Medical information platform driven clinical auxiliary decision support system and method

Abstract

The application discloses a medical information center-driven clinical auxiliary decision support system and method, and relates to the technical field of medical information auxiliary decision making. The system comprises six modules, namely, a clinical data multi-source acquisition module, a center data storage scheduling module, a clinical feature intelligent extraction module, a decision rule dynamic construction module, a multi-dimensional decision reasoning module and a decision result accurate output module, and the modules are cooperatively linked through a standardized data transmission protocol. The method sequentially collects multiple types of clinical data through the connection of heterogeneous medical systems, stores and dynamically schedules the data through a distributed architecture, extracts key clinical features, constructs and updates a decision rule set based on diagnosis and treatment specifications and data features, completes decision operation through multi-path reasoning and cross-validation, and finally presents the decision result in a structured manner according to a clinical scene. The application realizes multi-source data integration and dynamic decision support, improves the comprehensiveness and accuracy of clinical decision making, and provides systematic intelligent technical support for clinical diagnosis and treatment.

Description

Technical Field

[0001] This invention relates to the field of medical information-assisted decision-making technology, and in particular to a clinical auxiliary decision support system and method driven by a medical information platform. Background Technology

[0002] In clinical settings, medical data is scattered across heterogeneous medical systems, with inconsistent data formats and difficulties in sharing, resulting in a lack of comprehensive data support for clinical decision-making. With the advancement of medical informatization, the demand for precise and efficient decision support in clinical practice is increasingly urgent. Traditional decision-making models rely on individual physician experience, making it difficult to quickly integrate multi-source diagnostic and treatment data and synchronize with the latest treatment guidelines. Medical information platforms, with their advantages in data integration and scheduling, have become a key support for breaking down data silos. Based on this, clinical decision support systems and methods can be built, enabling centralized management and intelligent analysis of multi-source data. This provides technical support for disease diagnosis and treatment plan selection, aligning with the modern medical demand for standardized and precise diagnosis and treatment, and solving practical problems such as the difficulty of multi-source data fusion and the untimely updating of decision-making rules.

[0003] Existing technologies have two significant drawbacks: First, data processing lacks a systematic integration mechanism. After the collection of multi-source medical data, an efficient classification, storage, and dynamic scheduling system has not been established, leading to the omission of key information during feature extraction. This makes it difficult to fully mine the core elements in the data that are relevant to clinical decision-making, affecting the integrity of the decision-making basis. Second, the decision-making logic and optimization mechanism are not perfect. Decision rules are mostly statically set and cannot be dynamically updated according to new medical data and revisions to treatment guidelines. Furthermore, there is a lack of multi-dimensional reasoning and cross-validation. At the same time, a scientific adaptability and credibility assessment system has not been established, resulting in insufficient accuracy and applicability of decision-making results, making it difficult to meet the diverse needs of clinical diagnosis and treatment scenarios. Summary of the Invention

[0004] To address the technical problems of existing technologies, such as the lack of systematic data processing integration mechanisms and inadequate decision-making logic and optimization mechanisms, this invention provides a clinical auxiliary decision support system and method driven by a medical information platform. The technical solution is as follows: On the one hand, a clinical auxiliary decision support system driven by a medical information platform is provided, including: a multi-source clinical data acquisition module, a platform-based data storage and scheduling module, a clinical feature intelligent extraction module, a decision rule dynamic construction module, a multi-dimensional decision reasoning module, and a decision result accurate output module. The multi-source clinical data acquisition module acquires data related to electronic medical records, laboratory examinations, imaging reports, medication records, and treatment processes through a heterogeneous medical system interface adaptation mechanism. The platform-based data storage and scheduling module classifies and stores the acquired data based on a distributed architecture and dynamically schedules it according to access requests. The clinical feature intelligent extraction module mines labeling feature information related to clinical decision-making from the scheduled data. The decision rule dynamic construction module generates a dynamically updatable set of decision logic rules based on clinical treatment guidelines and medical data features. The multi-dimensional decision reasoning module calls the decision rule set to perform multi-path reasoning operations on the extracted clinical features. The decision result accurate output module presents the decision information obtained from the reasoning operations in a structured manner according to clinical application scenarios. Each module establishes a data interaction connection sequentially through a standardized data transmission protocol, providing full-process technical support from data acquisition to decision output.

[0005] Furthermore, the intelligent clinical feature extraction module includes: a high-dimensional data feature screening unit, a time-series diagnosis and treatment feature capture unit, a heterogeneous data feature fusion unit, and a calibration feature weight allocation unit. The high-dimensional data feature screening unit filters feature variables with clinical decision-making relevance from massive collected data through a feature correlation analysis mechanism. The time-series diagnosis and treatment feature capture unit performs time-dimensional correlation analysis on continuously generated diagnosis and treatment data to capture dynamically changing clinical features. The heterogeneous data feature fusion unit uses a feature space mapping method to map the feature information corresponding to different types of medical data to a unified feature space and complete the fusion process. The calibration feature weight allocation unit establishes a weight calculation mechanism based on the degree of influence of features on clinical decisions and assigns corresponding weight values ​​to each calibration feature.

[0006] Furthermore, the dynamic decision rule construction module includes: a treatment guideline rule parsing unit, a data feature association rule generation unit, a rule conflict detection and correction unit, and a rule version iteration and update unit. The treatment guideline rule parsing unit performs structured parsing of authoritative clinical treatment guidelines and industry standards to extract core rule elements. The data feature association rule generation unit generates decision rules adapted to medical data features through correlation analysis between features and treatment results. The rule conflict detection and correction unit performs logical conflict verification on the generated decision rules and eliminates conflict items through a rule adjustment mechanism. The rule version iteration and update unit iterates and replaces the existing rule set according to the new medical data and the update of treatment guidelines.

[0007] Furthermore, the multi-dimensional decision reasoning module includes: a rule-based reasoning path selection unit, a probabilistic reasoning operation unit, a case-matching reasoning unit, and a reasoning result cross-validation unit. The rule-based reasoning path selection unit selects an appropriate reasoning path from the rule set based on the clinical feature type and decision requirements. The probabilistic reasoning operation unit quantifies the uncertainty factors in the reasoning process based on a statistical probability model. The case-matching reasoning unit compares the similarity between the current clinical features and the features of historical treatment cases and extracts decision reference information from the matched cases. The reasoning result cross-validation unit verifies the consistency of the results obtained from different reasoning paths to obtain reliable reasoning conclusions.

[0008] Furthermore, the centralized data storage and scheduling module adopts a data scheduling priority calculation model: ,in, This is the data scheduling priority coefficient. The weighting coefficients are satisfied. , This quantifies the urgency of the current data access request. This is an average quantified value of the urgency of historical access requests. The amount of data to be scheduled. This represents the maximum data volume that can be scheduled in a single system session. The waiting time decay coefficient, The duration to wait for data requests.

[0009] Furthermore, the decision rule dynamic construction module employs a rule fit calculation model: ,in, This is the rule fit value. For the number of clinical features, For the number of decision rules, For the first The weight values ​​of each clinical feature For the first The confidence level of the decision rule. For the first The first clinical feature and the first The matching coefficient of each decision rule.

[0010] Furthermore, the multi-dimensional decision reasoning module employs a reasoning credibility calculation model: ,in, To ensure the credibility of the reasoning results, For the number of reasoning paths, For the first The reliability coefficient of each reasoning path. For the first The length of the reasoning path, To verify the sample size, For the first The inference error value of each validation sample.

[0011] Furthermore, the clinical data multi-source acquisition module employs a data acquisition integrity assessment model: ,in, This is the data collection integrity index. To effectively collect a number of data entries, To supplement the number of data entries collected, Theoretically, the total number of data entries should be collected. To supplement the data weighting coefficients, This is the error influence coefficient. To account for the total error in the collected data. This represents the maximum allowable error of the system.

[0012] Furthermore, the precise output module for decision results employs an output information optimization model: ,in, To optimize the output information coefficients, For clinical applicability weighting, To determine the degree of structure of the output information, Weighting based on information accuracy. To ensure the completeness of the output information, To output the response time, For clinical scenarios, the adaptation coefficient is used.

[0013] On the other hand, a clinical decision support method driven by a medical information platform is provided. This method is applied to a clinical decision support system driven by a medical information platform and includes the following steps: S101 connects to heterogeneous medical systems such as hospital information systems, laboratory information systems, medical image archiving and communication systems, and electronic medical record systems through a multi-source clinical data acquisition module. It uses interface adaptation technology to parse the data transmission protocols of different systems and acquire different types of clinical data, including patient basic information, diagnosis and treatment process records, test results, medication information, image data, and pathology reports. S102 transmits the collected multi-type clinical data to the central platform data storage and scheduling module. Based on a storage method that combines a distributed file storage architecture with a relational database, the data is classified and stored according to data type, generation time, and treatment stage. At the same time, a data indexing mechanism is established, and the data is dynamically scheduled and efficiently transmitted through a load balancing algorithm according to the data access requests of subsequent modules. S103, the clinical feature intelligent extraction module receives the scheduled data, uses feature recognition algorithms to mine the labeled features in the data that are related to disease diagnosis, treatment plan selection and prognosis assessment, and processes the feature information through feature screening, time series association and heterogeneous fusion techniques to provide data support for decision reasoning; S104, the dynamic construction module for decision rules is based on authoritative clinical diagnosis and treatment guidelines, industry standards and historical medical data. It uses a rule extraction algorithm to generate an initial set of decision rules, uses rule conflict detection technology to identify logical contradictions, establishes a rule update mechanism, and dynamically adjusts the rule set according to new medical data and revisions to diagnosis and treatment guidelines. S105, the multi-dimensional decision reasoning module calls the completed decision rule set, combines the extracted clinical features, and uses different reasoning methods such as rule reasoning, probabilistic reasoning and case matching reasoning to perform multi-path operations, and uses cross-validation technology of reasoning results to exclude unreliable conclusions; S106, the precise decision result output module, performs structured processing on the verified decision information, presents the decision results using visualization technology according to the clinical doctors' treatment habits and different application scenarios, transmits the results to the clinical workstation through a standardized interface, and establishes a data feedback channel to receive feedback information after clinical application, providing data support for subsequent system optimization.

[0014] The beneficial effects of the technical solution provided by this invention include at least the following: The system breaks down data barriers between heterogeneous medical systems through a multi-source clinical data acquisition module, and combines the classification storage and dynamic scheduling capabilities of a platform-based data storage and scheduling module to achieve comprehensive aggregation and efficient flow of multiple types of data, solving the problems of scattered data and inefficient access in traditional technologies; Through multi-unit collaborative processing of the intelligent clinical feature extraction module, core features are accurately mined and weights are reasonably allocated, making up for the shortcomings of incomplete feature extraction and omission of key information in existing technologies. The dynamic decision rule construction module supports rule parsing and generation, conflict correction, and iterative updates; The multi-dimensional decision reasoning module integrates multiple reasoning methods and performs cross-validation, and with the dedicated quantitative evaluation mechanism of each module, the adaptability of decision rules and the reliability of reasoning results are significantly improved, solving the problems of static solidification and insufficient accuracy of traditional decision rules. The structured presentation and scenario-based adaptation of the precise decision result output module, combined with standardized data interaction throughout the entire process, ensures the efficient implementation of decision information. Overall, it realizes full-chain optimization of multi-source data integration, dynamic rule construction, and precise reasoning output, providing comprehensive and reliable technical support for clinical decision-making. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A diagram illustrating the module composition of a clinical auxiliary decision support system driven by a medical information platform, as provided in an embodiment of the present invention. Figure 2 A flowchart illustrating the steps of a clinical auxiliary decision support method driven by a medical information platform, as provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] like Figure 1 As shown, the clinical decision support system driven by the medical information platform includes: a multi-source clinical data acquisition module, a platform-based data storage and scheduling module, a clinical feature intelligent extraction module, a decision rule dynamic construction module, a multi-dimensional decision reasoning module, and a decision result accurate output module. The multi-source clinical data acquisition module acquires data related to electronic medical records, laboratory tests, imaging reports, medication records, and treatment processes through a heterogeneous medical system interface adaptation mechanism. The platform-based data storage and scheduling module classifies and stores the acquired data based on a distributed architecture and dynamically schedules it according to access requests. The intelligent clinical feature extraction module mines labeling feature information related to clinical decision-making from the scheduled data. The dynamic decision rule construction module generates a dynamically updatable set of decision logic rules based on clinical treatment guidelines and medical data features. The multi-dimensional decision reasoning module calls the decision rule set to perform multi-path reasoning operations on the extracted clinical features. The accurate decision result output module presents the decision information obtained from the reasoning operations in a structured manner according to clinical application scenarios. Each module establishes a data interaction connection sequentially through a standardized data transmission protocol, providing full-process technical support from data acquisition to decision output.

[0023] The multi-source clinical data acquisition module serves as the core of the system's data input. Through a pre-defined heterogeneous medical system interface adaptation mechanism, it achieves stable integration with at least 10 mainstream medical systems, including hospital information systems, laboratory information systems, medical image archiving and communication systems, and electronic medical record systems. The interface adaptation supports more than 8 industry-standard data transmission protocols, such as HL7, DICOM, and FHIR. A protocol parser performs real-time parsing of the data packet structures from different systems, ensuring consistent data format conversion. This module features a dynamic data acquisition frequency adjustment mechanism, flexibly configurable between 1 second / acquisition and 60 seconds / acquisition according to clinical needs, automatically switching to the highest acquisition frequency for emergency treatment scenarios. The types of data collected include text records in electronic medical records, numerical indicators in laboratory tests, image metadata in imaging reports, drug codes and dosage information in medication records, and operation time nodes in the diagnosis and treatment process. During the collection process, a data integrity verification mechanism is used to detect missing fields in each data entry. When the missing rate exceeds 5%, a re-collection instruction is triggered, with a maximum of 3 re-collections and an interval of 5 seconds between each re-collection. At the same time, data security is ensured through data encryption transmission technology. The encryption algorithm adopts the AES-256 standard, and the data transmission latency is controlled within 100 milliseconds, ensuring efficient, complete, and secure aggregation of multi-source data and providing comprehensive raw data support for subsequent modules.

[0024] The platform-based data storage and scheduling module is built with a distributed architecture, including at least 16 storage nodes, each with a storage capacity of no less than 10TB. It supports data redundancy backup, with three backup copies stored on nodes in different physical locations to ensure data storage security. This module employs a hybrid storage approach combining a distributed file storage architecture and a relational database. Unstructured data, such as image files and text reports, is stored in the distributed file system, while structured data, such as test results and medication dosages, is stored in the relational database. Data is categorized into 6 main categories and 24 subcategories based on data type. Furthermore, data is partitioned by day based on its generation time. A three-level indexing mechanism is established, tagged with data attributes based on the treatment stage: the first-level index is categorized by data type, the second-level index by generation time, and the third-level index by treatment stage. The indexes are updated in real-time. When receiving data access requests from subsequent modules, the module allocates storage node resources through a load balancing algorithm. The load balancing threshold is set to 80% CPU utilization and 75% memory utilization for a single node. When the node load exceeds the threshold, the request is automatically distributed to a node with a lower load. The data scheduling response time is controlled within 50 milliseconds, achieving efficient data scheduling and transmission, and providing stable data access support for subsequent processes such as feature extraction and decision reasoning.

[0025] After receiving the integrated data from the centralized data storage and scheduling module, the clinical feature intelligent extraction module first preprocesses the data through a high-dimensional data feature screening unit. The screening window size is set to 1000 data points per batch. A feature correlation analysis mechanism is used to calculate the correlation between each data field and clinical decision-making. The correlation threshold is set to 0.6, and feature variables with correlation higher than the threshold are retained. After preliminary screening, the feature dimension is controlled within 500. Subsequently, the time-series diagnosis and treatment feature capture unit performs time-dimensional correlation analysis on the continuously generated diagnosis and treatment data. The time window is set to 24 hours, and time segments are divided by hours. The dynamic change features of the data within each time segment are captured to generate a time-series feature sequence. The heterogeneous data feature fusion unit uses a feature space mapping method to map the feature information of different types of medical data to a unified 128-dimensional feature space. A weighted fusion strategy is used during the fusion process, and the fusion weights are pre-set according to the importance of the data types. The feature weight allocation unit establishes a weight calculation mechanism based on the degree of influence of features on clinical decision-making. The weight of each feature is determined by the analytic hierarchy process, with the weight value ranging from 0 to 1 and the total weight being 1. Among them, the weight of core features such as core test indicators and key symptom descriptions is not less than 0.1. The number of final extracted features is controlled within 100, and the feature extraction accuracy is controlled above 95%, providing accurate feature support for decision rule construction and inference calculation.

[0026] The dynamic decision rule construction module is based on authoritative clinical treatment guidelines, industry standards, and historical medical data. The included treatment guidelines cover at least 50 specialties and over 300 diseases, with a historical medical data sample size of no less than 1 million records and a data span of no less than 5 years. This module uses a treatment guideline rule parsing unit to perform structured parsing of the guidelines and industry standards, refining the parsing granularity to the clause level, extracting core rule elements such as disease diagnosis conditions, scope of application of treatment plans, and contraindications for medication, forming an initial rule element library. The data feature association rule generation unit generates decision rules adapted to the characteristics of medical data through association analysis between features and treatment results, using association rule mining algorithms. The minimum support is set to 0.05, and the minimum confidence is set to 0.8 to ensure the effectiveness of the generated rules. The rule conflict detection and correction unit uses a logical consistency verification algorithm to verify each generated decision rule at a rate of 1000 rules per minute. When logical contradictions are detected between rules, a rule priority ranking mechanism is used for correction. Priorities are determined by the authority of the treatment guidelines and the support of the data, with higher-priority rules overriding lower-priority rules. The rule version iteration and update unit has a set update cycle, with a regular update cycle of one month. Emergency updates are triggered when there are major revisions to treatment guidelines or when the amount of new data exceeds 100,000. During the update process, an incremental update method is used, replacing only conflicting and newly added rules while retaining valid rules to ensure the dynamic optimization and timeliness of the decision rule set.

[0027] The multi-dimensional decision-making reasoning module calls upon the decision rule set generated by the dynamic decision rule construction module and combines it with the calibrated features extracted by the clinical feature intelligent extraction module. It employs a multi-path reasoning operation mode, supporting at least eight parallel reasoning paths, each corresponding to a different reasoning method. The rule reasoning path selection unit automatically selects the appropriate reasoning path based on the clinical feature type and decision requirements, with a path selection response time controlled within 20 milliseconds. The probabilistic reasoning operation unit quantifies the uncertainties in the reasoning process based on a statistical probability model, setting the confidence interval to 95% and quantifying the errors in the reasoning process. The case matching reasoning unit compares the current clinical features with the features of historical treatment cases. The case library includes at least 500,000 valid historical cases. The similarity calculation uses a cosine similarity algorithm, with a similarity threshold set to 0.85. Cases with similarity higher than the threshold are extracted as references, with no more than 20 reference cases per reasoning process. The cross-validation unit for inference results performs consistency checks on the results obtained from different inference paths. The validation indicators include result consistency and logical consistency. When the result consistency is lower than 90%, the inference process is restarted and the inference parameters are adjusted until a consistent inference conclusion is obtained. The entire inference process takes less than 3 seconds to ensure the reliability and efficiency of the inference results and provide accurate inference support for decision output.

[0028] After receiving reliable reasoning conclusions from the multi-dimensional decision reasoning module, the precise decision output module first structures the decision information, dividing it into six output templates according to clinical application scenarios, including disease diagnosis reference, treatment plan recommendation, and risk warning prompts. Each template includes 20 fixed core fields to ensure the standardization of the output information. This module uses visualization technology to present the decision results, supporting more than four display formats such as tables, line graphs, and bar charts. Clinicians can switch between display formats according to their operating habits, and the visualization interface response time is controlled within 300 milliseconds. The module transmits the structured decision results to the clinical workstation through a standardized interface. The interface transmission protocol adopts the HL7FHIR standard, with a transmission rate of no less than 10Mbps, ensuring the rapid application of results in clinical scenarios. Simultaneously, a data feedback channel is established to receive feedback information from clinicians after applying the decision results. The feedback information includes eight core fields such as result accuracy evaluation and supplementary needs suggestions. The feedback information collection cycle is 7 days, and the number of feedback collections for each decision result is no less than 3. The feedback data is stored in a dedicated database and is statistically analyzed regularly. The analysis cycle is 1 month, which provides direct data support for subsequent module optimization, rule updates, and feature adjustments of the system, forming a closed-loop mechanism of "output-feedback-optimization".

[0029] Preferably, the intelligent clinical feature extraction module includes: a high-dimensional data feature screening unit, a time-series diagnosis and treatment feature capture unit, a heterogeneous data feature fusion unit, and a calibrated feature weight allocation unit; The high-dimensional data feature screening unit filters out feature variables with clinical decision-making relevance from massive collected data through a feature correlation analysis mechanism. The time-series diagnosis and treatment feature capture unit performs time-dimensional correlation analysis on continuously generated diagnosis and treatment data to capture dynamically changing clinical features. The heterogeneous data feature fusion unit uses a feature space mapping method to map the feature information corresponding to different types of medical data to a unified feature space and complete the fusion process. The calibration feature weight allocation unit establishes a weight calculation mechanism based on the degree of influence of features on clinical decision-making and assigns corresponding weight values ​​to each calibration feature.

[0030] Specifically, the intelligent clinical feature extraction module comprises a high-dimensional data feature screening unit, a time-series diagnostic and treatment feature capture unit, a heterogeneous data feature fusion unit, and a calibrated feature weight allocation unit that work collaboratively. The high-dimensional data feature screening unit uses a sliding window mechanism to process input data, with a fixed window size of 1000 data points per batch. It calculates the correlation between each data field and the clinical decision-making objective using the Pearson correlation coefficient, setting a correlation threshold of 0.6. Only feature variables with correlation exceeding the threshold are retained, compressing the original data dimension from thousands to less than 500 dimensions. The screening process takes no more than 2 seconds per batch. The time-series diagnostic and treatment feature capture unit uses a 24-hour time period, dividing it into 36 time segments of 60 minutes each. It performs trend analysis and changes calculation on the diagnostic and treatment data within each segment, capturing dynamic features such as symptom exacerbation rate and indicator fluctuation amplitude, generating a time-series feature sequence including timestamps. The length of the feature sequence is consistent with the number of time segments. The heterogeneous data feature fusion unit employs feature space mapping technology to uniformly map different types of features, such as textual, numerical, and image metadata, to a 128-dimensional Euclidean feature space. During the fusion process, preset weights are assigned to each type of data: image features account for 35%, numerical test features for 30%, textual descriptive features for 25%, and other auxiliary features for 10%. The fusion operation is completed through weighted summation. The feature weight allocation unit uses the analytic hierarchy process (AHP) to construct a three-level evaluation system. The first-level indicators include diagnostic relevance, treatment guidance value, and prognostic predictive significance. The second-level indicators are further subdivided into 12 specific evaluation dimensions. Feature weights are calculated through pairwise comparison matrices, with weight values ​​ranging from 0 to 1 and a total sum of 1. Core clinical indicators, such as key parameters of complete blood counts and core symptom descriptions, are forcibly set with a weight of no less than 0.1. The final output is a weighted calibration feature set of no more than 100 features, with the overall feature extraction accuracy consistently above 95%.

[0031] Preferably, the decision rule dynamic construction module includes: a diagnosis and treatment standard rule parsing unit, a data feature association rule generation unit, a rule conflict detection and correction unit, and a rule version iteration and update unit; The diagnostic and treatment guidelines rule parsing unit performs structured parsing of authoritative clinical diagnostic and treatment guidelines and industry standards to extract core rule elements. The data feature association rule generation unit generates decision rules adapted to medical data features through the association analysis of features and diagnostic and treatment results. The rule conflict detection and correction unit performs logical conflict verification on the generated decision rules and eliminates conflict items through the rule adjustment mechanism. The rule version iteration and update unit iterates and replaces the existing rule set according to the new medical data and the update of diagnostic and treatment guidelines.

[0032] Specifically, the dynamic construction module for decision-making rules is seamlessly integrated with its diagnostic and treatment guidelines rule parsing unit, data feature association rule generation unit, rule conflict detection and correction unit, and rule version iteration and update unit. The diagnostic and treatment guidelines rule parsing unit uses XML structured format to parse over 300 disease treatment guidelines from 50 specialties, refining the parsing granularity to the clause level. It extracts core rule elements such as disease diagnosis thresholds, applicable conditions for treatment plans, drug dosage ranges, and contraindication lists. Each element is accompanied by an authoritative source marker and evidence level score, forming an initial library of over 2000 rule elements with a parsing efficiency of 50 pages per minute. The data feature association rule generation unit uses the Apriori algorithm to mine the correlation between features and treatment results, setting a minimum support of 0.05 and a minimum confidence of 0.8. Only strongly correlated rules that meet the threshold requirements are retained. The rule expression adopts an "IF-THEN" logical structure, clearly defining the preconditions, triggering conditions, and output conclusions. Rule mining for a single batch of 1 million historical data points takes no more than 30 minutes. The rule conflict detection and correction unit uses a predicate logic reasoning algorithm to verify each rule in the generated rule set at a speed of 1000 rules per minute. It identifies contradictory, duplicate, and redundant rules through logical consistency checks and corrects conflicts based on a priority mechanism: the authority of the treatment guidelines (national > provincial > institutional) and data support (sample size > 10,000 > 5,000 > 1,000). Higher-priority rules automatically override lower-priority conflicting rules. The rule version iteration and update unit employs a dual update mechanism. The regular update cycle is fixed at one month. When the number of revised national treatment guidelines or newly added valid medical data exceeds 100,000, an emergency update is triggered. This uses incremental update technology to replace only conflicting and outdated rules, supplementing newly added related rules. The update process does not interrupt system operation. After the updated rule set is verified through 1000 test cases, it officially takes effect, ensuring the timeliness and applicability of the rule set.

[0033] Preferably, the multi-dimensional decision reasoning module includes: a rule reasoning path selection unit, a probability reasoning operation unit, a case matching reasoning unit, and a reasoning result cross-validation unit; The rule reasoning path selection unit selects a suitable reasoning path from the rule set based on the clinical feature type and decision-making needs. The probability reasoning operation unit quantifies the uncertainty factors in the reasoning process based on the statistical probability model. The case matching reasoning unit compares the similarity between the current clinical features and the features of historical treatment cases and extracts the decision reference information of the matching cases. The reasoning result cross-validation unit performs consistency verification on the results obtained from different reasoning paths to obtain reliable reasoning conclusions.

[0034] Specifically, the multi-dimensional decision-making reasoning module's rule-based reasoning path selection unit, probabilistic reasoning operation unit, case matching reasoning unit, and reasoning result cross-validation unit work together to achieve accurate reasoning. The rule-based reasoning path selection unit has a built-in path matching algorithm that automatically matches an appropriate reasoning path based on the type of input clinical features (such as symptom-based, indicator-based, and imaging-based) and decision requirements (diagnosis, treatment, and prognosis). It supports simultaneous operation of up to eight parallel reasoning paths, with a path selection response time controlled within 20 milliseconds to ensure rapid startup of the reasoning process. The probabilistic reasoning operation unit quantifies the uncertainties in the reasoning process based on a Bayesian probability model, sets a 95% confidence interval as the criterion for judging the reliability of the results, and quantitatively evaluates the confidence level of the final reasoning result by calculating the error value and uncertainty coefficient of each reasoning step. Results with a confidence level below 80% are marked as pending verification. The case matching and reasoning unit accesses a case database containing 500,000 valid historical cases. This database is divided into over 300 sub-databases based on disease type. A cosine similarity algorithm is used to calculate the similarity between current clinical features and historical case features, with a similarity threshold set at 0.85. Each reasoning process extracts decision reference information from a maximum of 20 highly similar cases, with case retrieval and matching taking no more than 1 second. The reasoning result cross-validation unit employs dual validation criteria: a minimum 90% consistency rate and complete adherence to clinical diagnostic logic. Results from different reasoning paths are compared pairwise. If the consistency falls below the threshold, reasoning parameters (such as feature weights and rule confidence thresholds) are automatically adjusted, and the reasoning process is restarted until all paths reach consensus. The entire reasoning process is strictly controlled to within 3 seconds, ensuring efficient and reliable decision support for clinical practice.

[0035] Preferably, the centralized data storage and scheduling module adopts a data scheduling priority calculation model: ,in, This is the data scheduling priority coefficient. The weighting coefficients are satisfied. , This quantifies the urgency of the current data access request. Here, represents the average quantified value of the urgency of historical access requests, and represents the amount of data to be scheduled. This represents the maximum data volume that can be scheduled in a single system session. The waiting time decay coefficient, The duration to wait for data requests.

[0036] Specifically, the data scheduling priority calculation mechanism in the platform-based data storage scheduling module comprehensively evaluates the scheduling priority of data access requests through multi-dimensional parameters. The weighting coefficients are set to a fixed proportion: 60% for urgency and 40% for data volume, with a sum of 1 to ensure a balanced evaluation. The urgency level of data access requests is quantified into 10 levels based on clinical scenarios, increasing from 1 to 10. Emergency treatment scenarios such as emergency resuscitation and intensive care have a quantification value of 8-10, routine treatment scenarios have a value of 3-7, and non-treatment data queries have a value of 1-2. The average quantification value of historical access request urgency is derived from access data statistics over the past three months, and is updated daily. The data volume to be scheduled is divided into five ranges based on storage size, from less than 1MB to more than 10GB. The maximum data volume for a single system scheduling operation is set to 20GB to meet the scheduling needs of special data types such as large-scale imaging data. The waiting time decay coefficient is set to 0.05 to mitigate the priority decay of long-waiting requests, preventing important requests from being excessively downgraded due to excessive waiting time. The data request waiting time is counted from the moment the request is submitted, accurate to the second. This mechanism outputs a scheduling priority coefficient of 0-10 through dynamic input and calculation of the above parameters. Requests with higher coefficients are given priority in allocating storage node resources. The scheduling response time is controlled within 50 milliseconds, ensuring that urgent and important data requests receive rapid responses and improving the rationality and efficiency of overall data scheduling.

[0037] Preferably, the decision rule dynamic construction module employs a rule fit calculation model: ,in, This is the rule fit value. For the number of clinical features, For the number of decision rules, For the first The weight values ​​of each clinical feature For the first The confidence level of the decision rule. For the first The first clinical feature and the first The matching coefficient of each decision rule.

[0038] Specifically, the rule adaptation calculation mechanism in the dynamic construction module of decision rules takes the matching degree between clinical features and decision rules as the core evaluation target. The number of clinical features is dynamically adjusted according to the actual diagnosis and treatment scenario, ranging from 50 to 200. The number of decision rules corresponds to the included diagnosis and treatment guidelines, including more than 300 diseases in 50 specialties, with the total number of rules maintained at over 2000. The weight value of each clinical feature is determined by the analytic hierarchy process (AHP). The weight values ​​of core clinical indicators, such as key test values ​​and core symptom descriptions, are set at 0.1-0.3, while the weight values ​​of secondary features are set at 0.01-0.09, ensuring that core features play a dominant role in the adaptation assessment. The confidence level of decision rules is set based on the results of historical data validation. Through validation using more than 1 million historical diagnosis and treatment data, the rule confidence level is divided into five levels from 0.6 to 1.0. The confidence level of rules that have undergone large-scale data validation and comply with national diagnosis and treatment guidelines is set at 0.9-1.0, while the confidence level of newly added rules that have only undergone small-scale data validation is set at 0.6-0.7. The matching coefficient between clinical features and decision rules is divided into four levels: 0, 0.3, 0.6, and 1.0, based on the degree of matching. A perfect match is 1.0, a high match is 0.6, a partial match is 0.3, and no match is 0. The matching coefficient is automatically determined by comparing the features with the rule's prerequisite fields. This calculation mechanism outputs a rule fit value of 0-1 through the integration of the above parameters. Rules with a fit value higher than 0.8 are directly included in the decision rule set. Rules with a fit value between 0.6 and 0.8 require secondary verification to determine their inclusion. Rules with a fit value lower than 0.6 are marked as needing optimization, ensuring a high degree of fit between the decision rule set and the clinical features.

[0039] Preferably, the multi-dimensional decision reasoning module employs a reasoning credibility calculation model: ,in, To ensure the credibility of the reasoning results, For the number of reasoning paths, For the first The reliability coefficient of each reasoning path. For the first The length of the reasoning path, To verify the sample size, For the first The inference error value of each validation sample.

[0040] Specifically, the reasoning credibility calculation mechanism in the multi-dimensional decision-making reasoning module evaluates the reliability of the reasoning results through a dual dimension of reasoning path and validation samples. The number of reasoning paths is dynamically adjusted according to decision-making needs, with 3-5 paths for routine diagnosis and treatment scenarios and 6-8 paths for complex disease diagnosis and treatment scenarios. The reliability coefficient of each reasoning path is set based on the maturity of the corresponding reasoning algorithm. The reliability coefficient for rule-based reasoning paths with higher maturity is set at 0.9-1.0, while the reliability coefficient for auxiliary paths such as case matching reasoning is set at 0.7-0.8. The length of the reasoning path is calculated based on the number of reasoning steps, with the length of a routine path controlled at 5-8 steps and a maximum of 10 steps to avoid the accumulation of reasoning errors due to excessively long paths. The number of validation samples is divided according to disease type, with 1000-2000 validation samples for common and frequently occurring diseases and 300-500 validation samples for rare diseases. All validation samples come from historical diagnosis and treatment data that have been clinically confirmed to ensure the validity of the samples. The inference error value is calculated based on the degree of deviation between the inference result and the clinical diagnosis result. The deviation range is between 0 and 0.5. When they are completely consistent, the error value is 0, and when the deviation is large, the error value does not exceed 0.5. The calculation mechanism outputs a confidence level of 0-1 for the inference result through comprehensive calculation of the above parameters. Results with a confidence level higher than 0.85 are directly used as the decision output. Results with a confidence level between 0.7 and 0.85 need to be further judged in conjunction with the clinical physician's experience. Results with a confidence level lower than 0.7 restart the inference process and improve the confidence level by adjusting the inference parameters to ensure the reliability of the inference result.

[0041] Preferably, the clinical data multi-source acquisition module employs a data acquisition integrity assessment model: ,in, This is the data collection integrity index. To effectively collect a number of data entries, To supplement the number of data entries collected, Theoretically, the total number of data entries should be collected. To supplement the data weighting coefficients, This is the error influence coefficient. To account for the total error in the collected data. This represents the maximum allowable error of the system.

[0042] Specifically, the data acquisition integrity assessment mechanism in the multi-source clinical data acquisition module comprehensively evaluates acquisition quality through three dimensions: effective data volume, supplementary data volume, and error control. Effective data collection refers to the original data that meets data format requirements and has no missing fields. Supplementary data collection refers to the effective data obtained through supplementary collection commands after initial missing data. Supplementary collection is limited to a maximum of three times, with a 5-second interval between each collection to ensure the acquisition of as complete data as possible. The theoretically required total number of data entries is pre-set based on the clinical diagnosis and treatment process, including six major categories and 24 subcategories of data such as patient basic information, medical records, and laboratory test results. The number of required fields for each subcategory is set to 5-15 to ensure comprehensive data acquisition. The supplementary data weighting coefficient is set to 0.5, acknowledging the value of supplementary data while avoiding distortion of the integrity assessment due to over-reliance on supplementary data. The error impact coefficient is set to 0.3 to control the impact of data acquisition errors on the integrity assessment. The total data acquisition error is calculated cumulatively based on error types such as field errors and format mismatches. The system allows a maximum error of 5% of the total data volume; exceeding this threshold triggers a data verification alarm. The calculation mechanism outputs a data acquisition integrity index of 0-1 based on the above parameters. Data with an index higher than 0.9 directly enters the subsequent storage process. Data with an index between 0.8 and 0.9 needs to be partially acquired. Data with an index lower than 0.8 restarts the acquisition process to ensure that the data entering the system has high integrity.

[0043] Preferably, the decision result accurate output module employs an output information optimization model: ,in, To optimize the output information coefficients, For clinical applicability weighting, To determine the degree of structure of the output information, Weighting based on information accuracy. To ensure the completeness of the output information, To output the response time, For clinical scenarios, the adaptation coefficient is used.

[0044] Specifically, the output information optimization mechanism in the precise decision-making output module optimizes output effectiveness through five dimensions: clinical applicability, information structuring degree, accuracy, response time, and scenario adaptability. The weights for clinical applicability and information accuracy are both set to 0.3, serving as core evaluation indicators to ensure that the output information meets actual clinical application needs. The structuring degree of the output information is evaluated based on the field matching degree of a preset template. The structured template includes 20 core fields; a structuring degree value of 1.0 is achieved when the matching degree reaches 100%. The structuring degree value decreases by 0.05 for each missing non-critical field and by 0.1 for each missing critical field, ensuring the standardization of the output information. The completeness of the output information is evaluated based on the coverage of the core information required for decision-making. A completeness value of 1.0 is achieved when all 10 or more core information items are covered, 0.8 for 8-9 items, 0.6 for 6-7 items, and 0.4 for less than 6 items, ensuring no key information is omitted. Output response time begins from the time the inference result is transmitted to the output module. Response time for standard scenarios is controlled within 300 milliseconds, and for complex visualization scenarios, it does not exceed 500 milliseconds, avoiding impact on clinical efficiency due to slow response. The clinical scenario adaptation coefficient is set according to the degree of matching between the output format and the scenario: 0.9 for diagnostic reference scenarios, 0.85 for treatment plan recommendation scenarios, and 0.8 for risk warning scenarios, ensuring that the output format aligns with scenario requirements. This calculation mechanism integrates the above parameters to output an optimization coefficient of 0-2. Output information with coefficients higher than 1.5 is presented directly; coefficients between 1.2 and 1.5 undergo slight adjustments; and coefficients lower than 1.2 result in re-optimization of the output format and content, ensuring efficient implementation of decision-making results.

[0045] like Figure 2 As shown, a clinical decision support method driven by a medical information platform is described. This method is applied to a clinical decision support system driven by a medical information platform and includes the following steps: S101 connects to heterogeneous medical systems such as hospital information systems, laboratory information systems, medical image archiving and communication systems, and electronic medical record systems through a multi-source clinical data acquisition module. It uses interface adaptation technology to parse the data transmission protocols of different systems and acquire different types of clinical data, including patient basic information, diagnosis and treatment process records, test results, medication information, image data, and pathology reports. S102 transmits the collected multi-type clinical data to the central platform data storage and scheduling module. Based on a storage method that combines a distributed file storage architecture with a relational database, the data is classified and stored according to data type, generation time, and treatment stage. At the same time, a data indexing mechanism is established, and the data is dynamically scheduled and efficiently transmitted through a load balancing algorithm according to the data access requests of subsequent modules. S103, the clinical feature intelligent extraction module receives the scheduled data, uses feature recognition algorithms to mine the labeled features in the data that are related to disease diagnosis, treatment plan selection and prognosis assessment, and processes the feature information through feature screening, time series association and heterogeneous fusion techniques to provide data support for decision reasoning; S104, the dynamic construction module for decision rules is based on authoritative clinical diagnosis and treatment guidelines, industry standards and historical medical data. It uses a rule extraction algorithm to generate an initial set of decision rules, uses rule conflict detection technology to identify logical contradictions, establishes a rule update mechanism, and dynamically adjusts the rule set according to new medical data and revisions to diagnosis and treatment guidelines. S105, the multi-dimensional decision reasoning module calls the completed decision rule set, combines the extracted clinical features, and uses different reasoning methods such as rule reasoning, probabilistic reasoning and case matching reasoning to perform multi-path operations, and uses cross-validation technology of reasoning results to exclude unreliable conclusions; S106, the precise decision output module, performs structured processing on the verified decision information. According to the clinical doctors' diagnosis and treatment habits and different application scenarios, including disease diagnosis reference, treatment plan recommendation, risk warning prompts, etc., the decision results are presented using visualization technology. The results are transmitted to the clinical workstation through a standardized interface. At the same time, a data feedback channel is established to receive feedback information after clinical application, providing data support for subsequent system optimization.

[0046] The clinical decision support system and methodology driven by a medical information platform leverages modular collaborative design and end-to-end technological optimization to achieve multi-dimensional technological advantages and precisely overcome the shortcomings of previous technologies. The system breaks down data barriers between heterogeneous medical systems through a multi-source clinical data acquisition module. An interface adaptation mechanism enables comprehensive aggregation of various types of clinical data. Combined with the distributed classification storage and dynamic scheduling capabilities of the platform-based data storage and scheduling module, it solves the problems of scattered data and inefficient access in traditional technologies, ensuring high efficiency in data transmission and access. The intelligent clinical feature extraction module, through multi-unit collaborative operation, accurately filters highly correlated features, captures temporal dynamic features, integrates heterogeneous data features, and rationally allocates weights. This compensates for the incomplete feature extraction and omission of key information in traditional technologies, providing high-quality data support for decision-making.

[0047] The dynamic decision rule construction module achieves dynamic optimization of decision rules through parsing treatment guidelines, generating association rules, detecting and correcting conflicts, and iteratively updating versions, breaking the limitations of traditional static rules. The multi-dimensional decision reasoning module integrates multiple reasoning methods and performs cross-validation, coupled with the quantitative evaluation mechanisms of each core module, significantly improving the reliability and adaptability of decision results and solving the problem of insufficient accuracy in traditional decision-making. The structured presentation and scenario-based adaptation of the precise decision result output module, combined with standardized data interaction throughout the entire process, ensures the efficient implementation of decision information in clinical scenarios. The overall technical system achieves full-chain optimization from data integration, feature extraction, rule construction, reasoning computation to result output, comprehensively overcoming the shortcomings of existing technologies and providing systematic and intelligent technical support for clinical decision-making.

[0048] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0049] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0050] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0051] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0052] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0053] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0054] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0055] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0056] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A clinical auxiliary decision support system driven by a medical information platform, characterized in that, The system includes: a multi-source clinical data acquisition module, a platform-based data storage and scheduling module, a clinical feature intelligent extraction module, a decision rule dynamic construction module, a multi-dimensional decision reasoning module, and a decision result accurate output module; The multi-source clinical data acquisition module acquires relevant data from electronic medical records, laboratory tests, imaging reports, medication records, and treatment processes through a heterogeneous medical system interface adaptation mechanism. The platform-based data storage and scheduling module classifies and stores the acquired data based on a distributed architecture and dynamically schedules it according to access requests. The intelligent clinical feature extraction module mines labeling feature information related to clinical decision-making from the dynamically scheduled data. The dynamic decision rule construction module generates a dynamically updatable set of decision logic rules based on clinical treatment guidelines and medical data features. The multi-dimensional decision reasoning module calls the set of decision logic rules to perform multi-path reasoning operations on the extracted clinical features. The accurate decision result output module presents the decision information obtained from the reasoning operations in a structured manner according to clinical application scenarios. Each module establishes a data interaction connection sequentially through a standardized data transmission protocol, providing full-process technical support from data acquisition to decision output.

2. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The intelligent clinical feature extraction module includes: a high-dimensional data feature screening unit, a time-series diagnosis and treatment feature capture unit, a heterogeneous data feature fusion unit, and a calibrated feature weight allocation unit; The high-dimensional data feature screening unit filters out feature variables with clinical decision-making relevance from massive collected data through a feature correlation analysis mechanism. The time-series diagnosis and treatment feature capture unit performs time-dimensional correlation analysis on continuously generated diagnosis and treatment data to capture dynamically changing clinical features. The heterogeneous data feature fusion unit uses a feature space mapping method to map the feature information corresponding to different types of medical data to a unified feature space and complete the fusion process. The calibration feature weight allocation unit establishes a weight calculation mechanism based on the degree of influence of features on clinical decision-making and assigns corresponding weight values ​​to each calibration feature.

3. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The decision rule dynamic construction module includes: a diagnosis and treatment guideline rule parsing unit, a data feature association rule generation unit, a rule conflict detection and correction unit, and a rule version iteration and update unit; The diagnostic and treatment guidelines rule parsing unit performs structured parsing of authoritative clinical diagnostic and treatment guidelines and industry standards to extract core rule elements. The data feature association rule generation unit generates decision rules adapted to medical data features through the association analysis of features and diagnostic and treatment results. The rule conflict detection and correction unit performs logical conflict verification on the generated decision rules and eliminates conflict items through the rule adjustment mechanism. The rule version iteration and update unit iterates and replaces the existing decision logic rule set according to the new medical data and the update of diagnostic and treatment guidelines.

4. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The multi-dimensional decision reasoning module includes: a rule reasoning path selection unit, a probability reasoning operation unit, a case matching reasoning unit, and a reasoning result cross-validation unit; The rule reasoning path selection unit selects a suitable reasoning path from the decision logic rule set according to the clinical feature type and decision needs. The probability reasoning operation unit quantifies the uncertainty factors in the reasoning process based on the statistical probability model. The case matching reasoning unit compares the similarity between the current clinical features and the features of historical treatment cases and extracts the decision reference information of the matching cases. The reasoning result cross-validation unit performs consistency verification on the results obtained from different reasoning paths to obtain reliable reasoning conclusions.

5. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The data scheduling priority calculation model used in the platform-based data storage scheduling module is as follows: ,in, This is the data scheduling priority coefficient. The weighting coefficients are satisfied. , This quantifies the urgency of the current data access request. This is an average quantified value of the urgency of historical access requests. The amount of data to be scheduled. This represents the maximum data volume that can be scheduled in a single system session. The waiting time decay coefficient, The duration to wait for data requests.

6. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The decision rule dynamic construction module uses a rule fit calculation model: ,in, This is the rule fit value. For the number of clinical features, For the number of decision rules, For the first The weight values ​​of each clinical feature For the first The confidence level of the decision rule. For the first The first clinical feature and the first The matching coefficient of each decision rule.

7. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The multi-dimensional decision reasoning module employs a reasoning credibility calculation model: ,in, To ensure the credibility of the reasoning results, For the number of reasoning paths, For the first The reliability coefficient of each reasoning path. For the first The length of the reasoning path, To verify the sample size, For the first The inference error value of each validation sample.

8. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The clinical data multi-source acquisition module employs a data acquisition integrity assessment model: ,in, This is the data collection integrity index. To effectively collect a number of data entries, To supplement the number of data entries collected, Theoretically, the total number of data entries should be collected. To supplement the data weighting coefficients, This is the error influence coefficient. To account for the total error in the collected data. This represents the maximum allowable error of the system.

9. The clinical auxiliary decision support system driven by a medical information platform according to claim 1, characterized in that, The precise output module for decision results employs an output information optimization model. ,in, To optimize the output information coefficients, For clinical applicability weighting, To determine the degree of structure of the output information, Weighting based on information accuracy. To ensure the completeness of the output information, To output the response time, For clinical scenarios, the adaptation coefficient is used.

10. A clinical decision support method driven by a medical information platform, characterized in that, This method is implemented by a clinical auxiliary decision support system driven by a medical information platform as described in any one of claims 1-9, and includes the following steps: S101 connects to heterogeneous medical systems such as hospital information systems, laboratory information systems, medical image archiving and communication systems, and electronic medical record systems through a multi-source clinical data acquisition module. It uses interface adaptation technology to parse the data transmission protocols of different systems and acquire different types of clinical data, including patient basic information, diagnosis and treatment process records, test results, medication information, image data, and pathology reports. S102 transmits the collected multi-type clinical data to the central platform data storage and scheduling module. Based on a storage method that combines a distributed file storage architecture with a relational database, the data is classified and stored according to data type, generation time, and treatment stage. At the same time, a data indexing mechanism is established, and the data is dynamically scheduled and efficiently transmitted through a load balancing algorithm according to the data access requests of subsequent modules. S103, the clinical feature intelligent extraction module receives the scheduled data, uses feature recognition algorithms to mine the labeled features in the data that are related to disease diagnosis, treatment plan selection and prognosis assessment, and processes the feature information through feature screening, time series association and heterogeneous fusion techniques to provide data support for decision reasoning; S104, the dynamic construction module for decision rules is based on authoritative clinical diagnosis and treatment guidelines, industry standards and historical medical data. It uses a rule extraction algorithm to generate an initial set of decision rules, uses rule conflict detection technology to identify logical contradictions, establishes a rule update mechanism, and dynamically adjusts the rule set according to new medical data and revisions to diagnosis and treatment guidelines. S105, the multi-dimensional decision reasoning module calls the completed decision rule set, combines the extracted clinical features, and uses different reasoning methods such as rule reasoning, probabilistic reasoning and case matching reasoning to perform multi-path operations, and uses cross-validation technology of reasoning results to exclude unreliable conclusions; S106, the precise decision result output module, performs structured processing on the verified decision information, presents the decision results using visualization technology according to the clinical doctors' treatment habits and different application scenarios, transmits the results to the clinical workstation through a standardized interface, and establishes a data feedback channel to receive feedback information after clinical application, providing data support for subsequent system optimization.

Images