Smart hospital medical information management method and system

Abstract

The invention discloses an intelligent hospital medical information management method and system, and relates to the technical field of medical information, and the method comprises the steps: analyzing a medical direction and internal association in medical historical data to construct a partitioned medical knowledge graph, breaking a data island, analyzing the internal association of medical data to establish the partitioned medical knowledge graph, and carrying out the analysis of the internal association. A reliable basis is provided for subsequent determination of the disease change direction, partitioning is performed by considering the internal association of the medical condition, and the timeliness and effectiveness of medical knowledge graph search are ensured. And the multi-modal medical data of the patient is fused, so that the integrity of medical information integration is ensured, and the illness state is reflected more comprehensively. According to the method, the change direction of the illness state of the patient is determined, the change content is considered to speculate the change direction of the illness state, personalized medical scheme suggestions corresponding to each patient are generated according to the change direction of the illness state of the patient, the accuracy and adaptability of medical information early warning management are improved, and it is ensured that the medical information of the smart hospital reflects the illness state look-ahead.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a smart hospital medical information management method and system. Background Art

[0002] With the explosive growth of medical data (such as electronic medical records, imaging, and laboratory reports), traditional information management systems are facing challenges such as data silos, inefficiency, and insufficient decision support. For example, different departments use independent systems, resulting in data being unavailable for sharing. Doctors must manually integrate information, which is time-consuming and prone to errors. Furthermore, patients' expectations for healthcare quality and experience are increasing, demanding more precise and efficient diagnosis and treatment processes. Smart hospitals, by incorporating technologies such as the Internet of Things (IoT), big data, artificial intelligence (AI), and cloud computing, are building integrated medical information management platforms that enable data interconnection, intelligent assisted diagnosis, and dynamic resource allocation. For example, they leverage natural language processing (NLP) technology to automatically analyze medical records and, in conjunction with knowledge graphs, provide treatment recommendations. IoT devices are also used to monitor patients' vital signs in real time, providing early warnings of potential risks.

[0003] In the existing technology, patient condition analysis is performed only based on single-modal medical data, resulting in poor accuracy and adaptability of early warning management of medical information, which cannot meet the dynamic needs of modern smart hospitals.

[0004] Therefore, how to improve the accuracy and adaptability of early warning management of medical information is a technical problem that needs to be solved. Summary of the Invention

[0005] The purpose of the present invention is to solve the problem of poor accuracy and adaptability of early warning management of medical information in the existing technology, and to propose a smart hospital medical information management method, which includes: Obtain the medical directions and medical history data involved in the hospital, analyze the inherent correlation between medical directions and medical history data to build a partitioned medical knowledge graph; Acquire multimodal medical data of patients in the hospital, fuse the multimodal medical data of patients, build electronic medical records of patients, and update the electronic medical records regularly; Extract medical features from electronic medical records, establish a medical feature change sequence, and perform corresponding searches on the partitioned medical knowledge graph based on the medical feature change sequence to determine the direction of the patient's condition change; Generate personalized medical plan recommendations for each patient based on the direction of changes in the patient's condition, and use personalized medical plan recommendations to assist doctors in their reference.

[0006] In some embodiments of the present application, the inherent associations between medical directions and medical history data are analyzed to construct a partitioned medical knowledge graph, including: Use NLP tools in the medical field to identify entities in medical history data, define semantic relationships between different entities, represent entities as nodes, and represent semantic relationships between different entities as edges to describe the medical knowledge graph; The medical knowledge graph is preliminarily partitioned according to the medical pathology direction in the medical field. For each preliminarily partitioned medical knowledge graph, the association frequency parameters between entities and the causal association parameters between the order of entities are calculated. The association frequency parameters between entities and the causal relationship parameters between the entities before and after are integrated respectively to obtain the association strength index and the causal strength index; The thresholds of the association strength index and the causal strength index are set by the data characteristics of the association frequency parameter and the causal relationship parameter and the corresponding medical characteristics, and the preliminary partition is secondary partitioned based on the association strength index, the causal strength index and the corresponding thresholds.

[0007] In some embodiments of the present application, the thresholds of the association strength index and the causal strength index are set by the data characteristics of the association frequency parameter and the causal relationship parameter and the corresponding medical characteristics, including: Data characteristics include data sample size, data anomalies, and data coefficient of variation. Medical characteristics include disease mortality and infection rate. Determine the first interval threshold of the association strength index and the causal strength index according to the data sample size, data anomaly and data variation coefficient of the association frequency parameter and the causal relationship parameter; Determine the second interval threshold value of the association strength index and the causal strength index according to the mortality rate and the infection rate of the disease corresponding to the association frequency parameter and the causal relationship parameter; The threshold values ​​of the association strength indicator and the causal strength indicator are determined based on the first interval threshold value and the second interval threshold value of the association strength indicator and the causal strength indicator.

[0008] In some embodiments of the present application, the preliminary partition is secondary partitioned based on the association strength index, the causal strength index and the thresholds corresponding to the two, including: Based on the association strength index, causal strength index and their corresponding thresholds, the deviation of the association strength index and the causal strength index is determined respectively, and the causal association strength index is defined. The preliminary partition is then secondary partitioned by the causal association strength index.

[0009] In some embodiments of the present application, the multimodal medical data of the patient is fused, including: Determine the patient's disease type and evaluate the confidence of each modality of medical data based on the disease type. Calculate the attention between different modal medical data through the cross-modal attention mechanism, adjust the attention based on the confidence of the modal medical data, and fuse the different modal medical data based on the attention to obtain fused modal medical data.

[0010] In some embodiments of the present application, constructing a patient's electronic medical record includes: Define the electronic medical record template and fill the patient's basic information, multimodal medical data and fused modality medical data into the corresponding positions on the electronic medical record template.

[0011] In some embodiments of the present application, a medical feature change sequence is established, and a corresponding search is performed on the partitioned medical knowledge graph according to the medical feature change sequence, including: Medical features include single-modality medical features and fusion-modality medical features; An initial medical feature change sequence is constructed according to the temporal changes of the single-modality medical features and the fused-modality medical features, and the two initial medical feature change sequences are temporally aligned. Missing segments in the initial medical feature change sequence are interpolated to obtain a medical feature change sequence. Analyze the changes in the medical feature change sequences of the single-modality medical features and the fusion-modality medical features, and obtain the change trend characteristics of the single-modality medical features and the fusion-modality medical features; Calculate the matching degree of nodes of single-modal medical features and fusion-modal medical features on the partitioned medical knowledge graph.

[0012] In some embodiments of the present application, determining the direction of change in the patient's condition includes: The direction of change of the patient's condition is determined based on the matching degree of nodes of single-modality medical features and fusion-modality medical features as well as the change trend characteristics.

[0013] Correspondingly, this application also provides a smart hospital medical information management system, including: The first module is used to obtain the medical directions and medical history data involved in the hospital, analyze the inherent correlation between medical directions and medical history data to build a partitioned medical knowledge graph; The second module is used to obtain multimodal medical data of patients in the hospital, fuse the multimodal medical data of patients, build the electronic medical records of patients, and regularly update the electronic medical records; The third module is used to extract medical features from electronic medical records, establish a medical feature change sequence, and perform corresponding searches on the partitioned medical knowledge graph based on the medical feature change sequence to determine the direction of the patient's condition change; The fourth module is used to generate personalized medical plan recommendations for each patient based on the direction of changes in the patient's condition, and to assist doctors with reference through personalized medical plan recommendations.

[0014] The present invention has the following beneficial effects: 1. Analyze the inherent correlation between medical directions and medical historical data to build a partitioned medical knowledge graph, break the data silos, and analyze the inherent correlation of medical data to build a partitioned medical knowledge graph, providing a reliable basis for subsequent determination of the direction of disease changes. Partitioning takes into account the inherent correlation of medical conditions, ensuring the timeliness and effectiveness of medical knowledge graph searches.

[0015] 2. Integrate the patient's multimodal medical data to ensure the integrity of medical information integration and more comprehensively reflect the condition. Perform corresponding searches on the partitioned medical knowledge graph based on the sequence of changes in medical characteristics to determine the direction of the patient's condition change. Consider the content of the changes to infer the direction of the condition change. Based on the direction of the patient's condition change, generate personalized medical plan recommendations for each patient. This improves the accuracy and adaptability of medical information early warning management and ensures that smart hospital medical information reflects the foresight of the condition. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of a smart hospital medical information management method proposed by the present invention; Figure 2 This is a structural diagram of a smart hospital medical information management system proposed by the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0018] Reference Figure 1 , a smart hospital medical information management method, comprising the following steps: Step S101: obtain the medical directions and medical history data involved in the hospital, analyze the inherent correlation between the medical directions and medical history data to construct a partitioned medical knowledge graph.

[0019] In this implementation, medical data (e.g., department classification, disease area) and historical data (e.g., medical records, treatment notes) are combined to construct a structured, interpretable medical knowledge graph, supporting clinical decision-making, disease prediction, and personalized treatment. To ensure timely and effective use of the medical knowledge graph, the graph is partitioned according to associations and causal relationships.

[0020] In some embodiments of the present application, the inherent associations between medical directions and medical history data are analyzed to construct a partitioned medical knowledge graph, including: Use NLP tools in the medical field to identify entities in medical history data, define semantic relationships between different entities, represent entities as nodes, and represent semantic relationships between different entities as edges to describe the medical knowledge graph; The medical knowledge graph is preliminarily partitioned according to the medical pathology direction in the medical field. For each preliminarily partitioned medical knowledge graph, the association frequency parameters between entities and the causal association parameters between the order of entities are calculated. The association frequency parameters between entities and the causal relationship parameters between the entities before and after are integrated respectively to obtain the association strength index and the causal strength index; The thresholds of the association strength index and the causal strength index are set by the data characteristics of the association frequency parameter and the causal relationship parameter and the corresponding medical characteristics, and the preliminary partition is secondary partitioned based on the association strength index, the causal strength index and the corresponding thresholds.

[0021] In this example, NLP tools (such as MetaMap and ScispaCy) are used to extract entities (such as diseases, symptoms, medications, and test items) from text. For example, from the medical record "Patient complains of dizziness, blood sugar 7.8 mmol / L," extract "dizziness" (a symptom), "blood sugar" (a test item), and "7.8 mmol / L" (a numerical value). Define the relationship types between entities (such as "disease → cause → complication" and "medication → treatment → disease"). For example: "Diabetes" → "cause" → "diabetic nephropathy," "Insulin" → "treatment" → "Diabetes." Entities are stored as nodes and relationships as edges in a graph database (such as Neo4j).

[0022] The knowledge graph is initially partitioned according to medical pathology (e.g., endocrine, cardiovascular, and respiratory systems). Association frequency parameters describe the statistical strength of the co-occurrence of two entities in medical data, such as the number of co-occurrences or the probability of co-occurrence. Causal relationship parameters refer to the logical relationship whereby one entity (cause) leads to another entity (result), such as causal probability, conditional probability, and causal strength index. By integrating the association frequency parameters between entities and the causal relationship parameters between the order of entities, we obtain the Association Strength Index (ASI) and the Causal Strength Index (CSI), respectively. The data and medical characteristics of these parameters are considered to jointly set their respective thresholds, thereby performing secondary partitioning and separately classifying entity combinations with strong associations.

[0023] In some embodiments of the present application, the thresholds of the association strength index and the causal strength index are set by the data characteristics of the association frequency parameter and the causal relationship parameter and the corresponding medical characteristics, including: Data characteristics include data sample size, data anomalies, and data coefficient of variation. Medical characteristics include disease mortality and infection rate. Determine the first interval threshold of the association strength index and the causal strength index according to the data sample size, data anomaly and data variation coefficient of the association frequency parameter and the causal relationship parameter; Determine the second interval threshold value of the association strength index and the causal strength index according to the mortality rate and the infection rate of the disease corresponding to the association frequency parameter and the causal relationship parameter; The threshold values ​​of the association strength indicator and the causal strength indicator are determined based on the first interval threshold value and the second interval threshold value of the association strength indicator and the causal strength indicator.

[0024] In this embodiment, both data characteristics and medical characteristics may affect the threshold settings of the association strength index and the causal strength index, and therefore these factors are taken into consideration when setting the thresholds.

[0025] Data characteristics: Data sample size (N): The total number of cases in which entity pairs co-occur. The larger the sample size, the higher the threshold is set.

[0026] Outliers: Extreme values ​​(e.g., a pair of entities co-occurs significantly more frequently than others). The greater the degree of anomaly, the higher the threshold should be.

[0027] Data coefficient of variation (CV): standard deviation / mean, reflects the degree of data dispersion. The more stable the data, the higher the threshold is set.

[0028] The first interval thresholds of the association strength index and the causal strength index are determined by comprehensively considering the three aspects of data characteristics.

[0029] Medical Properties: Mortality Rate (MR): The proportion of deaths caused by the disease. The higher the mortality rate, the higher the threshold is set.

[0030] Transmission Rate (TR): The ability of a disease to spread among a population. The higher the transmission rate, the higher the threshold is set.

[0031] The second interval threshold of the association strength index and the causal strength index is comprehensively set based on the two comprehensive medical characteristics. The thresholds of the association strength index and the causal strength index are determined based on the first interval threshold and the second interval threshold of the association strength index and the causal strength index. The center value of the intersection interval of the two interval thresholds is used as the threshold. If there is no intersection between the two intervals, the center value of the distance between the two intervals is used as the threshold.

[0032] In some embodiments of the present application, the preliminary partition is secondary partitioned based on the association strength index, the causal strength index and the thresholds corresponding to the two, including: Based on the association strength index, causal strength index and their corresponding thresholds, the deviation of the association strength index and the causal strength index is determined respectively, and the causal association strength index is defined. The preliminary partition is then secondary partitioned by the causal association strength index.

[0033] In this embodiment, the deviation includes positive deviation and negative deviation. The positive deviation is the part of the indicator that exceeds the threshold, and the negative deviation is the part of the indicator that is lower than the threshold. The causal correlation strength index is defined as follows: ; in, For the An indicator of the strength of the causal relationship between groups of entities, 、 are the combined weights of the association strength index and the causal strength index, 、 For the The deviation of the association strength index and causal strength index between groups of entities, for 、 The maximum value in For the The first constant between groups of entities, It represents the correction of the sum of the two by the maximum value, and the first constant is used to balance the size of the correction function.

[0034] Step S102: Acquire the multimodal medical data of the patient in the hospital, fuse the multimodal medical data of the patient, construct the electronic medical record of the patient, and regularly update the electronic medical record.

[0035] In this embodiment, the multimodal medical data includes structured data and unstructured data; Structured data: laboratory test results (such as blood sugar, blood pressure), diagnosis codes (ICD-10), etc.

[0036] Unstructured data: doctor's notes (text), medical images (DICOM format), vital signs monitoring data (time series), etc.

[0037] Each modality of medical data reflects the patient's condition from one aspect and requires comprehensive consideration to help analyze the condition.

[0038] In some embodiments of the present application, the multimodal medical data of the patient is fused, including: Determine the patient's disease type and evaluate the confidence of each modality of medical data based on the disease type. Calculate the attention between different modal medical data through the cross-modal attention mechanism, adjust the attention based on the confidence of the modal medical data, and fuse the different modal medical data based on the attention to obtain fused modal medical data.

[0039] In this embodiment, the credibility of modal medical data varies depending on the disease type. For example, for pneumonia, imaging data is more convincing than text. The attention between medical data of different modalities is the correlation between the different modalities (a group). The confidence levels of the modalities in a group are integrated, and an adjustment coefficient is mapped based on the integrated confidence levels. The correlation (product) is adjusted using the adjustment coefficient to determine the weight of each modality, thereby achieving the fusion of medical data of different modalities.

[0040] In some embodiments of the present application, constructing a patient's electronic medical record includes: Define the electronic medical record template and fill the patient's basic information, multimodal medical data and fused modality medical data into the corresponding positions on the electronic medical record template.

[0041] In this embodiment, the FHIR (Fast Healthcare Interoperability Resources) standard or a customized XML / JSON template is used to ensure data scalability and interoperability.

[0042] Step S103: extract medical features from the electronic medical record, establish a medical feature change sequence, perform a corresponding search on the partitioned medical knowledge graph according to the medical feature change sequence, and determine the direction of the patient's condition change.

[0043] In some embodiments of the present application, a medical feature change sequence is established, and a corresponding search is performed on the partitioned medical knowledge graph according to the medical feature change sequence, including: Medical features include single-modality medical features and fusion-modality medical features; An initial medical feature change sequence is constructed according to the temporal changes of the single-modality medical features and the fused-modality medical features, and the two initial medical feature change sequences are temporally aligned. Missing segments in the initial medical feature change sequence are interpolated to obtain a medical feature change sequence. Analyze the changes in the medical feature change sequences of the single-modality medical features and the fusion-modality medical features, and obtain the change trend characteristics of the single-modality medical features and the fusion-modality medical features; Calculate the matching degree of nodes of single-modal medical features and fusion-modal medical features on the partitioned medical knowledge graph.

[0044] In this embodiment, the single-modality medical feature is a feature on a single modality, and the fused-modality medical feature is a feature that fuses multiple modalities.

[0045] For example, a unimodal medical feature: Structured data: blood sugar levels, blood pressure readings.

[0046] Unstructured data: Text description of "thickened lung markings" in a chest X-ray.

[0047] Fusion modality medical features: An “infection risk score” that combines imaging (likelihood of pneumonia) and laboratory data (white blood cell count).

[0048] The changing trend characteristics of single-modality medical features and fusion-modality medical features can be obtained in the following ways: Statistical characteristics: Mean and variance: Calculate the mean and variance of the feature within each time window to reflect the central tendency and dispersion of the data.

[0049] Slope and intercept: Through linear regression analysis, the slope and intercept of the feature in each time window are calculated to reflect the change rate and starting level of the feature.

[0050] Time Series Analysis: Moving Average: Calculates a moving average at each time point to smooth the data and observe its overall trend.

[0051] Exponential smoothing: Exponential smoothing is used to smooth the time series to capture long-term trends in the data.

[0052] Change point detection: CUSUM (Cumulative Sum Chart): Used to detect abrupt changes in a time series, identifying moments when a characteristic value suddenly increases or decreases.

[0053] Bayesian change point detection: Infers the location of change points in a time series using Bayesian methods, providing probabilistic change point detection results.

[0054] Trend direction analysis: Direction change statistics: Count the number of times the feature value rises, falls, or remains stable to analyze the overall change direction of the feature.

[0055] Trend Strength: Evaluate the strength of a trend by calculating the magnitude and frequency of changes in characteristic values.

[0056] The matching degree of the nodes of the partitioned medical knowledge graph for the single-modality medical features and the fusion-modality medical features can be calculated as follows: Cosine similarity: Calculate the cosine similarity between the medical feature vector and the node vector on the medical knowledge graph to measure the similarity between the two. The closer the cosine similarity is to 1, the more similar the two are.

[0057] Euclidean distance: Calculates the Euclidean distance between the medical feature vector and the node vector on the medical knowledge graph. The closer the distance, the more similar the two are. In practical applications, it may be necessary to normalize the distance to facilitate comparison.

[0058] Jaccard similarity coefficient: Applicable to discrete medical features, it measures similarity by calculating the ratio of the intersection to the union of two sets.

[0059] In some embodiments of the present application, determining the direction of change in the patient's condition includes: The direction of change of the patient's condition is determined based on the matching degree of nodes of single-modality medical features and fusion-modality medical features as well as the change trend characteristics.

[0060] In this embodiment, the direction of the evolution of the patient's condition may change to the relevant nodes on the knowledge graph, so corresponding early warning analysis is required.

[0061] A Markov chain is constructed based on the changing trend characteristics, with possible evolution directions (nodes) as different states. Based on historical data, the state transition probability of moving from one state to another is calculated. The overall matching degree is determined based on the matching degree of nodes based on single-modality medical features and fusion-modality medical features, as well as the state transition probability. The formula is as follows: ; in, For the The overall matching degree of each direction, 、 are the matching weights of single-modality medical features and fusion-modality medical features, 、 Respectively The matching degree between single-modality medical features and fusion-modality medical features in each direction, For the The state transition probability in each direction, For the Correction coefficients in each direction, Indicates that The correction coefficient obtained by mapping the state transition probability in each direction is used to correct the sum of the matching degrees of the two features.

[0062] Step S104: Generate personalized medical plan recommendations for each patient based on the direction of change in the patient's condition, and use the personalized medical plan recommendations to assist doctors in their reference.

[0063] In this embodiment, the directions with the highest overall matching degree are retained and analyzed to generate personalized medical plan recommendations, for example, Treatment strategies: drugs, surgery, radiotherapy, etc.

[0064] Dosage adjustment: Dynamic adjustment based on patient weight, liver and kidney function.

[0065] Monitoring plan: examination frequency, indicator thresholds (such as blood glucose monitoring frequency).

[0066] Emergency plan: emergency measures when the condition worsens.

[0067] Correspondingly, this application also provides a smart hospital medical information management system, such as Figure 2 Shown, including, The first module is used to obtain the medical directions and medical history data involved in the hospital, analyze the inherent correlation between medical directions and medical history data to build a partitioned medical knowledge graph; The second module is used to obtain multimodal medical data of patients in the hospital, fuse the multimodal medical data of patients, build the electronic medical records of patients, and regularly update the electronic medical records; The third module is used to extract medical features from electronic medical records, establish a medical feature change sequence, and perform corresponding searches on the partitioned medical knowledge graph based on the medical feature change sequence to determine the direction of the patient's condition change; The fourth module is used to generate personalized medical plan recommendations for each patient based on the direction of changes in the patient's condition, and to assist doctors with reference through personalized medical plan recommendations.

[0068] The present invention has the following beneficial effects: 1. Analyze the inherent correlation between medical directions and medical historical data to build a partitioned medical knowledge graph, break the data silos, and analyze the inherent correlation of medical data to build a partitioned medical knowledge graph, providing a reliable basis for subsequent determination of the direction of disease changes. Partitioning takes into account the inherent correlation of medical conditions, ensuring the timeliness and effectiveness of medical knowledge graph searches.

[0069] 2. Integrate the patient's multimodal medical data to ensure the integrity of medical information integration and more comprehensively reflect the condition. Perform corresponding searches on the partitioned medical knowledge graph based on the sequence of changes in medical characteristics to determine the direction of the patient's condition change. Consider the content of the changes to infer the direction of the condition change. Based on the direction of the patient's condition change, generate personalized medical plan recommendations for each patient. This improves the accuracy and adaptability of medical information early warning management and ensures that smart hospital medical information reflects the foresight of the condition.

[0070] Through the above description of the embodiments, those skilled in the art will clearly understand that the present invention can be implemented via hardware or via software combined with a necessary general-purpose hardware platform. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product. This software product can be stored on a non-volatile storage medium (such as a CD-ROM, USB flash drive, or external hard drive) and includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute the methods described in various implementation scenarios of the present invention.

[0071] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred implementation scenario, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0072] Those skilled in the art will appreciate that the modules in the system of the implementation scenario can be distributed in the system of the implementation scenario according to the implementation scenario description, or can be modified accordingly and located in one or more systems different from the implementation scenario. The modules of the above implementation scenario can be combined into one module or further divided into multiple submodules.

[0073] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A smart hospital medical information management method, characterized in that: include, Obtain the medical directions and medical history data involved in the hospital, analyze the inherent correlation between medical directions and medical history data to build a partitioned medical knowledge graph; Acquire multimodal medical data of patients in the hospital, fuse the multimodal medical data of patients, build electronic medical records of patients, and update the electronic medical records regularly; Extract medical features from electronic medical records, establish a medical feature change sequence, and perform corresponding searches on the partitioned medical knowledge graph based on the medical feature change sequence to determine the direction of the patient's condition change; Generate personalized medical plan recommendations for each patient based on the direction of changes in the patient's condition, and use personalized medical plan recommendations to assist doctors in their reference.

2. The smart hospital medical information management method according to claim 1, characterized in that: Analyze the inherent associations between medical directions and medical history data to build a partitioned medical knowledge graph, including: Use NLP tools in the medical field to identify entities in medical history data, define semantic relationships between different entities, represent entities as nodes, and represent semantic relationships between different entities as edges to describe the medical knowledge graph; The medical knowledge graph is preliminarily partitioned according to the medical pathology direction in the medical field. For each preliminarily partitioned medical knowledge graph, the association frequency parameters between entities and the causal association parameters between the order of entities are calculated. The association frequency parameters between entities and the causal relationship parameters between the entities before and after are integrated respectively to obtain the association strength index and the causal strength index; The thresholds of the association strength index and the causal strength index are set by the data characteristics of the association frequency parameter and the causal relationship parameter and the corresponding medical characteristics, and the preliminary partition is secondary partitioned based on the association strength index, the causal strength index and the corresponding thresholds.

3. The smart hospital medical information management method according to claim 2, characterized in that: The thresholds of the association strength index and the causal strength index are set by the data characteristics of the association frequency parameter and the causal relationship parameter and the corresponding medical characteristics, including: Data characteristics include data sample size, data anomalies, and data coefficient of variation. Medical characteristics include disease mortality and infection rate. Determine the first interval threshold of the association strength index and the causal strength index according to the data sample size, data anomaly and data variation coefficient of the association frequency parameter and the causal relationship parameter; Determine the second interval threshold value of the association strength index and the causal strength index according to the mortality rate and the infection rate of the disease corresponding to the association frequency parameter and the causal relationship parameter; The threshold values ​​of the association strength indicator and the causal strength indicator are determined based on the first interval threshold value and the second interval threshold value of the association strength indicator and the causal strength indicator.

4. The smart hospital medical information management method according to claim 2, characterized in that: The preliminary partition is then partitioned into two parts based on the association strength index, causal strength index and their corresponding thresholds, including: Based on the association strength index, causal strength index and their corresponding thresholds, the deviation of the association strength index and the causal strength index is determined respectively, and the causal association strength index is defined. The preliminary partition is then secondary partitioned by the causal association strength index.

5. The smart hospital medical information management method according to claim 1, characterized in that: Fusion of multimodal medical data of patients, including, Determine the patient's disease type and evaluate the confidence of each modality of medical data based on the disease type. Calculate the attention between different modal medical data through the cross-modal attention mechanism, adjust the attention based on the confidence of the modal medical data, and fuse the different modal medical data based on the attention to obtain fused modal medical data.

6. The smart hospital medical information management method according to claim 5, characterized in that: Build a patient's electronic medical record, including, Define the electronic medical record template and fill the patient's basic information, multimodal medical data and fused modality medical data into the corresponding positions on the electronic medical record template.

7. The smart hospital medical information management method according to claim 6, characterized in that: Establish a medical feature change sequence and perform corresponding searches on the partitioned medical knowledge graph according to the medical feature change sequence, including: Medical features include single-modality medical features and fusion-modality medical features; An initial medical feature change sequence is constructed according to the temporal changes of the single-modality medical features and the fused-modality medical features, and the two initial medical feature change sequences are temporally aligned. Missing segments in the initial medical feature change sequence are interpolated to obtain a medical feature change sequence. Analyze the changes in the medical feature change sequences of the single-modality medical features and the fusion-modality medical features, and obtain the change trend characteristics of the single-modality medical features and the fusion-modality medical features; Calculate the matching degree of nodes of single-modal medical features and fusion-modal medical features on the partitioned medical knowledge graph.

8. The smart hospital medical information management method according to claim 7, characterized in that: Determine the direction of the patient's condition, including: The direction of change of the patient's condition is determined based on the matching degree of nodes of single-modality medical features and fusion-modality medical features as well as the change trend characteristics.

9. A smart hospital medical information management system, characterized in that: include, The first module is used to obtain the medical directions and medical history data involved in the hospital, analyze the inherent correlation between medical directions and medical history data to build a partitioned medical knowledge graph; The second module is used to obtain multimodal medical data of patients in the hospital, fuse the multimodal medical data of patients, build the electronic medical records of patients, and regularly update the electronic medical records; The third module is used to extract medical features from electronic medical records, establish a medical feature change sequence, and perform corresponding searches on the partitioned medical knowledge graph based on the medical feature change sequence to determine the direction of the patient's condition change; The fourth module is used to generate personalized medical plan recommendations for each patient based on the direction of changes in the patient's condition, and to assist doctors with reference through personalized medical plan recommendations.

Images