An intelligent medical record data management system and method

Abstract

The application relates to the technical field of medical record data management, in particular to an intelligent medical record data management system and method; the method comprises the following steps: connecting with various information systems of a hospital, collecting patient medical record data in real time and preprocessing, constructing a dimension table after determining the dimension; taking the dimension as a process and the time-space sequence as a connection path to generate a medical record path template; constructing a sample set by sampling historical medical record data, analyzing a sample path diagram, and calculating a general difficulty value of the template; integrating process factor information, classifying samples, and calculating a feature management difficulty value; combining patient state information, the general difficulty value and the feature management difficulty value to construct a priority evaluation model and output an evaluation priority value; and adopting a hierarchical intelligent calling strategy to manage medical record data storage calling, optimizing calling performance and dynamically adjusting the storage level. The application optimizes the resource allocation of different medical record data through a hierarchical management strategy, and improves the management effect of different medical record data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical record data management, and particularly relates to an intelligent medical record data management system and method. BACKGROUND

[0002] With the development of medical treatment, medical record data also shows explosive growth, and reasonable management of medical record data becomes more and more important for medical informatization, and is crucial for medical quality and decision-making.

[0003] The traditional medical record management system has the problem of extensive storage and calling strategy, which leads to low data processing efficiency, unreasonable resource allocation, and difficulty in meeting the real-time, accuracy and intelligent analysis requirements of medical data, so that the structured processing, intelligent evaluation and hierarchical management of medical record data are implemented to realize efficient organization and utilization of medical record data, and to improve the quality of medical services and management efficiency, which has become an urgent problem to be solved. SUMMARY

[0004] The present application aims at the problems in the background art and provides an intelligent medical record data management system and method.

[0005] The technical scheme of the present application is an intelligent medical record data management method, comprising the following steps:

[0006] Real-time acquisition of various medical record data of patients, data preprocessing of various medical record data, determination of medical record data dimension, and construction of dimension table;

[0007] Taking the dimension in the medical record data as a process and the time and space sequence between the processes as a connection path, a plurality of medical record path templates are obtained by exhaustion;

[0008] Random sampling of medical record data, construction of a sample set, analysis of sample path graphs of each sample, calculation of general difficulty values of medical record path templates according to entity association conditions of the dimension table and the sample path graphs;

[0009] Obtaining information of each process factor in the sample path graph and integrating the information to obtain full-process information, processing the full-process information, classifying the sample path graph according to the processing result to obtain a sample class, and calculating a feature management difficulty value of the sample class;

[0010] Analyzing the log data of the patient to determine the patient state information, constructing a priority evaluation model combining the general difficulty value of the medical record path template and the feature management difficulty value of the sample class, and outputting an evaluation priority value of the patient medical record data;

[0011] The evaluation priority value of the medical record data adopts a hierarchical intelligent calling strategy, the medical record data is stored and called for management, the calling performance is optimized, and the storage level is dynamically adjusted.

[0012] Preferably, the method for analyzing the sample path graph of each sample comprises:

[0013] Randomly sampling the medical record data to construct a sample set, and obtaining the log data of each sample in the sample set;

[0014] Taking the data entities in the process as process factors, determining the process factors contained in different processes in the sample according to the log data, and determining the sample path graph according to the analysis of the medical record path template set.

[0015] Preferably, the method for calculating the general difficulty value of the medical record path template comprises:

[0016] Obtaining the association list of the process factor in the sample path graph, and counting the number of the association list as the association degree of the process factor;

[0017] Obtaining the association degree of the process factor in the sample path graph, the data transmission amount, the time consumption, the memory occupancy rate, the cache hit rate, the disk I / O throughput and the delay in the update, and taking them as the multi-dimensional representation parameters of the process factor, and constructing the corresponding representation parameter list based on the sample set;

[0018] Fitting the representation parameter list and the server load change list by using the least square method, determining the weight of each representation parameter, normalizing the element weight, weighting and summing the multi-dimensional representation parameters based on the processed weight, and obtaining the update difficulty value di of the process factor; i is the process factor number, and i is a positive integer.

[0019] Preferably, the method for calculating the general difficulty value of the medical record path template further comprises:

[0020] Determining the generation period and the corresponding time span of the sample according to the log data, counting the update times of the process factor in the generation period, and calculating the update frequency of each process factor in the sample based on the time span and the update times, calculating the average update frequency of the process factor based on the sample set and marking it as Ufi;

[0021] According to the entity definition of the dimension table, marking the process factor marked as the proxy key as the selected factor, calculating the query times of the selected factor in the corresponding medical log record of the sample set, calculating the query frequency based on the time span, calculating the average query frequency of the process factor based on the sample set and marking it as Cfi;

[0022] The operation difficulty value C of the sample path graph is calculated by the formula The operation difficulty value C of the sample path graph is calculated by the formula

[0023] In the sample set, according to the template attribute of the sample path diagram, the same type of medical record path diagram is obtained, the operation difficulty value C of the same type of medical record path diagram is calculated, and the general management difficulty value of the medical record data of the medical record path template is obtained.

[0024] Preferably, the state information includes an in-hospital state and a reservation state; the in-hospital state includes in-hospital and discharge; and the reservation state includes reservation and no reservation.

[0025] Preferably, the method for constructing the priority evaluation model in combination with the general difficulty value of the medical record path template and the characteristic management difficulty value of the sample type comprises:

[0026] The basic priority coefficient Bp of the patient is determined by using the priority coefficient formulation expression as follows:

[0027] In the formula, sp1, sp2 and sp3 are preset priority coefficients; sp1>sp2>sp3.

[0028] The basic priority value P of the medical record data of each patient ID is monitored, the current process is identified, the real-time medical record path template is determined, and the real-time general management difficulty value YC is obtained.

[0029] It is monitored whether the current process produces a keyword, and if a keyword is produced, a target keyword list is constructed.

[0030] The sample type real-time characteristic management difficulty value TC corresponding to the sample type is obtained based on the adapted sample type keyword list.

[0031] The evaluation priority value of the medical record data is calculated based on the formula Ep=Bp×(k1×YC+k2×TC); in the formula, k1 and k2 are weight coefficients.

[0032] Preferably, the method for analyzing and obtaining the sample type keyword list matched with the target keyword list comprises:

[0033] A word frequency vector set of the keyword set is created, and a union operation is performed on the target keyword list and the sample type keyword list to obtain a temporary keyword list.

[0034] The target keyword list and the temporary keyword list are respectively converted into target vector A and temporary vector B, and the cosine similarity of the target vector A and the temporary vector B is calculated by using the formula The cosine similarity of the target vector A and the temporary vector B is calculated and used as the matching degree of the target keyword list and the sample type keyword list.

[0035] The sample type keyword list corresponding to the maximum value of the matching degree between the target keyword list and the sample type keyword list is obtained, and the sample type keyword list matched with the target keyword list is obtained.

[0036] Preferably, the hierarchical intelligent calling strategy includes setting three priority levels according to the evaluation priority gradient of all case data, and four priority levels correspond to low priority data, medium priority data, high priority data and extremely high priority data respectively, and different implementation methods are implemented for different priority data.

[0037] The application further discloses an intelligent medical record data management system applying the intelligent medical record data management method.

[0038] The data acquisition and table construction module is used for collecting various medical record data of patients in real time, pre-processing the medical record data, determining the dimension of the medical record data, and constructing a dimension table.

[0039] The first data processing module is used for taking the dimensions in the medical record data as processes, taking the time and space sequence between the processes as a connection path, and exhaustively obtaining a plurality of medical record path templates.

[0040] The second data processing module is used for randomly sampling the medical record data, constructing a sample set, analyzing and obtaining a sample path graph of each sample, calculating a general difficulty value of the medical record path template according to the entity association of the dimension table and the sample path graph.

[0041] The third data processing module is used for obtaining information of each process factor in the sample path graph and integrating the information to obtain full-process information, processing the full-process information, classifying the sample path graph according to the processing result to obtain a sample class, and calculating a feature management difficulty value of the sample class.

[0042] The model construction module is used for analyzing the log data of the patient, determining the state information of the patient, combining the general difficulty value of the medical record path template and the feature management difficulty value of the sample class to construct a priority evaluation model, and outputting an evaluation priority value of the medical record data of the patient.

[0043] The data management module is used for adopting a hierarchical intelligent calling strategy for the evaluation priority value of the medical record data, managing storage and calling of the medical record data, optimizing calling performance, and dynamically adjusting the storage level.

[0044] Compared with the prior art, the above technical scheme of the application has the following beneficial technical effects:

[0045] By constructing a medical record path template and a medical record data sampling processing, the general management difficulty and the characteristic management difficulty are quantified, and a priority evaluation model is established by combining real-time state of patients and keyword analysis to dynamically calculate an evaluation priority value of the medical record data, a four-level management strategy is implemented based on the evaluation priority value, a targeted calling and storage method is adopted for different priority data, the problem of lag and jam in operation response of high-priority medical record data of inpatients is solved, resource allocation of different medical record data is optimized, and the management effect of different medical record data is improved. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 A method block diagram of the embodiment one proposed by the present application. DETAILED DESCRIPTION

[0047] Embodiment one, as shown in the figure, the present application proposes an intelligent medical record data management method, including the following steps: Figure 1

[0048] Real-time acquisition of various medical record data of patients, data preprocessing of each medical record data, determination of medical record data dimension, and construction of dimension table; wherein, various medical record data can be obtained by interfacing with each information system of the hospital;

[0049] Data preprocessing includes converting the collected data of different formats into a standard data format prepared in advance, which is convenient for subsequent processing and storage; by pre-set rules, the obviously wrong or repeated data is removed;

[0050] For example, check the patient age field, if there is negative number or data far beyond the human life range, it is considered as error data for marking and removing;

[0051] For example, the dimension table can include patient basic information table, initial diagnosis information table, examination information table, diagnosis information table, re-examination information table, treatment information table, etc.

[0052] The dimension in the medical record data is taken as a process, and the space-time sequence between the processes is taken as a connection path, and a plurality of medical record path templates are obtained by exhaustion; for example, the entities in the medical record data such as patients, visits, examinations, diagnoses and treatments are taken as processes, and the space-time sequence from the process patient to the visit is taken as a connection path;

[0053] For example, the medical record path template can be patient→initial diagnosis, patient→initial diagnosis→diagnosis, patient→initial diagnosis→examination→re-examination→diagnosis, examination→re-examination→second examination→second re-examination→diagnosis→treatment, diagnosis→treatment, etc.

[0054] Random sampling of medical record data, construction of sample set, and analysis of sample path diagram of each sample;

[0055] ​The sample path graph of each sample is analyzed, including obtaining log data of each sample in the sample set, determining the medical record path template of the application based on the sample log data;

[0056] The data entities in the process are taken as process factors, the process factors contained in the sample are determined according to the log data, the sample path graph is constructed by combining the medical record path template, and the operation difficulty value of the sample path graph is calculated according to the entity association of the dimension table;

[0057] The method for calculating the general difficulty value of the medical record path template comprises:

[0058] The association list of the process factor in the sample path graph is obtained, and the number of the association list is counted as the association degree of the process factor;

[0059] For example, the sample corresponds to a medical record path template: patient→initial diagnosis→diagnosis, wherein the several process factors of the process patient are patient ID, name, gender, male, age, and basic medical history;

[0060] The several process factors of the process initial diagnosis include registration method, registration subject, inquiry doctor, registration date, examination code, and whether to reconsult;

[0061] The several process factors of the process diagnosis include diagnosis code and treatment scheme code;

[0062] The patient ID in the process patient is linked to all the process factors in the processes of initial diagnosis and diagnosis; the inquiry doctor is linked to the diagnosis code and the treatment scheme code, which is convenient for subsequent responsibility tracing; the basic medical history is linked to the treatment scheme code to prevent conflicts between drugs and basic diseases in drug treatment;

[0063] The association degree of the process factor in the sample path graph, the data transmission amount, the time consumption, the memory occupation rate, the cache hit rate, the disk I / O throughput and the delay when updating are obtained and taken as the multi-dimensional representation parameters of the process factor, and the corresponding representation parameter list is constructed based on the sample set;

[0064] The corresponding server load change list of the sample set when the process factor is updated is obtained according to the log data;

[0065] The least square method is used to fit the representation parameter list and the server load change list, the weight of each representation parameter is determined, the element weight is normalized, the multi-dimensional representation parameters are weighted and summed based on the processed weight, and the update difficulty value di of the process factor is obtained; i is the process factor number, and i is a positive integer;

[0066] According to the log data, a generation period of the sample and a corresponding time span are determined, the number of updates of the process factor in the generation period is counted, and the update frequency of each process factor in the sample is calculated based on the time span and the number of updates; the update frequency average of the process factor is calculated according to the sample set and is marked as Ufi;

[0067] According to the entity definition of the dimension table, the process factor marked as the proxy key is marked as the selected factor, the number of queries of the selected factor in the corresponding medical record log record of the sample set is calculated, and the query frequency is calculated based on the time span; the query frequency average of the process factor is calculated according to the sample set and is marked as Cfi;

[0068] The number of queries includes the number of times that the patient queries the electronic medical record and the number of times that the hospital calls the electronic medical record;

[0069] The operation difficulty value C of the sample path graph is calculated by the formula The operation difficulty value C of the sample path graph is calculated by the formula

[0070] In the sample set, the sample path graphs are divided according to the template attributes of the sample path graphs to obtain the same type of medical record path graphs, and the operation difficulty value C of the same type of medical record path graphs is calculated to obtain the general management difficulty value of the medical record data applied to the medical record path template;

[0071] The information of each process factor in the sample path graph is obtained and integrated to obtain the whole process information, the whole process information is processed, the sample path graph is classified according to the processing result to obtain a sample class, and the characteristic management difficulty value of the sample class is calculated;

[0072] The method for classifying the sample path graph by using the natural language processing technology and the clustering algorithm on the whole process information to obtain the sample class includes:

[0073] The keywords are extracted by natural language processing on the whole process information to construct a keyword list;

[0074] The keyword lists of all sample medical record paths are summarized to obtain a keyword set;

[0075] The number of the same keywords in the keyword set is counted, the keyword types with a number less than a number threshold are removed, and the keyword list of the sample path graph is updated;

[0076] The sample path graph is classified by using the clustering algorithm and based on the keyword list to obtain a sample class, and the keyword lists of the sample classes are subjected to set operation to determine a class keyword list;

[0077] The characteristic management difficulty value of the sample class is obtained by calculating the average of the operation difficulty values C of the sample path graphs in the sample class, and taking the average as the characteristic management difficulty value of the sample class;

[0078] The log data of the patient is analyzed to determine patient state information, a priority evaluation model is constructed in combination with a general difficulty value of the medical record path template and a feature management difficulty value of the sample class, and an evaluation priority value of the medical record data of the patient is output.

[0079] The state information includes an in-hospital state and an appointment state; the in-hospital state includes in-hospital and discharged; the appointment state includes already booked and no booking; in which, the in-hospital state is analyzed by acquiring an interval length of the medical record data from the last update, when the interval length is greater than an interval length threshold, the patient is defined as discharged; otherwise, it is changed to in-hospital; for example, when the interval length of the medical record data of the patient is greater than 24 hours, the patient is defined as in the discharged state; the appointment state is whether there is appointment information such as registration record, recheck requirement, etc. in a future preset time length, if there is appointment information, it is defined as already booked; otherwise, it is changed to no booking;

[0080] The method for constructing the priority evaluation model in combination with the general difficulty value of the medical record path template and the feature management difficulty value of the sample class includes:

[0081] The basic priority coefficient Bp of the patient is determined by using a priority coefficient fitting expression, and the priority coefficient fitting expression is as follows:

[0082] In the formula, sp1, sp2 and sp3 are preset priority coefficients; sp1>sp2>sp3;

[0083] The basic priority value P of the medical record data is monitored according to each patient ID, the current process is identified, the real-time medical record path template is determined, and the real-time general management difficulty value YC is acquired;

[0084] It is monitored whether a keyword is generated in the current process, if a keyword is generated, a target keyword list is constructed;

[0085] The class keyword list adapted to the target keyword list is analyzed and acquired, and the real-time feature management difficulty value TC of the corresponding sample class is acquired based on the adapted class keyword list;

[0086] The evaluation priority value of the medical record data is calculated based on the formula Ep=Bp×(k1×YC+k2×TC); in the formula, k1 and k2 are weight coefficients;

[0087] The method for analyzing and acquiring the class keyword list matched with the target keyword list includes:

[0088] A word frequency vector set of the keyword set is created, a union operation is performed on the target keyword list and the class keyword list to obtain a temporary keyword list;

[0089] The target keyword list and the temporary keyword list are respectively converted into a target vector A and a temporary vector B, and the formula The cosine similarity of the target vector A and the temporary vector B is calculated and used as the matching degree of the target keyword list and the category keyword list;

[0090] The category keyword list corresponding to the maximum matching degree between the target keyword list and the category keyword list is obtained, and the category keyword list matched with the target keyword list is obtained;

[0091] The evaluation priority of the medical record data adopts a hierarchical intelligent calling strategy, the medical record data is stored and called, the calling performance is optimized, and the storage level is dynamically adjusted;

[0092] The hierarchical intelligent calling strategy includes setting three priority levels according to the evaluation priority gradient of all case data, and four priority levels corresponding to low priority data, medium priority data, high priority data and extremely high priority data, respectively. Different implementation methods are implemented for different priority data to realize targeted storage and calling;

[0093] For example, offline archiving storage is adopted for low priority data, and asynchronous reading mode is adopted for reading low priority medical record data;

[0094] The medium priority storage layer is used to read the data of the medium priority data, and the strategy of combining batch reading and cache preheating is adopted; wherein the medium priority storage layer can be a storage form of Ceph+SSD cache pool;

[0095] PCIe4.0SSD and enterprise-level SASSSD are combined to realize the hierarchical storage architecture for high priority data, and the intelligent pre-reading strategy based on management complexity is adopted;

[0096] The high-performance storage architecture based on memory computing cluster and all-flash array is adopted for extremely high priority data, and the strategy of global metadata index rapid positioning and multi-level index system acceleration is adopted for retrieval;

[0097] It should be noted that the above storage and calling methods of medical record data are prior art, and will not be described in detail here;

[0098] The calling performance is optimized, and the storage level is dynamically adjusted, including adjusting the storage level based on the evaluation priority of the medical record data, and adjusting the storage level regularly. For example, the medical record data with improved priority or increased management complexity is migrated from low-level storage to high-level storage, such as from medium priority storage to high priority storage;

[0099] By constructing a medical record path template and a medical record data sampling processing, the general management difficulty and the characteristic management difficulty are quantified, and a priority evaluation model is established by combining real-time state of the patient and keyword analysis to dynamically calculate an evaluation priority value of the medical record data, a four-level management strategy is implemented based on the evaluation priority value, a targeted calling and storage method is used for different priority data, the problem of lag and stuttering in operation response of high-priority medical record data of the inpatient is solved, resource allocation of different medical record data is optimized, and management effect of different medical record data is improved.

[0100] In the embodiment II, the intelligent medical record data management system is applied to the intelligent medical record data management method in the embodiment I, and specifically includes:

[0101] The data acquisition and table construction module is used for real-time acquisition of various medical record data of the patient, data preprocessing of the medical record data, determination of medical record data dimensions, and construction of a dimension table.

[0102] The first data processing module is used for taking the dimensions in the medical record data as a process, taking the time and space sequence between the processes as a connection path, and exhaustively obtaining a plurality of medical record path templates.

[0103] The second data processing module is used for random sampling of the medical record data, construction of a sample set, analysis of a sample path graph of each sample, calculation of a general difficulty value of the medical record path template according to the entity association condition of the dimension table and the sample path graph.

[0104] The third data processing module is used for obtaining information of each process factor in the sample path graph and integrating the information to obtain full-process information, processing the full-process information, classifying the sample path graph according to a processing result to obtain a sample class, and calculating a characteristic management difficulty value of the sample class.

[0105] The model construction module is used for analyzing log data of the patient, determining state information of the patient, constructing a priority evaluation model in combination with the general difficulty value of the medical record path template and the characteristic management difficulty value of the sample class, and outputting an evaluation priority value of the medical record data of the patient.

[0106] The data management module is used for adopting a hierarchical intelligent calling strategy for the evaluation priority value of the medical record data, storing and calling the medical record data, optimizing calling performance, and dynamically adjusting a storage level.

[0107] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited thereto, and various changes can be made within the knowledge range of those skilled in the art without departing from the purpose of the present application.

Claims

1. An intelligent medical record data management method, characterized in that, Includes the following steps: Collect various types of patient medical record data in real time, perform data preprocessing on each medical record data, determine the dimensions of the medical record data, and construct a dimension table; By taking the dimensions in the medical record data as the process and the spatiotemporal order between the processes as the connection path, multiple medical record path templates are obtained through exhaustive enumeration. Randomly sample medical record data to construct a sample set, analyze and obtain the sample path diagram for each sample, and calculate the general difficulty value of the medical record path template based on the entity association in the dimension table and the sample path diagram. Information on each process factor in the sample path graph is obtained and integrated to obtain full process information. The full process information is processed to obtain a keyword list for the path. The sample path graph is classified according to the keyword list to obtain sample classes. The feature management difficulty value of the sample classes is calculated. Analyze patient log data to determine patient status information, and construct a priority evaluation model by combining the general difficulty value of the medical record path template and the feature management difficulty value of the sample class, and output the evaluation priority value of the patient medical record data. The evaluation priority of medical record data adopts a hierarchical intelligent retrieval strategy, which manages the storage and retrieval of medical record data, optimizes the retrieval performance, and dynamically adjusts the storage level. Methods for calculating the general difficulty value of medical record path templates include: Obtain the association list of process factors in the sample path graph, and count the number of associations in the association list as the process factor association degree; The correlation degree, data transfer volume during update, time consumption, memory usage, cache hit rate, disk I / O throughput and latency of process factors in the sample path graph are obtained and used as multi-dimensional characterization parameters of process factors. A corresponding characterization parameter list is constructed based on the sample set. The least squares method is used to fit the list of characterization parameters and the list of server load changes to determine the weight of each characterization parameter. The element weights are then normalized. Based on the processed weights, the multidimensional characterization parameters are weighted and summed to obtain the update difficulty value di of the process factor; i is the process factor number, and i is a positive integer. The generation period and corresponding duration of the sample are determined based on the log data. The number of updates of the process factors during the generation period is counted. The update frequency of each process factor in the sample is calculated based on the duration and the number of updates. The mean update frequency of the process factors is calculated based on the sample set and labeled as Ufi. Based on the entity definition in the dimension table, process factors marked as surrogate keys are marked as selected factors. The number of queries for selected factors in the corresponding medical record logs in the sample set is calculated, and the query frequency is calculated based on the time span. The mean query frequency of process factors is calculated based on the sample set and marked as Cfi. Through formula The operational difficulty value C of the sample path graph is calculated; where α, β, and γ are weights. In the sample set, the sample path diagrams are divided according to their template attributes to obtain similar medical record path diagrams. The average operation difficulty value C of the similar medical record path diagrams is calculated to obtain the general difficulty value of medical record data that applies the medical record path template. Similarly, the feature management difficulty value of a sample class is obtained by calculating the mean of the operation difficulty value C of the sample path graph in the sample class.

2. The intelligent medical record data management method according to claim 1, characterized in that, Methods for analyzing and obtaining the sample path graph for each sample include: Randomly sample medical record data to construct a sample set, and obtain log data for each sample in the sample set; The data entities in the process are used as process factors. The process factors contained in different processes in the sample are determined based on log data, and the sample path diagram is determined based on the analysis of the medical record path template set.

3. The intelligent medical record data management method according to claim 1, characterized in that, Status information includes inpatient status and appointment status; inpatient status includes being in the hospital and discharged; appointment status includes having an appointment and not having an appointment.

4. The intelligent medical record data management method according to claim 3, characterized in that, Methods for constructing a priority evaluation model by combining the general difficulty value of medical record path templates and the feature management difficulty value of sample classes include: The patient's baseline priority coefficient Bp is determined using a priority coefficient formulation expression, which is as follows: In the formula, sp1, sp2, and sp3 are preset priority coefficients; sp1>sp2>sp3; Based on the basic priority value P of monitoring medical record data for each patient ID, identify the current process, determine the real-time medical record path template, and obtain the real-time general difficulty value YC. Monitor whether the current process generates keywords; if so, build a list of target keywords. Analyze and obtain a list of class keywords that match the target keyword list, and obtain the real-time feature management difficulty value (TC) of the corresponding sample class based on the matched list of class keywords; Based on formula The evaluation priority value of the medical record data is calculated; where k1 and k2 are weighting coefficients.

5. The intelligent medical record data management method according to claim 4, characterized in that, Methods for analyzing and obtaining a list of keyword categories that match the target keyword list include: Create a word frequency vector set for the keyword set, and perform a union operation on the target keyword list and the class keyword list to obtain a temporary keyword list; The target keyword list and the temporary keyword list are transformed into target vector A and temporary vector B, respectively, using the formula... The cosine similarity between the target vector A and the temporary vector B is calculated and used as the matching degree between the target keyword list and the class keyword list; By comparing and obtaining the list of class keywords corresponding to the maximum matching degree between the target keyword list and the target keyword list, a list of class keywords matching the target keyword list is obtained.

6. The intelligent medical record data management method according to claim 4, characterized in that, The hierarchical intelligent retrieval strategy includes setting four priority levels based on the gradient changes in the evaluation priority values ​​of all medical record data. The four priority levels correspond to low-priority data, medium-priority data, high-priority data, and very high-priority data, respectively, and different targeted storage and retrieval methods are implemented for data of different priorities.

7. The intelligent medical record data management method according to claim 4, characterized in that, Optimize call performance and dynamically adjust storage tiers, including changes in evaluation priority values ​​based on medical record data and periodic adjustments to storage tiers.

8. An intelligent medical record data management system, applied to the intelligent medical record data management method described in any one of claims 1 to 7, characterized in that, Specifically, it includes: The data acquisition and table construction module is used to collect various types of patient medical record data in real time, perform data preprocessing on each medical record data, determine the dimensions of the medical record data, and construct dimension tables; The first data processing module is used to take the dimensions in the medical record data as the process, the spatiotemporal order between the processes as the connection path, and exhaustively obtain multiple medical record path templates. The second data processing module is used to randomly sample medical record data, construct a sample set, analyze and obtain the sample path diagram of each sample, and calculate the general difficulty value of the medical record path template based on the entity association of the dimension table and the sample path diagram. The third data processing module is used to acquire and integrate the information of each process factor in the sample path diagram to obtain the full process information, process the full process information to obtain the keyword list of the path, classify the sample path diagram according to the keyword list to obtain the sample class, and calculate the feature management difficulty value of the sample class. The model building module is used to analyze patient log data, determine patient status information, and build a priority evaluation model by combining the general difficulty value of the medical record path template and the feature management difficulty value of the sample class, and output the evaluation priority value of the patient's medical record data. The data management module is used to prioritize medical record data using a hierarchical intelligent retrieval strategy, manage the storage and retrieval of medical record data, optimize retrieval performance, and dynamically adjust the storage hierarchy.

Images