Distributed architecture driven cross-platform medical information processing method and system

Abstract

The application discloses a distributed architecture driven cross-platform medical information processing method and system, relates to the related technical field of medical data processing, and comprises the following steps: constructing a distributed information processing network, integrating heterogeneous nodes of multiple medical information platforms, and collecting multi-modal medical data from different medical information platforms connected in real time; after standardization processing, the medical data is transmitted and stored by using a distributed message middleware; a medical analysis sample data is extracted by a data analysis node, a data mining task is cooperated based on a federal learning framework, a data mining result is generated, and a visual decision support report is aggregated. The technical problems of the prior art that cross-platform medical data is difficult to fuse and interwork, centralized architecture processing capacity is insufficient, and sensitive data sharing exists privacy security risks are solved, and the technical effects of improving the real-time performance and efficiency of medical data processing, guaranteeing data privacy security, and enhancing the accuracy and comprehensiveness of clinical decision support are achieved.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to a cross-platform medical information processing method and system driven by a distributed architecture. Background Technology

[0002] Medical institutions generally establish their own information management systems, such as Hospital Information Systems (HIS), Picture Archiving and Communication Systems (PACS), Laboratory Information Systems (LIS), and Electronic Medical Records (EMR), which greatly improve the efficiency of medical work and the level of data management. However, due to differences in the construction period, vendor standards, and application goals, these information management systems are often heterogeneous and isolated, forming "information silos" and "data islands." Most existing healthcare information platforms employ centralized architectures or single cross-platform data exchange interfaces. Faced with massive, high-speed, and multi-modal healthcare data, such as structured medical records, semi-structured examination reports, and unstructured medical images, their processing capabilities struggle to meet real-time requirements. Furthermore, healthcare data is highly sensitive and privacy-sensitive. Current cross-platform data sharing methods typically require physically centralizing data from different institutions at a central node for cleaning and analysis, which easily leads to data leakage risks during transmission and storage, and does not comply with the "data not leaving the domain" or minimal data flow compliance requirements. In addition, the different data standards used by different healthcare institutions (such as general hospitals, specialized hospitals, and community clinics) result in inconsistent data formats, encoding rules, and interface protocols, making cross-platform data fusion and interoperability extremely difficult. This prevents the formation of a continuous and complete patient diagnosis and treatment view, severely hindering big data-based clinical decision support.

[0003] Therefore, current technologies face challenges such as difficulty in integrating and interoperating data across medical platforms, insufficient processing capabilities of centralized architectures, and privacy and security risks associated with sharing sensitive data. Summary of the Invention

[0004] This application provides a distributed architecture-driven cross-platform medical information processing method and system, which solves the technical problems existing in the prior art, such as the difficulty in integrating and interoperating data across medical platforms, the insufficient processing capacity of centralized architecture, and the privacy and security risks of sharing sensitive data. It achieves the technical effects of improving the real-time performance and efficiency of medical data processing, ensuring data privacy and security, and enhancing the accuracy and comprehensiveness of clinical decision support.

[0005] This application provides a cross-platform medical information processing method driven by a distributed architecture. The method includes: constructing a distributed information processing network, which comprises multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes, wherein each node is distributed and deployed on different medical information platforms; acquiring multimodal medical data in real time from the different medical information platforms connected to by the multiple heterogeneous data acquisition nodes; standardizing the multimodal medical data and transmitting the standardized medical data to the data storage nodes through a distributed message middleware; extracting medical analysis sample data from the data storage nodes and collaboratively executing data mining tasks based on a federated learning framework to generate data mining results corresponding to each platform and transmitting them to the decision support nodes; and dynamically aggregating the data mining results from different platforms through the decision support nodes to generate a visualized decision support report.

[0006] In a possible implementation, multimodal medical data is acquired in real time from different medical information platforms through the multiple heterogeneous data acquisition nodes, including: connecting to medical monitoring equipment through medical device acquisition nodes among the multiple heterogeneous data acquisition nodes to acquire real-time patient monitoring data; connecting to the hospital information system through hospital system acquisition nodes among the multiple heterogeneous data acquisition nodes to acquire patient medical record data; receiving patient-reported data through mobile terminal acquisition nodes among the multiple heterogeneous data acquisition nodes; and integrating the patient real-time monitoring data, medical record data, and self-reported data into the multimodal medical data.

[0007] In a possible implementation, after standardizing the multimodal medical data, the standardized medical data is transmitted to the data storage node through a distributed message middleware. This includes: performing format conversion, semantic annotation, and time alignment on the multimodal medical data to generate standardized medical data; and transmitting the standardized medical data to the data storage node according to a preset routing rule through the distributed message middleware.

[0008] In a possible implementation, the data analysis node extracts medical analysis sample data from the data storage node and collaboratively executes data mining tasks based on a federated learning framework to generate data mining results for different platforms. This includes: extracting medical analysis sample data that meets the analysis conditions from the data storage node and transmitting it to the data analysis node; the data analysis node training a model on the medical analysis sample data locally to generate local model parameters; exchanging the local model parameters of each data analysis node through the federated learning framework and aggregating them to generate a global analysis model; and applying the global analysis model to mine the local medical analysis sample data of each node to generate data mining results corresponding to each platform.

[0009] In one possible implementation, the local model parameters of each data analysis node are exchanged through a federated learning framework and aggregated to generate a global analysis model. This includes: establishing a secure communication channel between the data analysis nodes; transmitting encrypted local model parameters through the secure communication channel; and using a parameter aggregation algorithm to perform a weighted average calculation on the received local model parameters of each node, and using the calculation result as the update parameter of the global analysis model.

[0010] In a possible implementation, the global analysis model is applied to mine the medical analysis sample data to generate data mining results for different platforms, including: inputting the local medical analysis sample data of each node into the updated global analysis model, performing medical data risk prediction, and generating data risk analysis results; generating data mining results containing prediction indicators and confidence assessments based on the data risk analysis results, and marking the data mining results with the corresponding source platform identifier.

[0011] In one possible implementation, the decision support node dynamically aggregates data mining results from different platforms to generate a visualized decision support report, including: receiving data mining results from various data analysis nodes; performing confidence-weighted fusion of conflicting data mining results based on the data mining results to generate fused data analysis results; and matching and validating the fused data analysis results with a clinical guideline library to generate a visualized decision support report.

[0012] This application also provides a cross-platform medical information processing system driven by a distributed architecture. The system includes: an information processing network construction module for building a distributed information processing network, which comprises multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes, wherein each node is distributed across different medical information platforms; a medical data acquisition module for acquiring multimodal medical data in real time from the different medical information platforms connected to by the multiple heterogeneous data acquisition nodes; a medical data processing and transmission module for standardizing the multimodal medical data and transmitting the standardized medical data to the data storage nodes via a distributed message middleware; a data mining result generation module for the data analysis nodes to extract medical analysis sample data from the data storage nodes, collaboratively execute data mining tasks based on a federated learning framework, generate data mining results corresponding to each platform, and transmit them to the decision support nodes; and a decision report generation module for dynamically aggregating the data mining results from different platforms through the decision support nodes to generate a visualized decision support report.

[0013] This application proposes a distributed architecture-driven cross-platform medical information processing method and system. This system constructs a distributed information processing network, integrating heterogeneous nodes from multiple medical information platforms. Multimodal medical data is collected in real-time from the different medical information platforms each platform connects to. After standardization processing, the data is transmitted and stored using a distributed message middleware. Data analysis nodes extract medical analysis sample data and collaborate on data mining tasks based on a federated learning framework to generate data mining results, which are then aggregated into a visualized decision support report. This addresses the technical challenges of cross-platform data integration and interoperability, insufficient processing capabilities of centralized architectures, and privacy and security risks associated with sharing sensitive data in existing technologies. It achieves the technical effects of improving the real-time performance and efficiency of medical data processing, ensuring data privacy and security, and enhancing the accuracy and comprehensiveness of clinical decision support. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 A schematic diagram of a distributed architecture-driven cross-platform medical information processing method provided in this application embodiment.

[0016] Figure 2 A schematic diagram of the structure of a cross-platform medical information processing system driven by a distributed architecture, provided in an embodiment of this application.

[0017] Figure labeling: Information processing network construction module 10, medical data acquisition module 20, medical data processing and transmission module 30, data mining result generation module 40, decision report generation module 50. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0019] This application provides a distributed architecture-driven cross-platform medical information processing method, such as... Figure 1 As shown, the method includes: Step S100: Construct a distributed information processing network, which includes multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes, wherein each node is distributed and deployed on different medical information platforms.

[0020] Preferably, a topology consisting of multiple servers or service instances with different functions is constructed to define a distributed information processing network. This network includes multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes. Each type of component is deployed in different physical locations within different medical institutions or medical information systems. All nodes communicate via TCP / IP. Heterogeneous data acquisition nodes refer to front-end machines or interface servers deployed within the intranet of each medical institution. For example, these are directly physically connected or connected via network cables to the data output ports of medical equipment (such as CT scanners and monitors) and the database interface of the hospital information system (HIS). Data storage nodes refer to database services deployed in each medical institution or on a private cloud. The data analysis nodes are computing servers or distributed file systems responsible for receiving and persisting standardized medical data, such as storage array servers with large-capacity hard drives; data analysis nodes are computing servers with GPUs (graphics processing units) or high-performance CPUs (central processing units) deployed in various medical institutions, responsible for executing local model training tasks in federated learning, such as rack servers; decision support nodes are central servers deployed in management institutions such as medical consortium data centers or health commission information centers, responsible for aggregating parameters from various analysis nodes and making decisions; all nodes are connected through a network to form a distributed information processing network, where the computation of each node only occurs locally, and only model parameters or non-privacy aggregation results are transmitted across platforms.

[0021] Step S200: Multimodal medical data is acquired in real time from the different medical information platforms to which each of the multiple heterogeneous data acquisition nodes is connected.

[0022] Step S200 further includes: connecting to medical monitoring equipment through medical device acquisition nodes among the plurality of heterogeneous data acquisition nodes to obtain real-time patient monitoring data; connecting to the hospital information system through hospital system acquisition nodes among the plurality of heterogeneous data acquisition nodes to obtain patient medical record data; receiving patient self-reported data through mobile terminal acquisition nodes among the plurality of heterogeneous data acquisition nodes; and integrating the patient real-time monitoring data, medical record data, and self-reported data into the multimodal medical data.

[0023] Preferably, the medical device acquisition nodes in the heterogeneous data acquisition nodes refer to medical IoT acquisition units deployed in ICU, CCU, or general wards. These units connect to medical monitoring devices, such as ECG monitors, ventilators, infusion pumps, and blood glucose meters, via serial cables, network cables, or Bluetooth. Following the communication protocols specified by the medical device manufacturers, they continuously read the data streams emitted by the medical monitoring devices to obtain the patient's real-time vital signs parameters, including heart rate waveforms, blood pressure values, blood oxygen saturation, and respiratory rate, generating real-time patient monitoring data. The hospital system acquisition nodes in the heterogeneous data acquisition nodes refer to interfaces deployed in the hospital's intranet server area or computer room. The server program establishes a connection with the backend database of the hospital information system through database connections or API interface calls. This may include the hospital information system (HIS) reading patient registration information, medical order records, and expense details; the electronic medical record system (EMR) reading patient chief complaints, present medical history, physical examination records, and progress notes; the laboratory information system (LIS) reading patient blood routine, biochemical indicators, and other test results; and the medical imaging system (PACS) reading patient CT, MRI, and X-ray images and report texts. This node extracts medical record data for specific patients within a specific time range according to a set timed task or trigger.

[0024] Preferably, the mobile terminal acquisition node in the heterogeneous data acquisition node refers to a mobile backend service program deployed on a public cloud server or in the hospital's DMZ. This program runs an HTTP / HTTPS server and receives data actively uploaded by patients from mobile devices such as mobile applications, WeChat mini-programs, or smart bracelets / watches through open API interfaces. This data includes at least symptom scales, medication records, and dietary diaries filled out by the patients themselves, as well as home blood pressure, blood sugar, weight, steps taken, and sleep quality data collected by the mobile device sensors. The data is then formatted and checked for completeness to ensure the reliability of the data source and the correctness of the format. Finally, using the patient's unique identifier such as outpatient number, inpatient number, or ID card number and timestamp, the patient's real-time monitoring data, medical record data, and self-reported data are integrated into multimodal medical data, which also includes machine-generated numerical signals, text records entered by doctors, and subjective information narrated by the patients.

[0025] Step S300: After standardizing the multimodal medical data, the standardized medical data is transmitted to the data storage node through a distributed message middleware.

[0026] Step S300 further includes performing format conversion, semantic annotation, and time alignment processing on the multimodal medical data to generate standardized medical data; and transmitting the standardized medical data to the data storage node according to preset routing rules through a distributed message middleware.

[0027] Preferably, multimodal medical data is standardized, including format conversion, semantic annotation, and time alignment. Specifically, format conversion refers to the acquisition node or standardization processor calling a format conversion program to unify heterogeneous data formats from different sources into a standardized format. This includes medical device data conversion, hospital system data conversion, and encoding standardization. For example, waveform data streams received from ECG monitors are converted into standard JSON objects or Parquet columnar storage formats, HL7 messages or custom text messages read from HIS databases are converted into system-wide XML or Protobuf data structures, and different encoding tables used by different hospitals are uniformly mapped to standard terms.

[0028] Preferably, semantic annotation processing refers to adding machine-understandable metadata tags to the converted data fields, transforming the data from simple numerical values ​​into clinically meaningful knowledge units. This includes ontology mapping, unit annotation, and relation annotation. For example, mapping field names to a standard medical ontology and annotating their standard concept codes, explicitly annotating the units of measurement for numerical fields, and annotating the relationships between data, such as labeling a certain ECG waveform data as "resting-state acquisition" and the associated diagnosis as "coronary atherosclerosis." Time alignment processing refers to unifying the timestamps of all data to a standard time zone and standard precision and reordering them according to patient ID and timeline. This includes time zone normalization and time series correction. For example, calibrating the timestamps from mobile terminals and hospital systems to Coordinated Universal Time (UTC), and aligning high-frequency data from medical devices, low-frequency data from hospital systems, and sporadic data reported by patients according to a unified timeline to ensure that the data is continuous and ordered in the time dimension.

[0029] Preferably, the standardized data is published to a message queue cluster and transmitted to data storage nodes via a distributed message middleware according to preset routing rules. The distributed message middleware refers to message queue software such as Kafka, RabbitMQ, or RocketMQ deployed in a distributed network, running on an independent server or container, responsible for receiving messages pushed by producers and temporarily storing them in a disk queue. The preset routing rules refer to the forwarding logic pre-configured in the message middleware, which may include routing by patient ID, by data source, or by data type. For example, patient IDs are sent to the corresponding storage shard after modulo operation, all ECG waveform data are sent to the time-series database storage node, and all medical record texts are sent to the document database storage node. The message middleware retrieves standardized medical data from the queue, determines the IP address and port of the target storage node according to the preset routing rules, establishes a TCP connection, and pushes the data there. If the target node is temporarily unavailable, the message middleware persistently saves the data locally and resends it after the node recovers, ensuring no data loss.

[0030] In step S400, the data analysis node extracts medical analysis sample data from the data storage node, and collaboratively executes data mining tasks based on the federated learning framework to generate data mining results corresponding to each platform, and transmits them to the decision support node.

[0031] Step S400 further includes: extracting medical analysis sample data that meets the analysis conditions from the data storage node and transmitting it to the data analysis node; training a model on the medical analysis sample data locally on the data analysis node to generate local model parameters; exchanging the local model parameters of each data analysis node through a federated learning framework and aggregating them to generate a global analysis model; and applying the global analysis model to mine the local medical analysis sample data of each node to generate data mining results corresponding to each platform.

[0032] Preferably, data analysis nodes deployed in the computer rooms of various medical institutions initiate data query requests to data storage nodes within the same institution. That is, the data analysis nodes execute pre-written SQL query statements or API calls to filter data that meets the analysis criteria. For example, they filter clinical records and test data of "patients diagnosed with type 2 diabetes within the past 30 days and aged between 18 and 75 years old" to determine the medical analysis sample data, including numerical features, text features, and labels, which are transmitted through fiber optic or network cables within the internal local area network. The data analysis nodes utilize local CPU / GPU computing power to load the extracted sample data into local memory, train machine learning models, and perform iterative calculations. That is, they run machine learning frameworks such as TensorFlow, PyTorch, or Spark MLlib, and perform calculations such as forward propagation, loss calculation, and backpropagation. After training, they output local model parameters, including the weight matrix and bias terms within the model, reflecting the degree of contribution of the institution's data features to the model.

[0033] Preferably, all data analysis nodes in different medical institutions communicate and fuse parameters through a federated learning coordination server in one or more rounds. Specifically, each data analysis node encrypts its local model parameters and sends them to the federated learning aggregation server via the network to execute the federated averaging algorithm. This includes weighting and summing the model parameters of each node according to the proportion of its local data to the total data to obtain new model parameters. For example, the parameter weights of hospital A with 1000 samples are twice that of hospital B with 500 samples. The calculated new model parameters are then sent back to the data analysis servers of all participating nodes. The training is then performed again based on the local data, and this process is repeated multiple times until the global model converges, and the global analysis model is generated.

[0034] Preferably, the local medical analysis sample data of each node is mined through a global analysis model. That is, the data analysis node inputs the feature data of local patients into the global analysis model to perform forward propagation calculation and outputs predicted values. For example, inputting a patient's current heart rate, blood pressure and blood sugar sequence, the output is "the probability of heart failure in the next 24 hours is 87%". The output numerical results are then encapsulated into structured data to generate data mining results containing patient identifiers, predicted indicator names, predicted values, confidence scores, etc. Finally, a tag field of the medical institution code to which the data analysis node belongs is added to generate data mining results corresponding to each platform, which are temporarily stored locally or sent directly to the decision support node.

[0035] Furthermore, step S400 also includes establishing a secure communication channel between the data analysis nodes; transmitting encrypted local model parameters through the secure communication channel; and using a parameter aggregation algorithm to perform a weighted average calculation on the received local model parameters of each node, and using the calculation result as the update parameter of the global analysis model.

[0036] Preferably, a secure communication channel is established between each data analysis node. Specifically, TLS is used between the aggregation server and the data analysis servers of each node. 1.3 (Transport Layer Security Protocol version 1.3) initiates a handshake, exchanges digital certificates, and verifies each other's identities. After successful verification, a session key is negotiated. Based on the negotiated key, all transmitted data is encrypted using a symmetric encryption algorithm (such as AES-256). Even if data packets are intercepted in the network, attackers cannot decrypt and recover the original content. Each node's data analysis server encrypts the trained local model parameters and transmits them to the aggregation server through a secure communication channel. This includes serializing the local model parameters into a byte stream and applying homomorphic encryption or differential privacy, adding random noise to the parameters, or encrypting the parameters using the aggregation server's public key. This ensures that even if the aggregation server cannot directly see the original plaintext parameter values, the parameters are sent from the medical institution's router to the gateway of the data center where the aggregation server is located via the Internet. After receiving the local model parameters from all participating nodes, the aggregation server decrypts them using its private key, deserializes the byte stream back into a weight matrix in memory, and performs a weighted average calculation using a federated averaging algorithm. The calculation result is used as the update parameters for the global analysis model.

[0037] Furthermore, step S400 also includes inputting the local medical analysis sample data of each node into the updated global analysis model, performing medical data risk prediction, and generating data risk analysis results; based on the data risk analysis results, generating data mining results containing prediction indicators and confidence assessments, and marking the data mining results with the corresponding source platform identifier.

[0038] Preferably, data analysis servers deployed within various medical institutions input locally stored medical analysis sample data, such as patients' age, gender, systolic blood pressure, diastolic blood pressure, fasting blood glucose, glycated hemoglobin, low-density lipoprotein, and creatinine levels, into an updated global analysis model to perform medical data risk prediction. This involves calculating a risk score for each patient through weight matrix operations and forward propagation using an activation function, serving as the data risk analysis result. For example, an output value of 0.89 might indicate "the patient has an 89% risk of experiencing a major adverse cardiovascular event within the next 12 months." The data risk analysis results are then post-processed, correlating the model output value with predefined clinical indicators. For instance, if the risk probability is high... If the value is equal to 0.7, a warning indicator for "high-risk cardiovascular disease" is generated. Then, a confidence score is calculated based on the model's built-in reliability, data integrity, or model consistency, thereby generating data mining results containing predictive indicators and confidence assessments. Finally, when generating the data mining result data package, the data analysis server reads the local configuration file or environment variables to obtain a pre-assigned unique institution code, and inserts this code as a fixed field into the result data package. This code is transmitted to the decision support node along with the data mining results. The decision support node can use this identifier to trace the source of the results. For example, when it finds that the prediction results of a certain platform conflict with those of other platforms, it can quickly locate the source or, when generating a visualization report, display the prediction results according to the medical institution classification.

[0039] Step S500: Through the decision support node, dynamically aggregate data mining results from different platforms to generate a visualized decision support report.

[0040] Step S500 further includes receiving data mining results from each data analysis node; performing confidence-weighted fusion of conflict data mining results based on the data mining results to generate fused data analysis results; and matching and validating the fused data analysis results with a clinical guideline library to generate a visual decision support report.

[0041] Preferably, the decision support server deployed in the management center's computer room receives data mining results from various data analysis nodes via message queues or API interfaces. It then decrypts and validates the data, extracting core information such as the patient's anonymized ID, risk indicator name, risk value, confidence level, source platform identifier, and timestamp. The server temporarily stores the parsed results in a memory queue or temporary buffer. Based on the data mining results, the data in the temporary queue is grouped and aggregated according to patient ID and time window. The system checks for situations where two or more risk indicators have the same name but significantly different values, or where the indicator names themselves contradict each other. When a conflict is detected, multiple conflicting data mining results for the same patient with inconsistent conclusions are weighted and merged according to confidence level to generate a final risk value as the fused data analysis result. Finally, the fused data analysis result is used as input for rule matching validation in a clinical guideline library. Based on the guideline recommendations, a visual decision support report is generated, such as data visualization charts, guideline recommendation text, and patient information summaries, which are then pushed to the attending physician's computer desktop or mobile application via the hospital intranet.

[0042] In the above text, refer to Figure 1 This paper describes in detail a cross-platform medical information processing method driven by a distributed architecture according to embodiments of the present invention. Next, reference will be made to... Figure 2 This invention describes a cross-platform medical information processing system driven by a distributed architecture according to embodiments of the present invention.

[0043] The distributed architecture-driven cross-platform medical information processing system according to embodiments of the present invention addresses the technical problems in existing technologies, such as the difficulty in integrating and interoperating data across medical platforms, the insufficient processing capabilities of centralized architectures, and the privacy and security risks associated with sharing sensitive data. It achieves the technical effects of improving the real-time performance and efficiency of medical data processing, ensuring data privacy and security, and enhancing the accuracy and comprehensiveness of clinical decision support. Figure 2 As shown, the cross-platform medical information processing system driven by a distributed architecture includes: an information processing network construction module 10, a medical data acquisition module 20, a medical data processing and transmission module 30, a data mining result generation module 40, and a decision report generation module 50.

[0044] The information processing network construction module 10 is used to construct a distributed information processing network, which includes multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes, wherein each node is distributed and deployed on different medical information platforms. The medical data acquisition module 20 is used to acquire multimodal medical data in real time from the different medical information platforms connected to by the multiple heterogeneous data acquisition nodes. The medical data processing and transmission module 30 is used to standardize the multimodal medical data and transmit the standardized medical data to the data storage nodes through a distributed message middleware. The data mining result generation module 40 is used by the data analysis nodes to extract medical analysis sample data from the data storage nodes, and collaboratively execute data mining tasks based on a federated learning framework to generate data mining results corresponding to each platform and transmit them to the decision support nodes. The decision report generation module 50 is used to dynamically aggregate the data mining results from different platforms through the decision support nodes to generate a visualized decision support report.

[0045] The specific configuration of the medical data acquisition module 20 will be described in detail below. The medical data acquisition module 20 further includes: connecting to medical monitoring equipment through a medical device acquisition node among the plurality of heterogeneous data acquisition nodes to acquire real-time patient monitoring data; connecting to a hospital information system through a hospital system acquisition node among the plurality of heterogeneous data acquisition nodes to acquire patient medical record data; receiving data voluntarily reported by patients through a mobile terminal acquisition node among the plurality of heterogeneous data acquisition nodes; and integrating the real-time patient monitoring data, medical record data, and voluntarily reported data into the multimodal medical data.

[0046] The specific configuration of the medical data processing and transmission module 30 will be described in detail below. The medical data processing and transmission module 30 further includes: performing format conversion, semantic annotation, and time alignment processing on the multimodal medical data to generate standardized medical data; and transmitting the standardized medical data to the data storage node according to preset routing rules through a distributed message middleware.

[0047] The specific configuration of the data mining result generation module 40 will be described in detail below. The data mining result generation module 40 further includes: extracting medical analysis sample data that meets the analysis conditions from the data storage node and transmitting it to the data analysis node; having the data analysis node perform model training on the medical analysis sample data locally to generate local model parameters; exchanging the local model parameters of each data analysis node through a federated learning framework and aggregating them to generate a global analysis model; and applying the global analysis model to mine the local medical analysis sample data of each node to generate data mining results corresponding to each platform.

[0048] The specific configuration of the data mining result generation module 40 will be described in detail below. The data mining result generation module 40 further includes: establishing a secure communication channel between the data analysis nodes; transmitting encrypted local model parameters through the secure communication channel; and using a parameter aggregation algorithm to perform a weighted average calculation on the received local model parameters of each node, using the calculation result as the update parameter of the global analysis model.

[0049] The specific configuration of the data mining result generation module 40 will be described in detail below. The data mining result generation module 40 further includes: inputting local medical analysis sample data from each node into the updated global analysis model, performing medical data risk prediction, and generating data risk analysis results; generating data mining results containing prediction indicators and confidence assessments based on the data risk analysis results, and marking the data mining results with the corresponding source platform identifier.

[0050] The specific configuration of the decision report generation module 50 will be described in detail below. The decision report generation module 50 further includes: receiving data mining results from various data analysis nodes; performing confidence-weighted fusion of conflict data mining results based on the data mining results to generate fused data analysis results; and matching and validating the fused data analysis results with a clinical guideline library to generate a visual decision support report.

[0051] The distributed architecture-driven cross-platform medical information processing system provided in this embodiment of the invention can execute the distributed architecture-driven cross-platform medical information processing method provided in this embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0052] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A cross-platform medical information processing method driven by a distributed architecture, characterized in that, The method includes: A distributed information processing network is constructed, which includes multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes, wherein each node is deployed in a distributed manner on different medical information platforms; Through the multiple heterogeneous data acquisition nodes, multimodal medical data is acquired in real time from the different medical information platforms they are connected to; After standardizing the multimodal medical data, the standardized medical data is transmitted to the data storage node through a distributed message middleware. The data analysis node extracts medical analysis sample data from the data storage node, and collaboratively executes data mining tasks based on the federated learning framework to generate data mining results corresponding to each platform, and transmits them to the decision support node. The decision support node dynamically aggregates data mining results from different platforms to generate a visual decision support report.

2. The distributed architecture-driven cross-platform medical information processing method as described in claim 1, characterized in that, Through the aforementioned multiple heterogeneous data acquisition nodes, multimodal medical data is acquired in real time from different medical information platforms, including: The medical device acquisition node among the multiple heterogeneous data acquisition nodes is connected to the medical monitoring device to obtain real-time monitoring data of the patient. By connecting the hospital system acquisition node among the multiple heterogeneous data acquisition nodes to the hospital information system, patient medical record data can be obtained. The mobile terminal acquisition node among the multiple heterogeneous data acquisition nodes receives data reported by patients voluntarily. The patient's real-time monitoring data, medical record data, and self-reported data are integrated into the multimodal medical data.

3. The distributed architecture-driven cross-platform medical information processing method as described in claim 1, characterized in that, After standardizing the multimodal medical data, the standardized medical data is transmitted to the data storage node via a distributed message middleware, including: The multimodal medical data is subjected to format conversion, semantic annotation, and time alignment to generate standardized medical data. The standardized medical data is transmitted to the data storage node according to preset routing rules through a distributed message middleware.

4. The distributed architecture-driven cross-platform medical information processing method as described in claim 1, characterized in that, The data analysis node extracts medical analysis sample data from the data storage node and collaboratively executes data mining tasks based on a federated learning framework to generate data mining results for different platforms, including: Medical analysis sample data that meets the analysis conditions is extracted from the data storage node and transmitted to the data analysis node; The data analysis node performs model training on the medical analysis sample data locally to generate local model parameters; The local model parameters of each data analysis node are exchanged through the federated learning framework and aggregated to generate a global analysis model. The global analysis model is applied to mine the local medical analysis sample data of each node, generating data mining results corresponding to each platform.

5. The distributed architecture-driven cross-platform medical information processing method as described in claim 4, characterized in that, By exchanging local model parameters among data analysis nodes through a federated learning framework, a global analysis model is generated, including: Establish a secure communication channel between the data analysis nodes; The encrypted local model parameters are transmitted through the secure communication channel. A parameter aggregation algorithm is used to calculate a weighted average of the local model parameters of each received node, and the calculation result is used as the update parameter of the global analysis model.

6. The distributed architecture-driven cross-platform medical information processing method as described in claim 5, characterized in that, The global analysis model is applied to mine the medical analysis sample data to generate data mining results for different platforms, including: The local medical analysis sample data of each node is input into the updated global analysis model to perform medical data risk prediction and generate data risk analysis results. Based on the data risk analysis results, data mining results containing predictive indicators and confidence assessments are generated, and the data mining results are marked with the corresponding source platform identifier.

7. The distributed architecture-driven cross-platform medical information processing method as described in claim 1, characterized in that, Through the aforementioned decision support node, data mining results from different platforms are dynamically aggregated to generate visualized decision support reports, including: Receive data mining results from various data analysis nodes; Based on the data mining results, the conflict data mining results are weighted and fused according to confidence level to generate fused data analysis results; The results of the fused data analysis are matched and validated with a clinical guideline database to generate a visual decision support report.

8. A cross-platform medical information processing system driven by a distributed architecture, characterized in that: The system is used to implement the distributed architecture-driven cross-platform medical information processing method according to any one of claims 1 to 7, the system comprising: An information processing network construction module is used to build a distributed information processing network, which includes multiple heterogeneous data acquisition nodes, data storage nodes, data analysis nodes, and decision support nodes, wherein each node is distributed and deployed on different medical information platforms. The medical data acquisition module is used to acquire multimodal medical data in real time from the different medical information platforms that they are connected to through the multiple heterogeneous data acquisition nodes; The medical data processing and transmission module is used to standardize the multimodal medical data and then transmit the standardized medical data to the data storage node through a distributed message middleware. The data mining result generation module is used by the data analysis node to extract medical analysis sample data from the data storage node, and collaboratively execute data mining tasks based on the federated learning framework to generate data mining results corresponding to each platform and transmit them to the decision support node. The decision report generation module is used to dynamically aggregate data mining results from different platforms through the decision support node to generate a visual decision support report.

Images