Patient diagnosis and treatment data management system based on big data

Abstract

The invention discloses a patient diagnosis and treatment data management system based on big data, and relates to the technical field of data management. Quality evaluation is carried out by calculating three types of indexes; the semantic deviation degree of the aggregated data and the original data is measured through a semantic inverse deduction ambiguity index; evaluating connectivity and integrity of diagnosis and treatment events in the aggregated data through an information integrity closed-loop index; and judging whether repeated redundant fields exist in the data or not through the information redundancy index. And calculating a qualification index by integrating the three indexes, judging whether the current aggregated data meets semantic mapping and fusion requirements or not, storing if the current aggregated data is qualified, and triggering a remapping and reaggregation process if the current aggregated data is not qualified, thereby ensuring that the finally stored diagnosis and treatment data is accurate in semantics, complete in structure and less in invalid data, improving the data quality, avoiding fracture of a diagnosis and treatment chain, and improving the diagnosis and treatment efficiency. And the negative influence on subsequent analysis and decision is reduced.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and more specifically to a patient diagnosis and treatment data management system based on big data. Background Technology

[0002] With the development of big data technology, patients' medical data can be automatically saved digitally, making treatment more convenient for doctors. Simultaneously, big data technology can aggregate structured and unstructured medical data generated by multiple business subsystems across various levels of medical institutions into a single main system. This main system, leveraging a unified data standard system, semantic mapping mechanism, and cleaning and aggregation algorithms, achieves unified integration, management, and utilization of patients' full-process, multi-dimensional medical data. The system aims to break down the fragmented and scattered nature of traditional medical information, constructing a comprehensive, complete, and consistent patient medical data resource pool. This breaks down data silos, improves the efficiency of medical resource collaboration, and provides high-quality data support for clinical decision support, medical quality control, scientific research analysis, and medical insurance supervision. By semantically mapping and cleaning the data from different subsystems, the main system not only breaks down data silos but also removes repetitive medical data, reducing redundant data and storage space consumption.

[0003] Therefore, in practical applications, the semantic mapping and aggregation of patient diagnosis and treatment data uploaded from various subsystems by the main system is crucial for the main system to completely and accurately store all patient diagnosis and treatment data. However, in practice, there is generally a lack of a self-checking mechanism to judge the quality of semantic mapping and aggregation. If the semantic mapping and aggregation results of the patient diagnosis and treatment data uploaded from various subsystems by the main system are unqualified, it may lead to low accuracy or completeness of the final patient diagnosis and treatment data stored by the main system, or even lead to some data breaks, and even affect subsequent diagnosis and treatment. Summary of the Invention

[0004] The purpose of this invention is to solve the problems mentioned above and provide a patient diagnosis and treatment data management system based on big data.

[0005] This invention proposes a patient diagnosis and treatment data management system based on big data, the system comprising:

[0006] Semantic ambiguity module: The patient diagnosis and treatment data of each subsystem aggregated by the main system are recorded as aggregated data, and the semantic back-back ambiguity index of the aggregated data is calculated based on the aggregated data; the semantic back-back ambiguity index is used to measure the degree of semantic difference between the diagnosis and treatment data derived from the aggregated data and the diagnosis and treatment data expressed by the original subsystem.

[0007] Information integrity closed-loop module: Extracts patient diagnosis and treatment event nodes from aggregated data, analyzes the diagnosis and treatment event nodes, and calculates the information integrity closed-loop index. The information integrity closed-loop index is used to measure the connectivity and integrity of each diagnosis and treatment event in the aggregated patient diagnosis and treatment data.

[0008] Information redundancy module: Extracts fields from each medical record from the aggregated data and calculates the information redundancy index, which is used to measure the degree of redundant data in the aggregated data;

[0009] Data Management Module: Calculates the qualification index of aggregated data based on semantic back-inference ambiguity index, information redundancy index, and information integrity closed-loop index, and determines the final management of aggregated data by the main system based on the qualification index.

[0010] Optionally, the semantic ambiguity module includes:

[0011] Semantic Unit Module: Semantically analyzes each medical record in the aggregated data and divides it into multiple semantic units;

[0012] Reverse Path Module: Constructs a reverse semantic path tree based on aggregated semantic units; uses all paths from leaf nodes to the root node in the path tree as the set of reverse semantic paths corresponding to all semantic units;

[0013] Path overlap ratio module: Counts the total number of duplicate paths in the reverse semantic path set. For the total number of all reverse semantic paths in the reverse semantic path set, the total number of duplicate paths is divided by the total number of all reverse semantic paths to obtain the path overlap ratio.

[0014] First calculation module: Calculates the semantic back-back ambiguity index based on the path overlap ratio.

[0015] 3. The patient diagnosis and treatment data management system based on big data according to claim 2, wherein the first calculation module comprises:

[0016] Root Mean Square Module: For each pair of semantic units, define the semantic edit distance between the corresponding reverse semantic paths, and calculate the root mean square of the semantic edit distance between all pairs of semantic units;

[0017] Semantic back-calculation ambiguity index module: based on root mean square and path overlap ratio The formula for calculating the semantic back-retrospective ambiguity index is as follows: In the formula, This is the semantic back-reasoning ambiguity index.

[0018] Optionally, the information complete closed-loop module includes:

[0019] Set module: This module groups patient treatment events from the aggregated data into a set in chronological order, where 𝑞 represents the total number of patient treatment events in the set; it also constructs a transition representation matrix. If the patient's medical events and If the conditions between them conform to the preset rules, then If it does not conform to the preset rules, then ;

[0020] Closed-loop access factor module: in all and If there exists a path from Departed and returned If a closed-loop path is found, then the path is recorded. For each closed-loop path pair, ... Calculate the number of event nodes traversed by the closed-loop path, denoted as . Constructing closed-loop access factors : ;

[0021] Closed-loop event coverage module: Extract the total number of event nodes covered by all closed-loop paths as the total number of covered nodes, and divide the total number of covered nodes by the total number of patient diagnosis and treatment events to obtain the closed-loop event coverage.

[0022] The second calculation module calculates the information completeness index based on the closed-loop accessibility factor and closed-loop event coverage.

[0023] Optionally, the second computing module includes:

[0024] Compression ratio module: from matrix Extracting submatrices that are only covered by closed-loop paths Count the number of all jump pairs with a value of 1 in the submatrix; then divide the number of all jump pairs with a value of 1 in the submatrix by the maximum number of jumps that a complete closed loop should have under ideal conditions to obtain the local loop compression ratio.

[0025] Information Completeness Closed-Loop Index Module: The information completeness closed-loop index is obtained by multiplying the mean of the closed-loop accessibility factors of all closed-loop path pairs by the closed-loop event coverage rate and the local loop compression rate.

[0026] Optionally, the information redundancy index includes:

[0027] Encoding Matrix Module: Extracts patient data from aggregated data Medical records, each record contains Each record is treated as a vector of key fields, forming a... The field matrix is ​​obtained by discretely encoding each column of the field and converting it into a pure integer representation.

[0028] Redundancy Coordination Strength Module: For any two fields in the encoding matrix and Calculate two fields and The redundancy coordination strength between them is used to obtain the field pair redundancy matrix; the calculation formula is: ,in, For fields and Redundancy coordination strength between them Indicates that in all record pairs, the field and The number of records with the same value at the same time; This represents the logarithm of all pairwise combinations of medical records;

[0029] Redundancy Core Set Module: Sorts the redundancy coordination strength of all field pairs by size and extracts the top-ranking redundancy pairs. The set of field pairs forms a redundant core set;

[0030] Optionally, the information redundancy index further includes:

[0031] Stable Redundancy Factor Module: For all medical records, if their values ​​on the field pairs involved in the set are completely consistent, they constitute a set of "redundant mappings"; let the total number of records formed be... There are such mapping volumes, each containing Medical records; calculate the stability redundancy factor: In the formula, As a stable redundancy factor;

[0032] Information Redundancy Index Module: Calculates the average redundancy coordination strength of all field pairs in the stable redundancy factor. And calculate the information redundancy index, the formula for which is: In the formula, This is the information redundancy index.

[0033] Optionally, the data management module includes:

[0034] The qualification index module normalizes the semantic back-derivative ambiguity index, information redundancy index, and information integrity closed-loop index of the aggregated data, removes units, and maps the values ​​to the 0-1 range. Then, it weights and sums the normalized semantic back-derivative ambiguity index, information redundancy index, and information integrity closed-loop index of the aggregated data to obtain the qualification index.

[0035] Optionally, the data management module further includes:

[0036] The first management module compares the pass index with the preset threshold. If the pass index is not less than the preset threshold, it means that the semantic mapping of the patient diagnosis and treatment data uploaded by each subsystem by the main system is qualified with the aggregated data. Then the main system stores and manages the current aggregated data.

[0037] Optionally, the data management module further includes:

[0038] Second management module: If the pass index is less than the preset threshold, it means that the semantic mapping of the patient diagnosis and treatment data uploaded by each subsystem by the main system and the aggregated data are not qualified. In this case, the main system needs to re-semantically map and aggregate the patient diagnosis and treatment data uploaded by each subsystem until the pass index of the aggregated data is not less than the preset threshold.

[0039] The beneficial effects of this invention are:

[0040] This invention proposes a patient diagnosis and treatment data management system based on big data. The main system aggregates patient diagnosis and treatment data uploaded from various subsystems into unified aggregated data and performs quality assessment by calculating three types of indices: a semantic back-engineering ambiguity index to measure the degree of semantic deviation between the aggregated data and the original data; an information integrity closed-loop index to assess the connectivity and completeness of diagnosis and treatment events in the aggregated data; and an information redundancy index to determine whether there are duplicate or redundant fields in the data. A pass index is calculated based on these three indicators to determine whether the current aggregated data meets the requirements for semantic mapping and fusion. If it passes, it is stored; if it fails, a remapping and re-aggregation process is triggered to ensure that the final stored diagnosis and treatment data is semantically accurate, structurally complete, and contains little invalid data, thereby improving data quality, avoiding breaks in the diagnosis and treatment chain, and reducing the negative impact on subsequent analysis and decision-making. Attached Figure Description

[0041] The invention will now be further described with reference to the accompanying drawings.

[0042] Figure 1 This is a framework diagram of a patient diagnosis and treatment data management system based on big data. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] This invention provides a patient diagnosis and treatment data management system based on big data. See also... Figure 1 , Figure 1A framework diagram of a patient diagnosis and treatment data management system based on big data provided in an embodiment of the present invention. The system includes:

[0045] Semantic Ambiguity Module: The patient diagnosis and treatment data of each subsystem aggregated by the main system are recorded as aggregated data. Based on the aggregated data, the semantic back-reasoning ambiguity index of the aggregated data is calculated to measure the degree of semantic difference between the diagnosis and treatment data derived from the aggregated data and the diagnosis and treatment data expressed by the original subsystem.

[0046] Information integrity closed-loop module: Extracts patient diagnosis and treatment event nodes from aggregated data, analyzes these nodes, and calculates the information integrity closed-loop index, which measures the connectivity and integrity of each diagnosis and treatment event in the aggregated patient diagnosis and treatment data.

[0047] Information Redundancy Module: Extracts fields from each medical record in the aggregated data and calculates the information redundancy index to measure the degree of redundant data in the aggregated data;

[0048] Data Management Module: Calculates the qualification index of aggregated data based on semantic back-inference ambiguity index, information redundancy index, and information integrity closed-loop index, and determines the final management of aggregated data by the main system based on the qualification index.

[0049] Based on the big data-based patient diagnosis and treatment data management system provided in this embodiment of the invention, the above-mentioned method, through the setting of a self-checking mechanism for semantic mapping and aggregation qualification judgment, determines whether the semantic mapping and aggregation results of the patient diagnosis and treatment data uploaded from each subsystem by the main system are qualified, ensuring that the accuracy or completeness of the final patient diagnosis and treatment data stored by the main system is high, and even reducing the possibility of data breakage in diagnosis and treatment, and reducing the impact on subsequent diagnosis and treatment effects.

[0050] In one embodiment, the semantic ambiguity module: records the patient diagnosis and treatment data of each subsystem aggregated by the main system as aggregated data, and calculates the semantic back-reasoning ambiguity index of the aggregated data based on the aggregated data, which is used to measure the degree of semantic difference between the diagnosis and treatment data derived from the aggregated data and the diagnosis and treatment data expressed by the original subsystem.

[0051] It's important to note that the main system aggregates patient treatment data by integrating data from multiple subsystems. Specifically, these subsystems include electronic health records, laboratory information management systems, medical imaging systems, and medication management systems. These subsystems record patient treatment information such as diagnoses, medication use, examination reports, and surgical records. The main system first receives structured and unstructured data uploaded from each subsystem through a unified data interface and standardized protocols. It then preprocesses and cleans the data, removing invalid or duplicate content. Next, the main system uses semantic mapping technology to convert data from different systems into a unified data format, ensuring compatibility and consistency between the data from each subsystem. During semantic mapping, the main system uniformly identifies and standardizes similar types of treatment information from different systems (such as a diagnosis of "hypertension" or the name of a certain drug), thereby achieving data fusion between different systems. For unstructured data (such as doctor's notes or imaging reports), the main system extracts key information using natural language processing technology and converts it into structured data. This processed data is integrated into a unified database, forming aggregated data. This aggregated dataset contains multi-dimensional, end-to-end patient treatment information.

[0052] In one implementation, the semantic ambiguity module includes:

[0053] Semantic unit module: aggregates data Each medical record in the database undergoes semantic parsing and is segmented into multiple semantic units (e.g., "diagnosis-location-severity" or "examination-location-indicator"); a set of semantic units is defined. ,in, For the first Each semantic unit is extracted from aggregated data. ; This represents the total number of semantic units extracted from the aggregated data, i.e., the number of elements in the semantic unit set;

[0054] Reverse Path Module: Based on the aggregated semantic units in the semantic unit set, a reverse semantic path tree is constructed. This tree represents the semantic restoration path structure tree constructed when mapping aggregated data back to the original subsystem data. It simulates the path set (such as the original structure, order, and representation) from which the original subsystem generates these semantic units. The path from all leaf nodes to the root node in the path tree is defined as follows: ,in, This is the set of reverse semantic paths corresponding to all semantic units. The reverse path used to reconstruct semantic units;

[0055] Path overlap ratio module: This module counts the total number of duplicate paths in the reverse semantic path set. For each reverse semantic path in the set, the total number of duplicate paths is divided by the total number of all reverse semantic paths to obtain the path overlap ratio. In the formula, This represents the path overlap ratio. This indicates the number of duplicate paths. More path overlap (i.e., multiple semantic units are incorrectly attributed to the same path) indicates a more severe ambiguity or flattening in the aggregation process; where the number of duplicate paths refers to the set... The number of times the same path appears in the text;

[0056] First calculation module: Calculates the semantic back-back ambiguity index based on the path overlap ratio.

[0057] In one implementation, the first computing module includes:

[0058] Root mean square module: for each pair of semantic units Define the semantic edit distance between semantic reconstruction paths. (This can be constructed based on differences in the concept hierarchy of the medical ontology, differences in description granularity, field jump count, etc.), calculate the root mean square of the semantic divergence between path pairs. , ; The larger the value, the greater the structural differences of each diagnostic element when reverse-engineered from the aggregated data, meaning the more distorted the aggregation process is;

[0059] Semantic back-calculation ambiguity index module: based on root mean square and path overlap ratio The formula for calculating the semantic back-retrospective ambiguity index is as follows: In the formula, This is the semantic back-reasoning ambiguity index.

[0060] It should be noted that the above calculation process is illustrated with an example. Semantic parsing is performed on each medical record in the aggregated data, segmenting it into semantic units with medical semantic structures, such as "chief complaint – persistent cough," "examination – chest X-ray – patchy shadow in the right lower lung," and "diagnosis – pneumonia." Semantic units can be extracted directly from structured fields, such as the "chief complaint" field in admission records and the "impression" field in examination reports. For unstructured text such as medical records and discharge summaries, natural language processing tools are used for sentence segmentation, phrase extraction, and entity recognition, segmenting complex text into phrase-type semantic units with diagnostic and treatment logic. All segmented semantic units constitute a set S, and the total number of units n in the set is an important parameter for subsequent calculations. For example, if an aggregated record contains "admitted due to fever for three days", "physical examination T 39℃", "laboratory test showed elevated white blood cell count", and "diagnosed as viral pneumonia", it can be divided into four semantic units: "chief complaint - fever", "examination - body temperature - 39℃", "laboratory test - blood routine - elevated white blood cell count", and "diagnosis - viral pneumonia", where n is 4.

[0061] After obtaining the set of semantic units, a reverse semantic restoration path is constructed for each semantic unit. This involves starting from the current aggregated expression and simulating the possible original subsystem paths from which it originates, forming a path tree T. The path tree construction is based on the data structure specifications, field hierarchy, and document templates of different subsystems in medical institutions, such as EMR, LIS, and PACS. For example, "chief complaint - fever" might originate from the "chief complaint" field in the admission record module of the EMR system; "examination - body temperature - 39℃" might originate from the physical signs section in the physical examination template; "laboratory test - complete blood count - elevated white blood cell count" comes from the laboratory test results item in the LIS system; and "diagnosis - viral pneumonia" originates from the diagnosis field or ICD code field in the EMR. In this way, for each semantic unit s1 to s2, a reverse semantic restoration path is constructed. n Construct the reverse path from p1 to p n This forms a path set P. The construction of paths is not a mechanical mapping, but rather a knowledge-based reasoning that combines the meaning of semantic units with the historical data structure, with the aim of reconstructing the possible expression channels of aggregated data in the original acquisition stage;

[0062] After path construction is completed, the path overlap ratio μ is calculated to determine whether multiple semantic units have been oversimplified and reduced to the same or similar paths, reflecting flattening or semantic compression during the aggregation process. First, all paths in the path set P are structurally normalized, for example, by stripping away specific values ​​and content, retaining only the "system-module-field" format. Then, the number of duplicate paths is counted, defined as the "number of duplicate paths." The number of duplicate paths is then divided by the total number of paths n to obtain the path overlap ratio μ. For example, if 3 out of 6 semantic units in a certain aggregated data point to the EMR.admission record / diagnosis field path, the number of duplicate paths is 2 (repetition count minus 1), and μ = 2 / 6 ≈ 0.33. A high path overlap ratio means that multiple semantic elements are merged into the same treatment path during the aggregation process, indicating semantic compression, which may mask the structural diversity that should exist in the original data. An edit distance matrix M between semantic paths is constructed, and the root mean square δ of semantic divergence is calculated.

[0063] The difference between each pair of semantic unit paths pᵢ and pⱼ is measured by defining the semantic edit distance dᵢⱼ. This difference can be quantified based on rules such as the number of hierarchical jumps in the medical ontology (e.g., a single-level jump from "pneumonia" to "viral pneumonia," and a two-level jump from "fever" to "respiratory infection"), expression granularity (e.g., specific pathogen vs. general diagnosis), and the number of cross-system path jumps (e.g., a single-system jump from LIS to EMR). Taking the paths "EMR.Admission Record.Diagnosis Field" and "PACS.Image Report.Description Field" as examples, they differ in system source, module name, and field level, so d can be set to 1.0; while "EMR.Progress Record.Diagnosis Field" and "EMR.Admission Record.Diagnosis Field" differ only in module, so d is set to 0.3. The sum of the d² values ​​between all pairs, divided by n(n-1) / 2, and then squared, yields the δ value. For example, if 4 semantic units form 6 path combinations, and the corresponding d² sum is 3.6, then δ = sqrt(23.6 / 43) = sqrt(0.6) ≈ 0.77. The larger the δ value, the greater the difference between semantic units during the reverse reconstruction process, and the lower the semantic consistency.

[0064] It's important to note that the Semantic Reverse Ambiguity Index is a quantitative indicator that measures whether aggregated data can be accurately and unambiguously reverse-engineered back to the semantic structure of the original subsystems. It reflects the degree of semantic loss, semantic flattening, or fuzzy mapping caused by the main system during the aggregation process of patient diagnosis and treatment data uploaded from various subsystems. In the process of aggregating data from multiple subsystems, due to differences in diagnosis and treatment structure, field naming, expression granularity, and semantic hierarchy among different subsystems, even if they describe the same or similar diagnosis and treatment facts, they may be integrated into seemingly consistent but semantically condensed or distorted records after unified aggregation by the main system. The Semantic Reverse Ambiguity Index comprehensively reflects the difficulty of semantic restoration by constructing a reverse semantic path tree and analyzing its path overlap (i.e., multiple diagnosis and treatment units from different sources or with different structures are incorrectly merged into the same path after aggregation), combined with the semantic edit distance between each pair of semantic paths (i.e., the minimum operational cost required to transform one representation into another in the semantic space). A higher index value indicates a more dispersed, overlapping, and farther edit distance in the path structure required to deduce the patient's treatment records from the original subsystem within the aggregated data. This suggests that the aggregation process weakens the structure and semantic clarity of the original treatment data. Such ambiguous or distorted aggregation leads to unclear and incoherent understanding of the patient's treatment history by the main system, reducing data usability, decision support accuracy, and the reliability of subsequent data mining and analysis. Therefore, a higher semantic back-dive fuzziness index indicates poorer quality of the current aggregated data, with poor semantic consistency, difficulty in structural reconstruction, and ambiguous context, directly reflecting the main system's inadequacy in the data semantic fusion process.

[0065] In one embodiment, the information integrity closed-loop module extracts patient diagnosis and treatment event nodes from the aggregated data, analyzes the diagnosis and treatment event nodes, and calculates the information integrity closed-loop index, which is used to measure the connectivity and integrity of each diagnosis and treatment event in the patient diagnosis and treatment data in the aggregated data.

[0066] In one implementation, the information integrity closed-loop module includes:

[0067] Collection module: This module organizes the patient's diagnosis and treatment events from the aggregated data into a collection in chronological order. ,in, Let represent the _n-th treatment event, and _n represent the total number of treatment events for the patient; construct a q×q transition representation matrix. Among them, if the patient's medical treatment event and The two events conform to a preset rule, meaning the preset rule is that the two events have a clear...

[0068] If medical causal logic or a reasonable temporal sequence (e.g., "examination" followed by "confirmation") is followed by... If there is no logical order or the event is invalid, then ;

[0069] Closed-loop accessibility factor module: Extracts closed-loop path pairs and constructs a set of closed-loop accessibility factors: in all and If there exists a path from After departing, other events occurred, and then it returned. If a closed-loop path is found, then record that path. For each closed-loop path pair; Calculate the number of event nodes traversed by the path (i.e., the length of the closed-loop path), denoted as . Constructing closed-loop access factors : ;

[0070] Closed-loop event coverage module: Extracts the total number of event nodes covered by all closed-loop paths as the total number of covered nodes, and divides the total number of covered nodes by the total number of patient diagnosis and treatment events. Obtain closed-loop event coverage , ; This indicates the percentage of medical events successfully captured by the closed-loop structure, measuring the "connectivity and coverage" of medical information.

[0071] Second calculation module: Based on closed-loop accessibility factor Closed-loop event coverage Calculate the information completeness closed-loop index.

[0072] In one implementation, the second computing module includes:

[0073] Compression ratio module: from matrix Extracting submatrices that are only covered by closed-loop paths Count the number of all jump pairs with a value of 1 in this submatrix, denoted as 𝜉; then divide 𝜉 by the maximum number of jumps that a complete closed loop should have under ideal conditions to obtain the local loop compression ratio. The higher the compression ratio, the more complete the information jump in the closed-loop region and the more compact the information structure.

[0074] Information Completeness Closed-Loop Index Module: The information completeness closed-loop index is obtained by multiplying the mean of the closed-loop accessibility factors of all closed-loop path pairs by the closed-loop event coverage rate and the local loop compression rate. The formula is as follows: In the formula, This is the information completeness and closed-loop index. A higher information completeness and closed-loop index value indicates that there are more effective closed loops between events in the treatment pathway, and that the structure is compact and comprehensive.

[0075] It should be noted that the above calculation process is illustrated with an example. In calculating the information integrity closed-loop index, the first step is to collect patient treatment data from various medical information systems, including electronic medical records, examination and testing systems, prescription systems, and inpatient systems. This data is then uniformly sorted by timestamps to form a complete sequence of treatment events arranged chronologically. For example, a patient's treatment record might sequentially include events such as "initial outpatient visit," "complete blood count," "tumor marker test," "CT scan," "diagnosis," "chemotherapy," "follow-up," and "follow-up." These events are numbered to form an ordered set with treatment events as nodes. The system analyzes whether there is a reasonable medical logical relationship or temporal order between any two treatment events. For example, "examination" is followed by "diagnosis," and "diagnosis" is followed by "treatment," a sequence supported by medical causal logic. If a clear logical relationship exists between two events, it is marked as a valid jump in the jump representation matrix; if there is no direct logical relationship between the two events or the event itself is invalid (e.g., duplicate entry, incorrect operation, irrelevant treatment), it is marked as no jump. This creates a matrix that reflects the jump structure of the diagnosis and treatment process, which is used to support the subsequent extraction of closed-loop paths.

[0076] Based on this jump structure, the system identifies event combinations that can form closed-loop paths from all possible event pairs. A closed-loop path refers to a logical or temporal loop that starts from a certain event, passes through several other events, and then returns to the original event. For example, starting from "initial diagnosis," passing through "examination," "diagnosis," "treatment," and "follow-up," and returning to "initial diagnosis," constitutes a complete diagnosis-treatment closed loop. This type of closed loop is common in actual medical practice, especially in scenarios requiring periodic assessment and adjustment of treatment strategies, such as chronic disease management or cancer treatment. All such path pairs are recorded, forming a set of closed-loop path pairs. For each closed-loop path, the system calculates the number of diagnosis-treatment events experienced in the path, which is the length of the path. The shorter the path, the closer the connection between events, and the more compact the closed loop. The system assigns an accessibility factor to each path; the larger this factor, the higher the information flow efficiency of the path, indicating a more standardized and coherent diagnosis-treatment process.

[0077] The system then extracts the set of diagnostic and treatment events covered by all closed-loop paths. In other words, it counts which event nodes appear in at least one closed-loop path. Comparing the size of this set with the total number of events yields the closed-loop event coverage rate, reflecting the proportion of event nodes in the diagnostic and treatment data that participate in an effective closed-loop structure. For example, if a patient has a total of 10 diagnostic and treatment events, and 6 of them appear in some closed-loop path, then the closed-loop event coverage rate is 60%. The higher this proportion, the more complete and organized the patient's diagnostic and treatment path.

[0078] Furthermore, the system extracts a submatrix from the original jump matrix, consisting only of events covered by the closed-loop path, to analyze the jump structure within the closed-loop region. In this submatrix, the system counts all instances indicating jumps between events, i.e., the number of valid jumps. This number is then compared to the theoretically maximum number of jumps between these events to obtain a local loop compression ratio. A higher compression ratio indicates a tighter connection between events within these closed-loop regions, more complete information jumps, and a more compact structure. Finally, the system integrates the compactness of the closed-loop path, the coverage of events, and the structural density within the closed-loop region to evaluate the closed-loop integrity of the entire diagnostic and treatment data in terms of information structure. A higher information integrity closed-loop index indicates a tighter data chain in the diagnostic and treatment process, clearer relationships between events, more logical consistency, and better overall data quality; conversely, a lower index suggests potential missing, disconnected, or invalid jumps in diagnostic and treatment information. The data acquisition throughout the calculation process relies on the actual patient's historical medical history. Through the standardization of medical behavior coding, the pre-setting of event logic rules, and the strict arrangement of time series, the calculation of the information integrity closed-loop index is ensured to have a clear data foundation and logical basis.

[0079] It's important to note that the Information Completeness and Loop Retention Index is an indicator used to measure whether the diagnostic and treatment events in aggregated data form a semantically reasonable, structurally coherent, and sufficiently comprehensive closed-loop path structure. Its core focus is on the "semantic and temporal closure and connectivity of diagnostic and treatment information." The underlying concept is that a patient's complete diagnostic and treatment process should reflect a series of events with logical sequence, causal chains, and feedback links, such as "initial diagnosis—examination—diagnosis—treatment—follow-up." If the transitions between these events can form a loop structure in terms of time and semantics—that is, if the information has the characteristics of "starting point—development—feedback—closure"—it indicates that the diagnostic and treatment path is sufficiently covered and logically closed, and the data has high completeness and usability. Conversely, a smaller Information Completeness and Loop Retention Index means that many diagnostic and treatment events exist in isolation, failing to form an organic structure and lacking a semantic path closure and feedback mechanism. The diagnostic and treatment information may exhibit fragmented and disconnected characteristics, which significantly weakens the semantic quality of the aggregated data. On the one hand, such aggregated data cannot accurately present the context of the original diagnosis and treatment process of each subsystem, which can easily lead to semantic misjudgment and path loss. The main system will then struggle to identify whether a patient has completed the key diagnosis and treatment loop during analysis. On the other hand, when facing applications such as subsequent clinical decision support, path optimization recommendations, and disease progression analysis, the models built by the main system based on this data with missing loops may also be misleading. Therefore, a lower information completeness loop index not only indicates structural defects in the aggregated data but also essentially reflects that the original diagnosis and treatment logic has been weakened or omitted during the aggregation process, thus becoming an important reference for measuring whether the main system's data aggregation quality is up to standard. When the index is high, it indicates that the diagnosis and treatment data uploaded by each subsystem has been largely successfully incorporated into a reasonable closed-loop path during the aggregation process, and the information possesses strong structural consistency, causal completeness, and temporal coherence. The aggregation results have good traceability and semantic reliability.

[0080] In one embodiment, the information redundancy module extracts fields from each medical record in the aggregated data and calculates an information redundancy index to measure the degree of redundant data in the aggregated data.

[0081] In one implementation, the information redundancy module includes:

[0082] Encoding Matrix Module: Extracts patient data from aggregated data Medical records, each record contains Key fields, such as "visit time," "department," and "diagnosis code," are used to construct a vector of fields for each record. The field matrix is ​​used for diagnosis and treatment; each column field is discretely encoded and converted into a pure integer representation to obtain the encoding matrix;

[0083] Redundancy Coordination Strength Module: For any two fields in the encoding matrix and Calculate two fields and The redundancy coordination strength between them is used to obtain the field pair redundancy matrix; the calculation formula is: ,in, For fields and Redundancy coordination strength between them This indicates that in all medical record pairs, the field and The number of records with the same value at the same time; This represents the logarithm of all pairwise combinations of medical records;

[0084] Redundancy Core Set Module: Sorts the redundancy coordination strength of all field pairs by size and extracts the top-ranking redundancy pairs. The set of field pairs forms a redundancy core set; it represents the region where the strongest redundancy structure exists between fields.

[0085] Stable Redundancy Factor Module: For all medical records, if their values ​​on the field pairs involved in the set are completely consistent, they constitute a set of "redundant mappings"; let the total number of records formed be... There are such mapping volumes, each containing Medical records; calculate the stability redundancy factor: In the formula, As a stable redundancy factor;

[0086] Information Redundancy Index Module: Calculates the average redundancy coordination strength of all field pairs in the stable redundancy factor. And calculate the information redundancy index, the formula for which is: In the formula, This is the information redundancy index.

[0087] It should be noted that the above calculation process is illustrated with an example. The core field information of each record is extracted from the aggregated raw medical data. These fields typically include the consultation time, the department performing the treatment, the diagnosis code, the cost category, and the treatment items. Each record is considered a field vector composed of these fields, and all records form a field matrix. Next, each column of fields is discretized, that is, different values ​​in each field are mapped to different integer codes, forming an encoded field matrix. The purpose is to make subsequent calculations more efficient and comparable. For example, suppose there are three records, all with a consultation time of "July 1, 2025", departments of "Internal Medicine", "Internal Medicine", and "Surgery" respectively, and diagnosis codes of "J18.0", "J18.0", and "K35.9" respectively. After encoding, "Internal Medicine" can be recorded as 1, "Surgery" as 2, "J18.0" as 1, and "K35.9" as 2, resulting in an encoded matrix. Next, any two fields (e.g., "Department of Execution" and "Diagnosis Code") will be randomly selected from this coding matrix. All medical records will be iterated through in pairs. For each pair, the values ​​of these two fields will be checked to see if they are identical. If they are identical, a redundancy coordination will be accumulated. After iterating through all record pairs, the redundancy degree between these two fields will be calculated. This comparison will be performed on all pairs of fields, generating a redundancy coordination matrix for field pairs. The redundancy values ​​of all field pairs in this matrix will then be sorted, and several pairs of fields with the highest redundancy degree (e.g., 3 pairs) will be selected to form a "redundancy core set," representing the field combination area with the strongest repetition and the most obvious potential redundancy structure in the medical records. Subsequently, for all medical records, a field combination label will be constructed on these redundant field pairs. For example, if a record has "Internal Medicine + J18.0" for "Department of Execution" and "Diagnosis Code," and another record has the same value for these two fields, then these two records are considered to belong to the same "redundancy mapping body." After grouping all records according to these field value combination labels, a series of redundant mapping bodies can be identified. Each mapping body contains several records with completely identical field values. For example, in the three records mentioned above, records 1 and 2 form one mapping body because their "Department of Execution" and "Diagnosis Code" are identical, while record 3 is in an independent group. By counting the number of all mapping bodies and the number of records contained in each, we can further measure whether the redundancy distribution is concentrated and stable. Finally, by combining the average redundancy coordination strength between redundant core field pairs with the stability of the redundant mapping body structure, an overall information redundancy index is calculated. The level of this index directly reflects the severity of redundant content in the aggregated data. If the index is high, it indicates that a large number of records have high repetition in key fields, posing a potential risk of data redundancy. This can trigger actions such as optimizing the aggregation algorithm, data cleaning, or adjusting the fusion rules.Throughout the process, each data acquisition point relies on the structure of the original fields in the medical records, the consistency of the coding, and the clear definition of the combination rules.

[0088] It's important to note that the information redundancy index is a composite indicator used to measure the degree of redundant information distribution in aggregated data. It reflects whether, after semantic mapping and aggregation of patient treatment data uploaded from multiple subsystems, a large number of data patterns emerge with highly consistent values ​​in key fields, a lack of diversity, and significant repetitive structures. A higher information redundancy index indicates that more records in the treatment data have completely identical values ​​in redundant core field pairs. This high degree of consistency often does not stem from the natural similarity of patient behavior or treatment paths in the real world, but may be due to data fusion errors caused by factors such as inaccurate semantic mapping, distorted field correspondences, and loss of key information during structure transformation. For example, fields with the same meaning in different subsystems may have inconsistent formats before uploading. If the main system fails to accurately distinguish or merge these semantic differences during semantic mapping, records that should belong to different categories may be incorrectly aggregated into a seemingly identical group, resulting in structural redundancy. Similarly, if the actual meanings of "department" and "diagnosis code" differ slightly in different subsystems, but are simply mapped to the same value after aggregation, multiple patient records may form redundant mappings for these two fields. These phenomena lead the main system to mistakenly believe that a large number of records are structurally identical, thus increasing the information redundancy index. Therefore, a higher information redundancy index often indicates a problem with the semantic mapping of the main system, and that the data fusion process has failed to accurately reproduce the true differences in the information uploaded by the subsystems. This results in reduced data quality, lack of representativeness, or misleading subsequent analysis, leading to the conclusion that the quality of the aggregated data is increasingly substandard. This is why the information redundancy index serves as an important basis for evaluating the accuracy of the semantic mapping of aggregated data and the quality of fusion.

[0089] In one embodiment, the data management module includes:

[0090] The qualification index module normalizes the semantic back-draft ambiguity index, information redundancy index, and information integrity closed-loop index of the aggregated data, removing units and mapping the values ​​to the 0-1 range. It then weights and sums these normalized indices to obtain the qualification index. The calculation formula is as follows: In the formula The passing index, These are the normalized semantic back-reasoning ambiguity index, information redundancy index, and information integrity closed-loop index, respectively. These represent the preset weight coefficients for the semantic back-reasoning ambiguity index, information redundancy index, and information integrity closed-loop index of the aggregated data after multiple normalization processing, respectively. All are greater than 0; To minimize the value and prevent the denominator from being zero, it is generally set to a value of [value missing]. ;

[0091] It should be noted that commonly used methods for removing dimensions include Min-Max normalization and Z-Score standardization, which will not be elaborated here. Settings should be set according to the actual situation, generally They are equal and their sum is 1, for example, All can be , , .

[0092] In one embodiment, the data management module further includes:

[0093] The first management module compares the pass index with the preset threshold. If the pass index is not less than the preset threshold, it means that the semantic mapping of the patient diagnosis and treatment data uploaded by each subsystem by the main system is qualified with the aggregated data. Then the main system stores and manages the current aggregated data.

[0094] In one embodiment, the data management module further includes:

[0095] Second management module: If the pass index is less than the preset threshold, it means that the semantic mapping of the patient diagnosis and treatment data uploaded by each subsystem by the main system and the aggregated data are not qualified. In this case, the main system needs to re-semantically map and aggregate the patient diagnosis and treatment data uploaded by each subsystem until the pass index of the aggregated data is not less than the preset threshold.

[0096] It should be noted that during the aggregated data quality management process, the main system calculates and analyzes a pass / fail index to determine whether the current aggregated data meets the quality requirements for semantic mapping and fusion consistency. First, the main system compares the pass / fail index of the current aggregated data with a preset threshold. If the pass / fail index is not less than the threshold, it indicates that the patient diagnosis and treatment data uploaded by each subsystem has been well mapped and fused in terms of semantic expression and structural organization, and the aggregation result has high semantic consistency and data completeness. Therefore, the main system can directly store and archive this aggregated data as an effective data source for subsequent analysis and decision-making. However, if the pass / fail index is lower than the preset threshold, it means that there are significant problems with the current aggregation result. This may be due to semantic deviations or unclear mappings in the encoding, classification, and field definitions of the data uploaded by the subsystems, leading to inconsistent fusion. It may also be due to inaccurate application of aggregation rules or structural conflicts between data. In this case, the main system does not store the current aggregation result but initiates a correction process. It re-semantically maps the original diagnosis and treatment data uploaded by each subsystem, adjusts the field correspondences, unifies the semantic label system, and re-aggregates the data based on the new mapping relationships. The new qualification index is then recalculated and compared with the threshold again. If it still fails to meet the threshold, the process of iterative correction and aggregation continues until the qualification index of the generated aggregated data reaches or exceeds the threshold standard. Through this process, the data stored in the main system is ensured to have high-quality semantic consistency and structural integrity, thereby providing reliable basic data support for subsequent intelligent analysis, assisted diagnosis and treatment, and knowledge extraction.

[0097] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A patient diagnosis and treatment data management system based on big data, characterized in that, The system includes: Semantic ambiguity module: The patient diagnosis and treatment data of each subsystem aggregated by the main system are recorded as aggregated data, and the semantic back-back ambiguity index of the aggregated data is calculated based on the aggregated data; the semantic back-back ambiguity index is used to measure the degree of semantic difference between the diagnosis and treatment data derived from the aggregated data and the diagnosis and treatment data expressed by the original subsystem. Information integrity closed-loop module: Extracts patient diagnosis and treatment event nodes from aggregated data, analyzes the diagnosis and treatment event nodes, and calculates the information integrity closed-loop index. The information integrity closed-loop index is used to measure the connectivity and integrity of each diagnosis and treatment event in the aggregated patient diagnosis and treatment data. Information redundancy module: Extracts fields from each medical record from the aggregated data and calculates the information redundancy index, which is used to measure the degree of redundant data in the aggregated data; Data Management Module: Calculates the qualification index of aggregated data based on semantic back-inference ambiguity index, information redundancy index, and information integrity closed-loop index, and determines the final management of aggregated data by the main system based on the qualification index.

2. The patient diagnosis and treatment data management system based on big data according to claim 1, characterized in that, The semantic ambiguity module includes: Semantic Unit Module: Semantically analyzes each medical record in the aggregated data and divides it into multiple semantic units; Reverse Path Module: Constructs a reverse semantic path tree based on aggregated semantic units; uses all paths from leaf nodes to the root node in the path tree as the set of reverse semantic paths corresponding to all semantic units; Path overlap ratio module: Counts the total number of duplicate paths in the reverse semantic path set. For the total number of all reverse semantic paths in the reverse semantic path set, the total number of duplicate paths is divided by the total number of all reverse semantic paths to obtain the path overlap ratio. First calculation module: Calculates the semantic back-back ambiguity index based on the path overlap ratio.

3. The patient diagnosis and treatment data management system based on big data according to claim 2, characterized in that, The first computing module includes: Root Mean Square Module: For each pair of semantic units, define the semantic edit distance between the corresponding reverse semantic paths, and calculate the root mean square of the semantic edit distance between all pairs of semantic units; Semantic back-calculation ambiguity index module: based on root mean square and path overlap ratio The formula for calculating the semantic back-reasoning ambiguity index is as follows: In the formula, This is the semantic back-reasoning ambiguity index.

4. The patient diagnosis and treatment data management system based on big data according to claim 1, characterized in that, The complete information closed-loop module includes: Set module: This module groups patient treatment events from the aggregated data into a set in chronological order, where 𝑞 represents the total number of patient treatment events in the set; it also constructs a transition representation matrix. If the patient's medical events and If the conditions between them conform to the preset rules, then If it does not conform to the preset rules, then ; Closed-loop access factor module: in all and If there exists a path from Departed and returned If a closed-loop path is found, then the path is recorded. For each closed-loop path pair, ... Calculate the number of event nodes traversed by the closed-loop path, denoted as . Constructing closed-loop access factors : ; Closed-loop event coverage module: Extract the total number of event nodes covered by all closed-loop paths as the total number of covered nodes, and divide the total number of covered nodes by the total number of patient diagnosis and treatment events to obtain the closed-loop event coverage. The second calculation module calculates the information completeness index based on the closed-loop accessibility factor and closed-loop event coverage.

5. The patient diagnosis and treatment data management system based on big data according to claim 4, characterized in that, The second calculation module includes: Compression ratio module: from matrix Extracting submatrices that are only covered by closed-loop paths Count the number of all jump pairs with a value of 1 in the submatrix; then divide the number of all jump pairs with a value of 1 in the submatrix by the maximum number of jumps that a complete closed loop should have under ideal conditions to obtain the local loop compression ratio. Information Completeness Closed-Loop Index Module: The information completeness closed-loop index is obtained by multiplying the mean of the closed-loop accessibility factors of all closed-loop path pairs by the closed-loop event coverage rate and the local loop compression rate.

6. The patient diagnosis and treatment data management system based on big data according to claim 1, characterized in that, The information redundancy module includes: Encoding matrix module: Extracts patient data from aggregated data Medical records, each record contains Each record is treated as a vector of key fields, forming a... The field matrix is ​​obtained by discretely encoding each column of the field and converting it into a pure integer representation. Redundancy Coordination Strength Module: For any two fields in the encoding matrix and Calculate two fields and The redundancy coordination strength between them is used to obtain the field pair redundancy matrix; the calculation formula is: ,in, For fields and Redundancy coordination strength between them This indicates that in all medical record pairs, the field and The number of records with the same value at the same time; This represents the logarithm of all pairwise combinations of medical records; Redundancy Core Set Module: Sorts the redundancy coordination strength of all field pairs by size and extracts the top-ranking redundancy pairs. The set of field pairs forms a redundant core set.

7. The patient diagnosis and treatment data management system based on big data according to claim 6, characterized in that, The information redundancy module also includes: Stable Redundancy Factor Module: For all medical records, if their values ​​on the fields involved in the set are completely identical, then a set of redundant mappings is formed; let the total number of records formed be... There are such mapping volumes, each containing Medical records; calculate the stability redundancy factor: In the formula, As a stable redundancy factor; Information Redundancy Index Module: Calculates the average redundancy coordination strength of all field pairs in the stable redundancy factor. And calculate the information redundancy index, the formula for which is: In the formula, This is the information redundancy index.

8. The patient diagnosis and treatment data management system based on big data according to claim 1, characterized in that, The data management module includes: The qualification index module normalizes the semantic back-derivative ambiguity index, information redundancy index, and information integrity closed-loop index of the aggregated data, removes units, and maps the values ​​to the 0-1 range. Then, it weights and sums the normalized semantic back-derivative ambiguity index, information redundancy index, and information integrity closed-loop index of the aggregated data to obtain the qualification index.

9. The patient diagnosis and treatment data management system based on big data according to claim 8, characterized in that, The data management module also includes: The first management module compares the pass index with the preset threshold. If the pass index is not less than the preset threshold, it means that the semantic mapping of the patient diagnosis and treatment data uploaded by each subsystem by the main system is qualified with the aggregated data. Then the main system stores and manages the current aggregated data.

10. The patient diagnosis and treatment data management system based on big data according to claim 9, characterized in that, The data management module also includes: Second management module: If the pass index is less than the preset threshold, it means that the semantic mapping of the patient diagnosis and treatment data uploaded by each subsystem by the main system and the aggregated data are not qualified. In this case, the main system needs to re-semantically map and aggregate the patient diagnosis and treatment data uploaded by each subsystem until the pass index of the aggregated data is not less than the preset threshold.

Images