Remote data fusion system based on real world data platform

Abstract

The invention provides a remote data fusion system based on a real world data platform, which comprises an equipment acquisition module, an edge calculation module, a data transmission and integration module and an application interface module, and is characterized in that physiological parameters and environmental data acquired by wearable equipment are acquired, and local model analysis is carried out through a federal learning technology; anonymously uploading to a real world data platform, transmitting original data by adopting a low-delay communication protocol, and integrating the original data and multi-source data in the real world data platform and storing the integrated data in a database; the method can be called by multiple application layer interfaces. Real-time physiological parameters are obtained by adopting wearable equipment, federated learning and edge calculation are deployed, data privacy is protected, the data are fused with a real world data platform, multi-modal data are provided for multiple practical application interfaces, and various data platforms and systems are provided for clinical tests and the like.

Description

Technical Field

[0001] This application relates to the technical field of medical data platforms, and particularly to a remote data fusion system based on a real-world data platform. Background Art

[0002] A real-world data platform records various healthcare activities in the real world. These data are generated during people's daily medical practices and healthcare processes, and are authentic and representative, capable of reflecting patient characteristics, the effectiveness of treatment measures, and clinical conditions. Among them, real-world data includes, but is not limited to, electronic medical records, medical insurance data, drug sales, etc. Most existing real-world medical data platforms have collected data such as clinical diagnoses and imaging files, and are connected to various actual medical application scenarios; however, for remote clinical trial data collection, it has not been integrated into the real-world data platform, and such clinical trial data cannot be effectively studied and utilized. Summary of the Invention

[0003] In view of the above problems, the present invention provides a remote data fusion system based on a real-world data platform, which uses wearable devices to obtain real-time physiological parameters, and deploys federated learning and edge computing to protect data privacy, and integrates with the real-world data platform to provide multi-modal data for multiple actual application interfaces, and provides diverse data platforms and systems for clinical trials, etc.

[0004] In a first aspect, the present invention provides a remote data fusion system based on a real-world data platform, including the following modules:

[0005] A device acquisition module, which acquires physiological parameters and environmental data collected by wearable devices to obtain raw data;

[0006] An edge computing module, which performs data preprocessing on the raw data, performs local model analysis through federated learning technology, and anonymously uploads it to the real-world data platform;

[0007] A data transmission and integration module, which uses a low-latency communication protocol to transmit the raw data, and integrates the raw data with multi-source data in the real-world data platform and stores it in a database;

[0008] An application interface module, where the database can be called by multiple application layer interfaces. Among them, when visualization is called, a dynamic trend graph of the raw data can be displayed on the visualization panel.

[0009] Further, the federated learning technology, the specific working process includes:

[0010] Initialize a global model on the cloud server and distribute it to each device acquisition end;

[0011] Each device acquisition end independently trains and predicts or analyzes a model using local data, and calculates model updates;

[0012] The device acquisition end sends the calculated model update parameters to the cloud server;

[0013] The cloud server collects the model updates from each device acquisition end, integrates them using a specific aggregation algorithm, and updates the global model;

[0014] The cloud server redistributes the updated global model to each device acquisition end again and conducts the next round of model training.

[0015] Furthermore, the key steps of the aggregation algorithm include the following:

[0016] The server end initializes the global model parameters ;

[0017] Send the global model to each device end, and each device end conducts several rounds of training using local data and updates the model parameters ;

[0018] The server end collects the model parameters of each device end, introduces a regularization term in the objective function, which is expressed as:

[0019]

[0020] , and updates the global model parameters for the new round .

[0021] Furthermore, the low-latency communication protocol includes wifi, 5G or 6G network communication.

[0022] In a second aspect, a computer-readable storage medium has instructions stored thereon, and when the instructions are executed by a processor, the remote data fusion system based on the real-world data platform described in any one of the above is implemented.

[0023] Through the above implementation, the following advantages or beneficial effects are achieved:

[0024] (1) Wearable devices are adopted to obtain physiological parameters in real time and integrate them with the real world, providing diverse clinical trial data;

[0025] (2) Using federated learning and edge computing not only protects data privacy but also provides distributed data collection;

[0026] (3) A platform and system for multi-modal data are provided, providing a standardized and sufficient data source for the research and practical application of clinical trials.

[0027] Other features and advantages of the present application will be described in the subsequent specification, and in part will be obvious from the specification, or can be understood by implementing the present application. The objectives and other advantages of the present application can be achieved and obtained through the structures specifically pointed out in the written specification, claims, and drawings. Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0029] Figure 1 It is a schematic diagram of a remote data fusion system based on a real-world data platform provided by the present invention. Detailed Embodiments

[0030] To make the objectives, technical solutions, and advantages of the present application clearer and more understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0031] Embodiment 1

[0032] As Figure 1 shown, a remote data fusion system based on a real-world data platform includes the following modules:

[0033] The device acquisition module acquires physiological parameters and environmental data collected by wearable devices to obtain raw data;

[0034] The edge computing module preprocesses the raw data, performs local model analysis through federated learning technology, and then anonymously uploads it to the real-world data platform;

[0035] The data transmission and integration module transmits the raw data using a low-latency communication protocol, and integrates the raw data with multi-source data in the real-world data platform and stores it in a database;

[0036] The application interface module, the database can be called by multiple application layer interfaces. Among them, when visualizing, a dynamic trend graph of the original data can be displayed on the visualization panel.

[0037] Obtain the physiological parameters and environmental data collected by the wearable device to get the original data; among them, the wearable device includes a smart watch, a continuous glucose monitor, an electrocardiogram patch, a motion sensor, etc., and the data collection includes physiological parameters such as heart rate, blood oxygen, steps, blood glucose, blood pressure, sleep quality, etc., and environmental information such as temperature and location.

[0038] The edge computing module preprocesses the original data, performs local model analysis through federated learning technology, and anonymously uploads it to the real-world data platform.

[0039] Perform preprocessing operations on the data collected by the device, including data denoising, outlier filtering, etc. And apply federated learning technology locally to analyze the collected data. In specific implementations, use blood glucose data to monitor the diabetic condition of patients and reasonably optimize diet management; use electrocardiogram patches to monitor electrocardiogram data changes, predict whether a heart disease alarm is triggered, and provide timely treatment. Among them, in order to protect local data privacy, federated learning technology is adopted, and the specific workflow includes:

[0040] Initialize a global model on the cloud server and distribute it to each device collection end;

[0041] Each device collection end independently trains and predicts or analyzes the model using local data and calculates the model update;

[0042] The device collection end sends the calculated model update parameters to the cloud server;

[0043] The cloud server collects the model updates from each device collection end, integrates and updates the global model using a specific aggregation algorithm; among them, the aggregation algorithm includes but is not limited to FedAvg, FedProx.

[0044] The cloud server redistributes the updated global model to each device collection end and conducts the next round of model training.

[0045] The key steps of the FedProv aggregation algorithm include:

[0046] The server side initializes the global model parameters ;

[0047] Send the global model to each device end, and each device end uses local data for several rounds of training and updates the model parameters ;

[0048] The server side collects the model parameters of each device end , introduce a regularization term in the objective function, expressed as:

[0049]

[0050] , update the global model parameters for a new round .

[0051] Adopt federated learning technology to collect data from local devices for model training, prediction and analysis. In specific implementation, dynamic management of chronic diseases, observation of clinical trials or health management, etc. can be carried out.

[0052] The data transmission and integration module uses a low-latency communication protocol to transmit the original data, and integrates the original data with multi-source data in the real-world data platform and stores it in a database.

[0053] In specific implementation, the communication protocol includes wifi, 5G or 6G communication to achieve low-latency transmission, and transmits the anonymized data to the real-world data platform and integrates it with the multi-source data in the platform. Among them, the multi-source data includes various modalities of data such as clinical diagnosis and treatment, image data of image files, etc. The storage solutions adopted include relational databases, time-series databases, and knowledge graphs with relational networks, etc. In specific implementation, there are knowledge graphs of disease-gene-drug relational networks, relational databases of patient diagnosis records and basic information, etc., storage of special format DICOM files, and storage of time-series data of wearable devices.

[0054] The application interface module, the database can be called by multiple application layer interfaces. Among them, when visualizing is called, the dynamic trend chart of the original data can be displayed on the visualization panel.

[0055] In specific implementation, in the real-world data platform, multi-modal data is collected and stored, associated and managed with various adapted solutions. When various modalities of data are docked to the application layer, they will be associated with many actual applications, including data calls from clinical research, hospital systems or public health organizations, specifically including diagnostic-level imaging clouds, intelligent medical insurance cost control, epidemic prevention and control, etc. And by calling the time-series data of wearable devices, a dynamic trend chart is drawn and displayed on the visualization panel.

[0056] Embodiment 2

[0057] A computer-readable storage medium, on which instructions are stored, and when the instructions are executed by a processor, the remote data fusion system based on the real-world data platform as described in any one of the above is implemented.

[0058] Through the above solutions, the following advantages or beneficial effects are achieved:

[0059] (1) Adopt wearable devices to obtain physiological parameters in real time and integrate them with the real world, providing diverse clinical trial data;

[0060] (2) By using federated learning and edge computing, it not only protects data privacy but also provides distributed data collection;

[0061] (3) It provides platforms and systems for multi-modal data, offering a standardized and sufficient data source for the research and practical application of clinical trials.

[0062] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A remote data fusion system based on a real-world data platform, characterized in that, It includes the following modules: The device acquisition module acquires physiological parameters and environmental data collected by the wearable device to obtain raw data; The edge computing module performs data preprocessing on the raw data, conducts local model analysis through federated learning technology, and anonymously uploads it to the real-world data platform; The data transmission and integration module transmits the raw data using a low-latency communication protocol, and integrates the raw data with multi-source data in the real-world data platform and stores it in the database; The application interface module, the database can be called by multiple application layer interfaces. Among them, when visualization is called, a dynamic trend chart of the raw data can be displayed on the visualization panel.

2. The remote data fusion system based on the real-world data platform according to claim 1, wherein The specific workflow of the federated learning technology includes: Initialize a global model on the cloud server and distribute it to each device acquisition end; Each device acquisition end independently trains and predicts or analyzes the model using local data and calculates the model update; The device acquisition end sends the calculated model update parameters to the cloud server; The cloud server collects the model updates of each device acquisition end, integrates them using a specific aggregation algorithm, and updates the global model; The cloud server redistributes the updated global model to each device acquisition end and conducts the next round of model training.

3. The remote data fusion system based on the real-world data platform according to claim 2, wherein, The key steps of the aggregation algorithm include The following: The server end initializes the global model parameters; Send the global model to each device end, and each device end uses local data for several rounds of training and updates the model parameters; The server-side collects the model parameters of each device-side, and introduces a regularization term in the objective function, which is expressed as: ; Update the global model parameters for a new round .

4. The remote data fusion system based on the real-world data platform according to claim 1, wherein The low-latency communication protocol includes wifi, 5G or 6G network communication.

5. A computer-readable storage medium, on which instructions are stored, characterized in that, When the instruction is executed by the processor, it implements the remote data fusion system based on the real-world data platform as described in any one of claims 1-4.

Images