[0113] S4. Based on a predefined early warning rule base, perform multi-dimensional dynamic matching on the combined indicator set to obtain the early warning classification information of the combined indicator set;

[0114] In this embodiment of the invention, the step of performing multi-dimensional dynamic matching on the combined indicator set based on a predefined early warning rule base to obtain the early warning classification information of the combined indicator set includes:

[0115] The combined index set is arrayed in multiple dimensions to obtain the combined index matrix of the combined index set;

[0116] Eigenvalue decomposition is performed on the combined index matrix to obtain the multidimensional feature vector of the combined index set;

[0117] The difference between the feature components in the multidimensional feature vector and the multidimensional reference standard of the early warning rule is compared to obtain the single-dimensional matching degree of the feature component.

[0118] Based on the current stage of the disease in the patients with the specific disease, assign clinical decision weights to the characteristic components;

[0119] Based on the clinical decision weights, calculate the weighted comprehensive matching degree of the single-dimensional matching degree;

[0120] The weighted comprehensive matching degree is mapped to the early warning threshold range to confirm the early warning classification information of the combined indicator set.

[0121] The formula for calculating the weighted overall matching degree is as follows:

[0122] ;

[0123] In the formula, The overall matching degree, For the first Weighting of clinical decision-making in each dimension. For the first Single-dimensional matching degree of each dimension The total number of dimensions, This is the weighted average of the matching degree of the single dimension. This is the preset matching degree adjustment coefficient. For summation operations, This is for square root operations.

[0124] First, we sort out all the disease-related indicator combinations included in the combined indicator set, and clarify the clinical assessment dimensions corresponding to each indicator combination. For example, for diabetes, there are blood glucose control assessment combinations, renal function impairment assessment combinations, retinopathy assessment combinations, and cardiovascular risk assessment combinations. Among them, the blood glucose control assessment combination includes fasting blood glucose value in millimoles per liter, 2-hour postprandial blood glucose value in millimoles per liter, and glycated hemoglobin value in percentage. The renal function impairment assessment combination includes serum creatinine value in micromoles per liter, urinary microalbumin value in milligrams per liter, and estimated glomerular filtration rate in milliliters per minute per 1.73 square meters of body surface area. Each clinical assessment dimension is treated as an independent row in the combined indicator matrix. Then, according to the patient order in the disease-specific patient list, all specific indicator values ​​for each patient under each clinical assessment dimension are extracted and organized into columns of the matrix according to the correspondence between patient ID and indicator value. For example, if patient A's values ​​under the blood glucose control assessment dimension are 5.2, 7.1, and 5.6, these values ​​are then filled into the cells corresponding to the row of the blood glucose control assessment dimension and the column corresponding to patient A in sequence. This ensures that the data in each cell accurately corresponds to a specific clinical assessment dimension, a specific patient, and a specific indicator. The final result is a combined indicator matrix in which rows represent clinical assessment dimensions, columns represent patients, and cells store specific indicator values. All data in the matrix has a uniform format and clear relationships.

[0125] First, the data in each column of the combined indicator matrix is ​​centered. The mean of all indicator values ​​in each column is calculated. For example, if the data in a column is 5.2, 6.1, 5.8, and 6.3, its mean is the sum of 5.2, 6.1, 5.8, and 6.3 divided by 4, which equals 5.85. Then, each indicator value in that column is subtracted from the mean of the corresponding column to obtain the centered data as -0.65, 0.25, -0.05, and 0.45. This step eliminates the magnitude bias caused by individual differences in patient data and avoids excessive interference from high values ​​of a particular patient in the overall analysis. Next, calculate the covariance between any two columns of data in the centered matrix. Taking the centered data of the first column (-0.65, 0.25, -0.05, 0.45) and the centered data of the second column (-0.5, 0.3, 0.1, 0.1) as an example, first multiply the centered values ​​at corresponding positions in the two columns. This gives -0.65 multiplied by -0.5 equals 0.325, 0.25 multiplied by 0.3 equals 0.075, -0.05 multiplied by 0.1 equals -0.005, and 0.45 multiplied by 0.1 equals 0.045. Summing these products gives 0.325 plus 0.075 minus 0.005 plus 0.045 equals 0.44. Dividing this by the number of data points minus one (4 minus 1) equals 3, resulting in a covariance of approximately 0.147. Calculate the covariance between all columns in the matrix in this way to construct the complete covariance matrix. Subsequently, the numerical distribution of the covariance matrix was analyzed, and several eigenvalues ​​with the largest values ​​in the covariance matrix were selected. Each eigenvalue corresponds to an eigenvector that can reflect the core trend of data change. For example, the top three eigenvalues ​​with the largest values ​​were selected, and their corresponding eigenvectors reflected the main change patterns of blood glucose control, kidney function, and cardiovascular risk, respectively. These eigenvectors were arranged in descending order of their corresponding eigenvalues ​​to form a multidimensional eigenvector that can accurately represent the core features of the combined indicator set and whose dimensions are consistent with the number of selected eigenvalues.

[0126] First, a multi-dimensional reference standard corresponding to each feature component in the multi-dimensional feature vector is retrieved from a predefined early warning rule base. Each feature component's reference standard is based on authoritative clinical guidelines and extensive historical diagnostic data in the specific disease area. It includes the ideal value of the feature component in a healthy state, the acceptable range of fluctuation in clinical practice, and the critical value indicating abnormal conditions. For example, a feature component reflecting kidney function, namely the estimated glomerular filtration rate, has an ideal value of 90 ml / min per 1.73 m² body surface area, an acceptable fluctuation range of 60 to 120 ml / min per 1.73 m² body surface area, and an abnormal critical value of 50 ml / min per 1.73 m² body surface area. Next, calculate the difference between the actual value of each feature component and the ideal value in the reference standard. If the difference is within the allowable fluctuation range, it is mapped to a value between 0.8 and 1 based on the proportion of the difference to the allowable fluctuation range. For example, if the actual value is 85 and the difference between it and the ideal value of 90 is 5, the allowable fluctuation range is 60, which is 120 minus 60. The difference accounts for approximately 8.3%, so it is mapped to 1 minus 0.083, which is approximately 0.917. If the difference exceeds the allowable fluctuation range but does not reach the abnormal threshold, for example, if the actual value is 55 and the difference is 35, exceeding the lower limit of the allowable range by 5, which is 60 minus 55, it is mapped to 0.65, which is between 0.5 and 0.8. If the difference reaches or exceeds the abnormal threshold, for example, if the actual value is 45 and the difference is 45, reaching the threshold by 5, which is 50 minus 45, it is mapped to 0.3, which is between 0 and 0.5. Through this mapping process, the degree of fit between each feature component and the reference standard is accurately quantified, and the single-dimensional matching degree corresponding to each feature component is obtained.

[0127] First, based on standardized data of patients with specific diseases, we comprehensively determine their current disease stage. Specifically, we combine information such as the duration of symptoms (e.g., whether cough lasts for more than 2 weeks), the trend of changes in test indicators (e.g., whether blood sugar levels rise occasionally or continuously), the progression of lesions in imaging data (e.g., whether tumor lesions are localized to the primary site or have metastasized), and the treatment response in medical records (e.g., whether blood pressure drops to the normal range after using antihypertensive drugs). Patients with specific diseases are then divided into different disease stages. For example, diabetic patients are divided into pre-diabetes, acute phase, chronic complication phase, and remission phase, while cancer patients are divided into early, middle, late, and recovery phases. Then, a clinical decision weight allocation standard that perfectly corresponds to the current disease stage is retrieved from a predefined early warning rule base. This standard clarifies the importance level of different feature components in the diagnosis and treatment decisions at this stage. For example, in the acute phase of diabetes, the clinical decision weight of feature components directly related to rapid disease progression, such as the blood ketone feature component reflecting the risk of ketoacidosis, is set to 0.4; the feature component directly related to blood glucose control is set to 0.3; the auxiliary feature component related to nutritional status assessment is set to 0.1; the auxiliary feature component related to lifestyle is set to 0.1; and the feature component related to other secondary assessment dimensions is set to 0.1. The sum of the weight values ​​of all feature components is fixed at 1. If the patient is in the remission phase of diabetes, the weight of feature components related to complication monitoring will be adjusted to 0.35; the weight of feature components related to blood glucose stability will be adjusted to 0.25; and the weight of feature components related to lifestyle intervention will be adjusted to 0.3. This ensures that the weight allocation is in line with the diagnosis and treatment focus of the current disease stage, and finally, a specific and fixed clinical decision weight value is assigned to each feature component.

[0128] First, calculate the product of the clinical decision weight and the corresponding single-dimensional matching degree for each feature component. For example, if a feature component has a weight of 0.4 and a matching degree of 0.8, its product is 0.32; if the weight is 0.3 and the matching degree is 0.9, the product is 0.27; if the weight is 0.1 and the matching degree is 0.7, the product is 0.07; if the weight is 0.1 and the matching degree is 0.65, the product is 0.065; and if the weight is 0.1 and the matching degree is 0.75, the product is 0.075. Add these products together for all feature components to get a total of 0.32 + 0.27 + 0.07 + 0.065 + 0.075 = 0.8. Then calculate the sum of the clinical decision weights for all feature components, which is 0.4 + 0.3 + 0.1 + 0.1 + 0.1 = 1. Dividing the sum of the former by the sum of the latter, we get the weighted average matching degree of the single-dimensional matching degree as 0.8 divided by 1, which equals 0.8. Next, we calculate the correction factor used to adjust the result. First, we calculate the difference between the single-dimensional matching degree and the weighted average matching degree of each feature component. In the example above, the differences are 0 (0.8 minus 0.8), 0.1 (0.9 minus 0.8), -0.1 (0.7 minus 0.8), -0.15 (0.65 minus 0.8), and -0.05 (0.75 minus 0.8). We then square each difference to get 0, 0.01, 0.01, 0.0225, and 0.0025. Finally, we multiply each squared result by the corresponding feature component. The clinical decision weights yield the following values: 0 x 0.4 = 0, 0.01 x 0.3 = 0.003, 0.01 x 0.1 = 0.001, 0.0225 x 0.1 = 0.00225, and 0.0025 x 0.1 = 0.00025. Adding these results together gives 0 + 0.003 + 0.001 + 0.00225 + 0.00025 = 0.0065. Dividing this sum by the sum of the clinical decision weights (1) gives 0.0065. Taking the square root of 0.0065 gives approximately 0.0806. The matching degree adjustment coefficient corresponding to the specific disease is retrieved from the predefined early warning rule base. For example, the adjustment coefficient for diabetes is 0.8. The square root result is multiplied by the adjustment coefficient, resulting in 0.0806 multiplied by 0.8, which is approximately 0.0645. This value is then subtracted from 1, yielding a correction factor of 1 minus 0.0645, which is approximately 0.9355. Finally, the weighted average matching degree is multiplied by the correction factor, i.e., 0.8 multiplied by 0.9355, which is approximately 0.748, to obtain the weighted comprehensive matching degree of the combined indicator set.

[0129] First, the system retrieves the warning threshold range that perfectly corresponds to the current specific disease from a predefined warning rule base. This range is based on a large number of clinical cases and evidence-based medicine data related to the specific disease. Different warning levels are corresponding to the weighted comprehensive matching degree. For example, a value range of 0.8 to 1.0 corresponds to no warning level. At this level, the patient's condition is stable, and it is recommended to monitor indicators according to the regular cycle, such as once a month. A value range of 0.5 to 0.8 corresponds to a mild warning level. At this level, the patient has slight abnormalities in indicators. It is recommended to increase indicator monitoring once a week and adjust lifestyle factors such as diet and exercise as needed. A value range of 0.2 to 0.5 corresponds to a moderate warning level. At this level, the patient's indicators are significantly abnormal. It is recommended to have a follow-up visit to a specialist clinic within 3 days for the doctor to assess whether the medication plan needs to be adjusted. A value range of 0 to 0.2 corresponds to a severe warning level. At this level, the patient's condition is at risk of deterioration, and it is recommended to go to the hospital immediately to initiate the emergency intervention process. The calculated weighted overall matching degree, such as 0.748, is then precisely compared with these warning threshold ranges to determine that 0.748 belongs to the mild warning level range of 0.5 to 0.8. Combined with the clinical recommendations corresponding to this warning level, information including the warning level name, key abnormal indicators, and specific treatment suggestions is compiled to finally obtain the warning classification information of the combined indicator set.

[0130] The beneficial effects are as follows: By arraying the combined indicator set in multiple dimensions, the scattered indicator combinations are organized into a combined indicator matrix according to clinical assessment dimensions and patient correspondences, making the disease-specific data structure more regular and the correlation clearer, providing an ordered data foundation for subsequent feature extraction; when performing feature decomposition on the combined indicator matrix, data bias is eliminated through centering, covariance is calculated to construct an association matrix, and multidimensional feature vectors reflecting the core features of the data are accurately extracted, reducing the interference of redundant information on the analysis results; when comparing the difference between feature components and reference standards, the single-dimensional matching degree is mapped according to the range of numerical differences, quantifying the degree of fit of each feature and improving the matching results. The accuracy of the results is ensured by allocating clinical decision weights according to the disease stage, so that the weight settings are aligned with the treatment priorities at different stages and avoid secondary dimensions interfering with the core assessment. When calculating the weighted comprehensive matching degree, the weighted average and dispersion correction are combined to highlight the influence of key dimensions while taking into account the stability of the results, ensuring that the comprehensive matching degree can truly reflect the overall fit. The comprehensive matching degree is mapped to the warning threshold range and supplemented with clinical suggestions, making the warning grading information more clinically instructive. The entire process is progressively refined and improved, significantly enhancing the accuracy, reliability and practicality of the warning grading information, and providing a scientific and reliable warning basis for the subsequent generation of personalized intervention plans.

[0131] S5. Based on the early warning classification information, perform strategy fusion on the combined indicator set to generate a personalized intervention plan for the patient with the specific disease.

[0132] In this embodiment of the invention, the step of performing strategy fusion on the combined indicator set based on the early warning classification information to generate a personalized intervention plan for the specific disease patient includes:

[0133] The target elements of the combined indicator set are separated to obtain the core intervention targets and auxiliary regulation indicators of the combined indicator set.

[0134] Based on the aforementioned early warning classification information, select the strategy elements corresponding to the core intervention target from the preset disease-specific intervention plan;

[0135] Based on the clinical importance of the core intervention targets and the regulation coefficients of the auxiliary regulatory indicators, the intervention logic of the strategy elements is reconstructed to obtain the basic intervention plan for the patients with the specific disease.

[0136] The basic intervention plan is clinically validated, and based on the validation results, a personalized intervention plan is generated for the patients with the specific disease.

[0137] The intervention logic is reconstructed based on the clinical importance of the core intervention target and the adjustment coefficient of the auxiliary regulatory index to obtain the basic intervention plan for the patient with the specific disease, including:

[0138] Based on the clinical importance of the core intervention targets and the adjustment coefficients of the auxiliary adjustment indicators, parameter adjustment data for the strategy elements are generated;

[0139] Path simulation is performed on the parameter adjustment data to obtain the path deviation result of the parameter adjustment data;

[0140] Based on the path deviation results, the parameter adjustment data is dynamically corrected to obtain the optimized parameter configuration of the strategy elements;

[0141] The optimized parameters are configured and combined to form the basic intervention plan for the patients with the specific disease.

[0142] First, we analyzed the correlation between all indicator combinations in the set of combined indicators and the pathological mechanisms, disease progression, and prognosis of specific diseases. Then, combining authoritative clinical guidelines and treatment consensus in the specific disease field, we determined the nature of the impact of each indicator combination on the diagnosis and treatment of the specific disease. Indicator combinations that directly affect the core pathological aspects of the specific disease and determine the direction of disease progression were identified as core intervention targets, such as the combination of low-density lipoprotein cholesterol levels and the degree of coronary artery stenosis for coronary heart disease, and the combination of serum creatinine levels and urinary protein quantification for chronic kidney disease. Indicator combinations that indirectly affect the stability of the disease and assist the core intervention targets in exerting their effects were identified as auxiliary regulatory indicators, such as the combination of high-density lipoprotein cholesterol levels for coronary heart disease, and the combination of hemoglobin levels for chronic kidney disease. Through this judgment based on pathological correlation and clinical significance, we completed the separation of target elements, obtaining the core intervention targets and auxiliary regulatory indicators of the combined indicator set.

[0143] From a predefined disease-specific intervention program library, the corresponding intervention program framework is retrieved based on the current warning level information. Different warning levels correspond to different intensities and types of program frameworks: mild warning corresponds to a framework that focuses on lifestyle intervention and is supplemented by drug intervention, such as a low-salt diet plus regular exercise plus regular blood pressure monitoring for mild hypertension; moderate warning corresponds to a framework that combines lifestyle intervention and basic drug intervention, such as a low-sugar diet plus daily exercise plus oral hypoglycemic agents plus weekly blood glucose monitoring for moderate diabetes; severe warning corresponds to a framework that strengthens drug intervention, multidisciplinary collaborative intervention, and high-frequency monitoring, such as an inhaled corticosteroid combined with bronchodilators plus allergen avoidance plus daily pulmonary function monitoring plus respiratory follow-up for severe asthma. Within the retrieved framework, strategic elements directly corresponding to the core intervention target are selected. For example, if the core intervention target is a combination of low-density lipoprotein cholesterol level and coronary artery stenosis degree and the warning level is moderate, strategic elements directly related to the core target, such as the frequency of regular follow-up examinations for coronary artery stenosis when using statins, are selected to ensure that the selected strategic elements can accurately act on the core intervention target, thus obtaining strategic elements corresponding to the core intervention target.

[0144] First, based on the priority classification of core intervention targets in the clinical guidelines for specific diseases, the clinical importance score of each core intervention target was determined, with the score range set from 0 to 1. Among them, the score of core targets that directly determine the patient's life safety or whether the condition worsens should not be lower than 0.8, the score of targets that play a key role in stabilizing the condition but do not directly endanger life should be between 0.5 and 0.7, and the score of targets that have some impact on the condition but are not critical should be lower than 0.5. At the same time, the adjustment coefficient of auxiliary adjustment indicators was retrieved from the predefined disease-specific intervention protocol library. This coefficient is based on a large amount of clinical data statistics and reflects the auxiliary adjustment indicators' ability to improve the core intervention targets. The coefficient range is 0 to 0.5, with a coefficient of not lower than 0.3 for strong auxiliary ability, between 0.15 and 0.29 for moderate auxiliary ability, and lower than 0.15 for weak auxiliary ability. Following the logic that the parameter adjustment range equals the clinical importance score of the core intervention target multiplied by 1 plus the adjustment coefficient of the auxiliary adjustment index, the parameter adjustment data for each strategy element is calculated. For example, if the clinical importance score of the core intervention target low-density lipoprotein cholesterol value is 0.8, the corresponding adjustment coefficient of the auxiliary adjustment index high-density lipoprotein cholesterol value is 0.3. If the original baseline dose of the strategy element statin dosage is 10 mg per day, the parameter adjustment range is 0.8 multiplied by 1.3 equals 1.04, and the original dose is adjusted to 10.4 mg per day, which is rounded to 10 mg or 10.5 mg per day according to clinical drug dosage standards. If the original baseline duration of the strategy element weekly exercise time is 150 minutes, the parameter adjustment range is 0.8 multiplied by 1.3 equals 1.04, and it is adjusted to 156 minutes per week, thus obtaining the parameter adjustment data for the strategy element.

[0145] First, an ideal path model for the progression of a specific disease is constructed. This model is based on a large amount of clinical data of patients with the specific disease, consistent with the current patient warning level, disease stage, and initial state of the core intervention target. It clarifies the ideal trend and range of change of the core intervention target within the next 1 to 3 months after the parameter adjustment data is implemented. For example, for diabetic patients, the ideal path model is set at a weekly decrease of 0.5 mmol / L in fasting blood glucose, with weekly blood glucose fluctuations not exceeding 1.0 mmol / L; for hypertensive patients, the ideal path model is set at a monthly decrease of 5 mmHg in systolic blood pressure, with daily systolic blood pressure fluctuations not exceeding 10 mmHg. The calculated parameter adjustment data is then substituted into this ideal path model. A simulation logic is used to calculate daily or weekly indicator changes based on the patient's current indicator data according to the parameter adjustment requirements, and then summed to obtain the staged change results. This simulates the actual change path of the core intervention target within the next 1 to 3 months. The simulated actual change path is then compared stage by stage with the ideal trend in the ideal path model. If the improvement rate of the core intervention target in the simulated path is slower than the ideal path, for example, if the simulated fasting blood glucose of a diabetic patient decreases by only 0.2 mmol / L per week, which is lower than the ideal 0.5 mmol / L, or if the fluctuation range of the core intervention target exceeds the ideal range, for example, if the simulated daily systolic blood pressure fluctuation of a hypertensive patient reaches 15 mmHg, which exceeds the ideal 10 mmHg, then this situation is recorded as a path deviation result. The time node of the deviation is marked in detail, such as the deviation starting in the second week of the simulation, the deviation range, such as the deviation of blood glucose improvement rate of 0.3 mmol / L per week, and the corresponding parameter adjustment data item, such as the parameter item of daily exercise duration of 30 minutes, to obtain the path deviation result of parameter adjustment data.

[0146] For the deviation parameters marked in the path deviation results, first analyze the specific reasons for the deviation: If the deviation is due to insufficient parameter adjustment, such as insufficient daily exercise duration leading to slower blood sugar improvement than the ideal path, then increase the parameter value by 20% of the original adjustment range, for example, adjust the original daily exercise duration from 30 minutes to 36 minutes; if the deviation is due to excessive parameter adjustment, such as excessive medication dosage causing blood pressure fluctuations exceeding the ideal range, then decrease the parameter value by 15% of the original adjustment range, for example, adjust the original daily antihypertensive medication dosage from 10 mg to 8.5 mg; if the deviation is due to insufficient synergy between parameters, such as mismatch between dietary carbohydrate intake and exercise duration causing blood sugar fluctuations, then adjust two or more related parameters simultaneously, for example, adjust the original daily carbohydrate intake from 250 grams to 220 grams, and simultaneously adjust the original daily exercise duration from 30 minutes to 35 minutes. After parameter adjustment, the new parameter data is substituted into the ideal path model for disease progression again for simulation. The difference between the new simulated path and the ideal path is compared. The above analysis and adjustment simulation process is repeated until the deviation between the simulated path and the ideal path is less than the preset threshold. For example, the deviation of the blood glucose improvement rate for diabetic patients is less than 0.1 mmol / L per week, and the deviation of the blood pressure fluctuation amplitude for hypertensive patients is less than 2 mmHg per day. The parameter data at this time is the optimized parameter configuration of the strategy elements.

[0147] The optimization parameters of all strategy elements are integrated according to a fixed logical structure of core intervention measures, auxiliary adjustment measures, and monitoring plans. The core intervention measures section clarifies the specific operational content corresponding to each core intervention target, including drug intervention (e.g., drug name, daily dosage, frequency, and timing of administration), dietary intervention (e.g., daily intake of key nutrients and dietary restrictions), and surgical or physical intervention (e.g., surgical type and frequency). The auxiliary adjustment measures section clarifies the intervention methods corresponding to each auxiliary adjustment indicator, including exercise intervention (e.g., exercise type, daily exercise duration, and frequency), psychological intervention (e.g., frequency and methods of psychological counseling), and lifestyle adjustments (e.g., sleep schedule, smoking cessation, and alcohol abstinence requirements). The monitoring plan section clarifies the monitoring time for core intervention targets and auxiliary adjustment indicators (e.g., daily fasting blood glucose monitoring before breakfast, monthly blood lipid monitoring in the first week), monitoring frequency (e.g., blood glucose once daily, blood lipids once monthly), and data recording requirements (e.g., recording accompanying symptoms when monitoring values ​​are abnormal). This content is organized into a clearly structured and explicitly stated plan text, which serves as the basic intervention plan for patients with specific diseases.

[0148] Three or more experts with associate chief physician or higher titles and over five years of clinical experience in the specific disease area were invited to conduct clinical validation and evaluation of the basic intervention plan. The experts evaluated the plan from three core dimensions: first, the alignment of the plan with the patient's current disease stage and warning level, determining whether the intervention intensity matched the patient's disease risk; second, the safety of the parameters configured in the plan, verifying whether the drug dosage was within the clinically safe range and whether the intervention measures posed any potential health risks; and third, the feasibility of the plan, considering factors such as the patient's age, physical condition, and living environment to determine whether the patient could consistently implement the intervention requirements. Simultaneously, historical cases with similar core target statuses at the current patient's warning level were retrieved from the disease-specific case database to analyze the implementation effects of similar plans in similar cases. If the expert evaluation indicated problems such as excessive drug dosage or exercise intensity exceeding the patient's physical capacity, or if the case analysis showed poor blood sugar control with similar plans, the plan was adjusted based on expert advice and case experience, such as adjusting high-dose medication to a safe dose and reducing exercise intensity. After multiple rounds of evaluation and adjustment, a personalized intervention plan was developed that fully suited the individual patient's condition and ability to implement the plan.

[0149] The beneficial effects include: accurately distinguishing core intervention targets and auxiliary regulatory indicators through target element separation, providing a clear direction for subsequent strategy selection and ensuring that the intervention focus does not deviate; selecting corresponding strategy elements based on early warning levels, enabling precise matching of intervention intensity and disease risk to avoid over-intervention or under-intervention; generating parameter adjustment data through clinical importance and regulation coefficients, and obtaining optimized parameters through path simulation and dynamic correction, thereby improving the scientificity and rationality of the plan; effectively avoiding safety risks and implementation obstacles by combining expert evaluation and case reference in the clinical validation stage; and finally generating personalized intervention plans that can fully adapt to the individual patient's disease characteristics and execution ability, significantly improving the pertinence and effectiveness of disease-specific interventions, and helping to stabilize the patient's condition and improve prognosis.

[0150] S6. When the patient with the specific disease is receiving treatment according to the personalized intervention plan, real-time disease data of the patient with the specific disease is collected synchronously and continuously, and the real-time disease data is fed back to the disease screening rule base and the early warning rule base.

[0151] First, based on the core intervention targets and auxiliary regulatory indicators specified in the personalized intervention plan, the scope of real-time disease-specific data collection is determined. This includes key indicators corresponding to the core targets, such as fasting blood glucose and 2-hour postprandial blood glucose levels for diabetic patients, and systolic and diastolic blood pressure for hypertensive patients, as well as related data corresponding to auxiliary indicators, such as body mass index for diabetic patients and blood lipid levels for hypertensive patients. Simultaneously, the data collection method is clarified. For indicators that can be monitored at home, patients are instructed to use home medical devices with real-time data transmission capabilities, such as smart blood glucose meters and electronic blood pressure monitors, to collect data at preset frequencies, such as blood glucose three times daily and blood pressure twice daily. For indicators requiring monitoring by medical institutions, such as imaging data and molecular sequencing data, test results are obtained in real-time through the interface of the medical institution's diagnostic and treatment equipment. For diagnostic and treatment procedures and medication records, medical staff enter the data into the system within one hour after implementing the intervention measures, ensuring that data collection is synchronized and continuous with the treatment process, without data delays or omissions.

[0152] The collected real-time data is transmitted to the disease-specific data processing system in real time via an encrypted transmission protocol such as HTTPS. The system first performs a data integrity check, verifying that each data entry contains necessary fields such as the patient's unique identifier, collection time, indicator name, and indicator value. If any fields are missing, the system immediately sends feedback to the data acquisition end for supplementation. Next, the system performs an accuracy check, comparing the data value with the clinically reasonable range for that indicator. For example, the reasonable range for blood glucose is 2.8 to 22.2 mmol / L. If the value exceeds the range, it is marked as abnormal data, and medical staff verify whether to retain or correct it. Finally, according to the previously standardized data format requirements, the data units are unified, such as unifying the unit for blood pressure to millimeters of mercury and the unit for blood glucose to millimoles per liter, resulting in standardized real-time disease-specific data that can be directly used for rule base updates. Standardized real-time disease-specific data is imported into the update module of the disease-specific screening rule base. The module first calculates the proportion of patients in the newly added real-time data who meet the original screening criteria set. If the proportion is below 80% for a continuous month (e.g., the original diabetes screening criteria set a fasting blood glucose threshold of 7.0 mmol / L, while the newly added data shows that most confirmed patients have a fasting blood glucose threshold of 6.8 mmol / L), then the rule optimization process is initiated. Based on the indicator characteristics of confirmed patients in the newly added data, the test indicator conditions in the screening criteria set are adjusted, such as adjusting the fasting blood glucose threshold to 6.8 mmol / L. At the same time, the newly added disease-specific patient diagnosis and treatment cases are added to the case library of the rule base to enrich the clinical data support of the rule base, so that subsequent disease-specific screening based on the rule base can better fit the current clinical reality and improve the accuracy of screening.

[0153] The standardized real-time disease-specific data is input into the dynamic optimization module of the early warning rule base. The module first analyzes the actual distribution of each feature component in the real-time data. If it finds that the clinically reasonable range of a certain feature component deviates from the original multi-dimensional reference standard by more than 10% (e.g., the ideal value in the original reference standard for renal function feature component was 90 ml / min / 1.73 m² body surface area, but the new data shows that the value for most healthy patients is 85 ml / min / 1.73 m² body surface area), then the reference standard for that feature component is adjusted, for example, the ideal value is adjusted to 85. Then, based on the prognosis of patients at different disease stages in the real-time data, the clinical decision weight of the feature component is adjusted. For example, if the new data shows that a certain feature component has a greater impact on prognosis in the acute phase, then its weight is increased. At the same time, based on the fit between the early warning level in the real-time data and the actual disease development, the matching degree adjustment coefficient is adjusted. For example, if the original coefficient was 0.8, and the new data shows that the early warning level is too conservative, then the coefficient is adjusted to 0.7, ensuring that the early warning rule base can be continuously optimized with the update of clinical data, thereby improving the accuracy of subsequent early warning level information.

[0154] The beneficial effects include the synchronous and continuous collection of real-time disease-specific data from patients, ensuring complete synchronization between the acquired data and the implementation of personalized intervention plans. This avoids diagnostic biases caused by data lag and provides the latest data sources that align with actual clinical scenarios for subsequent rule base updates. Standardized real-time disease-specific data is fed back to the disease screening rule base, allowing for optimization of existing screening condition sets and enrichment of the case database based on new clinical data. This makes subsequent disease-specific patient screening more consistent with current clinical characteristics, reducing missed or incorrect screenings. Feedback to the early warning rule base dynamically adjusts the reference standards, clinical decision weights, and matching degree adjustment coefficients of feature components, making early warning grading information more accurately matched to the patient's real-time condition, avoiding over- or under-warning. Simultaneously, this process forms a closed loop of "intervention execution - data collection - rule optimization," ensuring the entire intelligent disease-specific data association and integration process continuously aligns with clinical practice, constantly improving the targeting and effectiveness of disease-specific diagnosis and treatment, and providing reliable support for more patients with specific diseases requiring precise treatment.

[0155] like Figure 2 The diagram shown is a functional block diagram of an intelligent disease-specific data association and integration system provided in an embodiment of the present invention.

[0156] The intelligent disease-specific data association and integration system 100 described in this invention can be installed in an electronic device. Depending on the functions implemented, the intelligent disease-specific data association and integration system 100 may include a data standardization module 101, a real-time screening module 102, an indicator analysis module 103, an early warning matching module 104, a strategy fusion module 105, and a feedback optimization module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0157] In this embodiment, the functions of each module / unit are as follows:

[0158] The data standardization module 101 is used to clean the patient's original medical data to obtain standardized data of the original medical data.

[0159] The real-time screening module 102 is used to perform real-time screening and judgment on the standardized data based on a preset disease-specific screening rule base to obtain a list of patients with specific diseases.

[0160] The indicator analysis module 103 is used to extract the disease data of the disease patients in the disease patient list from the standardized data, and to perform indicator correlation analysis on the disease data to obtain a set of combined indicators for the disease data.

[0161] The early warning matching module 104 is used to perform multi-dimensional dynamic matching of the combined indicator set based on a predefined early warning rule base to obtain the early warning classification information of the combined indicator set.

[0162] The strategy fusion module 105 is used to perform strategy fusion on the combined indicator set based on the early warning classification information to generate a personalized intervention plan for the patient with the specific disease.

[0163] The feedback optimization module 106 is used to simultaneously and continuously collect real-time disease data of the disease patient when the disease patient is receiving treatment according to the personalized intervention plan, and to feed the real-time disease data back to the disease screening rule base and the early warning rule base.

[0164] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0165] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0166] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0167] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0168] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent method for linking and integrating disease-specific data, characterized in that, The method includes: S1. Perform data cleaning on the patient's original medical data to obtain standardized data of the original medical data; S2. Based on a preset disease screening rule base, the standardized data is screened and judged in real time to obtain a list of patients with specific diseases. S3. Extract disease-specific data of patients in the disease-specific patient list from the standardized data, and perform indicator correlation analysis on the disease-specific data to obtain a set of combined indicators for the disease-specific data, including: Based on the medical correlations between clinical indicators in the disease-specific data, an indicator correlation network for the clinical indicators is constructed. Identify indicator nodes in the indicator association network that are related to the current disease stage of the patient with the specific disease; Starting from the indicator node, traverse the associated edges in the indicator association network to obtain the indicator combination with strong association in the indicator association network. The aforementioned combination of indicators is compiled into a combined indicator set for the specific disease data; S4. Based on a predefined early warning rule base, perform multi-dimensional dynamic matching on the combined indicator set to obtain the early warning classification information of the combined indicator set, including: The combined index set is arrayed in multiple dimensions to obtain the combined index matrix of the combined index set; Eigenvalue decomposition is performed on the combined index matrix to obtain the multidimensional feature vector of the combined index set; The difference between the feature components in the multidimensional feature vector and the multidimensional reference standard of the early warning rule is compared to obtain the single-dimensional matching degree of the feature component. Based on the current stage of the disease in the patients with the specific disease, assign clinical decision weights to the characteristic components; Based on the clinical decision weights, calculate the weighted comprehensive matching degree of the single-dimensional matching degree; The weighted comprehensive matching degree is mapped to the early warning threshold range to confirm the early warning classification information of the combined indicator set; S5. Based on the early warning classification information, perform strategy fusion on the combined indicator set to generate a personalized intervention plan for the patient with the specific disease. S6. When the patient with the specific disease is receiving treatment according to the personalized intervention plan, real-time disease data of the patient with the specific disease is collected synchronously and continuously, and the real-time disease data is fed back to the disease screening rule base and the early warning rule base.

2. The intelligent disease-specific data association and integration method as described in claim 1, characterized in that, The process of cleaning the patient's original medical data to obtain standardized data includes: Based on the data source, the patient's original diagnosis and treatment data are divided into clinical text data, medical imaging data, and molecular sequencing data; The clinical text data is corrected using medical terminology to obtain standardized text data of the original diagnosis and treatment data. The medical image data is structured to obtain unified image data of the original diagnostic and treatment data; Low-quality data is filtered from the molecular sequencing data to obtain high-quality molecular data from the original diagnostic and treatment data. The standardized text data, the unified image data, and the high-quality molecular data are collected in chronological order to obtain the standardized data of the original diagnostic and treatment data.

3. The intelligent disease-specific data association and integration method as described in claim 1, characterized in that, The standardized data is screened and judged in real time based on a preset disease-specific screening rule base to obtain a list of patients with specific diseases, including: The diagnostic conditions, test indicator conditions, examination result conditions, and medication conditions in the preset disease screening rule base are integrated into the screening condition set for the patient; Based on the set of screening criteria, patient records in the standardized data that meet the set of screening criteria are selected; Clinical phenotype clustering is performed on the patient records to obtain the disease classification results of the patient records; The patient records are arranged hierarchically according to the disease classification results to obtain a list of patients with specific diseases.

4. The intelligent disease-specific data association and integration method as described in claim 1, characterized in that, The step of constructing an indicator association network among clinical indicators based on the medical correlations among clinical indicators in the disease-specific data includes: Medical semantic analysis was performed on the clinical indicators to obtain their interaction relationships. Based on the aforementioned interaction relationships, the nature and strength of the association between the clinical indicators are determined; Using the clinical indicators as nodes, the correlation properties as connecting edges, and the correlation strength as the weight of the connecting edges, an initial network structure is constructed among the clinical indicators. By removing spurious associations from the initial network structure, the association network between the clinical indicators is obtained.

5. The intelligent disease-specific data association and integration method as described in claim 1, characterized in that, The predefined early warning rule base is used to perform multi-dimensional dynamic matching on the combined indicator set to obtain the early warning classification information of the combined indicator set, including: The combined index set is arrayed in multiple dimensions to obtain the combined index matrix of the combined index set; Eigenvalue decomposition is performed on the combined index matrix to obtain the multidimensional feature vector of the combined index set; The difference between the feature components in the multidimensional feature vector and the multidimensional reference standard of the early warning rule is compared to obtain the single-dimensional matching degree of the feature component. Based on the current stage of the disease in the patients with the specific disease, assign clinical decision weights to the characteristic components; Based on the clinical decision weights, calculate the weighted comprehensive matching degree of the single-dimensional matching degree; The weighted comprehensive matching degree is mapped to the early warning threshold range to confirm the early warning classification information of the combined indicator set.

6. The intelligent disease-specific data association and integration method as described in claim 1, characterized in that, The formula for calculating the weighted overall matching degree is as follows: ； In the formula, The overall matching degree, For the first Weighting of clinical decision-making in each dimension. For the first Single-dimensional matching degree of each dimension The total number of dimensions, This is the weighted average of the matching degree of the single dimension. This is the preset matching degree adjustment coefficient. For summation operations, This is the square root operation.

7. The intelligent disease-specific data association and integration method as described in claim 1, characterized in that, The step of fusing strategies based on the early warning classification information and the combined indicator set to generate a personalized intervention plan for the specific disease patient includes: The target elements of the combined indicator set are separated to obtain the core intervention targets and auxiliary regulation indicators of the combined indicator set. Based on the aforementioned early warning classification information, select the strategy elements corresponding to the core intervention target from the preset disease-specific intervention plan; Based on the clinical importance of the core intervention targets and the regulation coefficients of the auxiliary regulatory indicators, the intervention logic of the strategy elements is reconstructed to obtain the basic intervention plan for the patients with the specific disease. The basic intervention plan is clinically validated, and based on the validation results, a personalized intervention plan is generated for the patients with the specific disease.

8. The intelligent disease-specific data association and integration method as described in claim 7, characterized in that, The intervention logic is reconstructed based on the clinical importance of the core intervention target and the adjustment coefficient of the auxiliary regulatory index to obtain the basic intervention plan for the patient with the specific disease, including: Based on the clinical importance of the core intervention targets and the adjustment coefficients of the auxiliary adjustment indicators, parameter adjustment data for the strategy elements are generated; Path simulation is performed on the parameter adjustment data to obtain the path deviation result of the parameter adjustment data; Based on the path deviation results, the parameter adjustment data is dynamically corrected to obtain the optimized parameter configuration of the strategy elements; The optimized parameters are configured and combined to form the basic intervention plan for the patients with the specific disease.

9. An intelligent system for linking and integrating disease-specific data, characterized in that, The system for implementing the intelligent disease-specific data association and integration method according to claim 1 includes: The data standardization module is used to clean the patient's original medical data to obtain standardized data from the original medical data. The real-time screening module is used to perform real-time screening and judgment on the standardized data based on a preset disease-specific screening rule base to obtain a list of patients with specific diseases. The indicator analysis module is used to extract disease-specific data of patients in the disease-specific patient list from the standardized data, and to perform indicator correlation analysis on the disease-specific data to obtain a set of combined indicators for the disease-specific data. The early warning matching module is used to perform multi-dimensional dynamic matching of the combined indicator set based on a predefined early warning rule base to obtain the early warning classification information of the combined indicator set. The strategy fusion module is used to perform strategy fusion on the combined indicator set based on the early warning classification information to generate a personalized intervention plan for the patient with the specific disease. The feedback optimization module is used to synchronously and continuously collect real-time disease data of the patients with the specific disease when they receive diagnosis and treatment according to the personalized intervention plan, and to feed the real-time disease data back to the disease screening rule base and the early warning rule base.

Images