Hospital intranet-based kidney transplantation single-disease follow-up database system

Abstract

The technical scheme of the present application discloses a kidney transplantation single disease follow-up database system based on a hospital intranet environment, comprising a data integration layer, an intelligent analysis layer and a doctor-patient linkage layer. Through the cooperative operation of the three-layer technical architecture designed by the present application, the three core problems of the prior art are comprehensively solved, and the following technical effects are achieved: at the data integration level, automatic and real-time integration of multi-port medical data of kidney transplantation patients is realized, the data integration efficiency is improved by 90%, the accuracy rate reaches 99.5%, and the traceability and privacy security are significantly improved; at the data analysis level, the AI-driven analysis model realizes automatic generation of patient whole-cycle health reports, large-sample group feature mining and real-time abnormality early warning, and the doctor data analysis workload is reduced by 80%, and the abnormal data response time is shortened to the minute level.

Description

Technical Field

[0001] This invention relates to a single-disease follow-up database system for kidney transplantation. Background Technology

[0002] Among the effective renal replacement therapies for patients with end-stage renal disease, kidney transplantation is one of the key methods. The quality of life and prognosis of kidney transplant patients depend heavily on a scientific and efficient follow-up management system.

[0003] In the current era of rapid development in medical informatization, single-disease medical management is gradually becoming an important direction for improving the efficiency and quality of diagnosis and treatment. In the field of kidney transplantation, the patient's diagnosis and treatment process involves multiple scenarios such as outpatient, inpatient, and emergency departments. The resulting medical data is scattered across more than ten portals, including HIS systems, electronic medical records, laboratory systems, imaging systems, and prescription systems. At the same time, the data types cover multimodal content such as medical history, laboratory test results, medication records, surgical information, comorbidities and complications.

[0004] To address the technical shortcomings of current medical data systems, such as scattered patient information and difficulties in multi-port integration, doctors need to manually enter the diagnosis and follow-up data of kidney transplant patients using traditional Excel software, or entrust engineers to retrieve diagnosis and treatment data from multiple hospital databases, integrate them into multiple documents, and then manually combine them. These methods have problems such as high data usage costs, poor timeliness, difficulty in acquisition, high error rate, high risk of loss, low dimensionality, difficulty in cleaning, lack of traceability, difficulty in collaborative operation across different ports, and difficulty in protecting privacy data.

[0005] Current medical data systems can only collect raw data and cannot clean, integrate, or intelligently analyze large-sample, multimodal data. Therefore, their medical value to both doctors and patients is extremely limited. Doctors need to spend a significant amount of manpower analyzing and interpreting patient data, and patients rely on doctors for all medical advice. Furthermore, the lack of AI-based early warning systems for handling critical and abnormal medical data creates potential medical risks.

[0006] Current medical case follow-up systems are doctor-centric, lacking effective and accurate channels for collecting patient-side symptoms, signs, observation indicators, and local data. Furthermore, while current technology supports one-way patient communication, the lack of real-time, intelligent management on the doctor's end prevents collaborative doctor-patient interaction, resulting in delayed and inefficient remote patient management.

[0007] Specifically, to address the single-disease management needs of kidney transplantation, a management solution combining traditional medical data processing methods with a general-purpose medical follow-up system is currently adopted. Its core architecture and operating logic are as follows: (I) Data Acquisition and Integration Solution In the existing scheme, the integration of medical data of kidney transplant patients lacks a unified system support and mainly relies on a "manual-led + fragmented retrieval from multiple systems" model to achieve data aggregation.

[0008] 1. Data Acquisition Methods: Doctors need to collect patients' full-cycle medical data through two methods: First, manual entry, using office software such as Excel to input the patient's outpatient and inpatient medical records, test results, medication information, etc., into a table to form a personal patient data document; Second, cross-system retrieval and integration, for data scattered across different ports such as the hospital's outpatient system, inpatient system, laboratory system, and imaging system, it is necessary to entrust the hospital's information technology engineer to extract the data from each system separately through database retrieval, export it into multiple independent documents, and then have medical staff manually screen, copy and paste to integrate it, and finally form a complete patient data file.

[0009] 2. Data storage format: The integrated patient data is stored in local documents (such as Excel spreadsheets, Word documents, PDF reports) or in the form of scattered records in the hospital's general electronic medical record system. There is no centralized and structured database specifically for kidney transplantation. The data documents of different patients are independent of each other, making it impossible to achieve cross-patient and cross-cycle data association and centralized monitoring.

[0010] (II) Data Analysis and Application Solution Existing solutions can only achieve the "raw accumulation" of medical data, lacking the professional and intelligent analysis capabilities for kidney transplantation.

[0011] 1. Data Analysis Subject and Method: Data analysis relies entirely on manual work by doctors. During diagnosis and treatment or follow-up, doctors need to review the patient's historical data documents one by one, compare the test indicators (such as serum creatinine, blood urea nitrogen, immunosuppressant concentration, etc.), symptom changes, and medication adjustments at different times, and judge the trend of the patient's condition through personal clinical experience to form diagnosis and treatment recommendations.

[0012] 2. Scope of Data Application: The analysis results are only used for single-point medical decisions and cannot generate a full-cycle health report for patients. Furthermore, it cannot be used for population characteristic analysis of large sample patient data, such as the incidence of complications in different age groups or differences in the efficacy of specific immunosuppressants. Additionally, there is no automatic early warning mechanism for abnormal patient data, such as sudden increases or decreases in immunosuppressant concentrations or abnormal fluctuations in renal function indicators. These abnormalities require doctors to manually review the data periodically, making it impossible to detect risks in real time.

[0013] (III) Doctor-Patient Collaboration and Follow-up Management Plan The existing follow-up management system mainly relies on "passive doctor response + one-way patient communication" and lacks a two-way linkage intelligent management mechanism.

[0014] 1. Patient-side data feedback: Patients can only provide feedback to the medical system through traditional methods. These include: 1) in-person follow-up feedback, where patients verbally describe recent symptoms and changes in vital signs to their doctor during regular outpatient checkups; and 2) one-way online communication, using the hospital's general registration app, WeChat official account message function, or contacting the department nurse by phone to inform them of their discomfort. However, patients cannot directly upload visual information such as vital sign data (e.g., blood pressure, weight), or symptom photos, nor is there a dedicated channel to submit a health status report to the system.

[0015] 2. Medical Management Approach: Doctors cannot access patients' health data in real time while at home; they can only obtain information when patients proactively communicate or during offline follow-ups, and then adjust treatment plans accordingly. Patient feedback information must be manually compiled and recorded in patient data documents, lacking AI-assisted automatic classification and prioritization functions. This can lead to medical risks in critical situations (such as severe rejection symptoms) due to delayed information transmission and untimely processing.

[0016] Existing technical solutions have many core defects in practical applications, including the following three major technical problems: (i) Low efficiency and poor quality of multi-port data integration for single diseases Existing solutions rely on manual data entry and cross-system manual data integration, which presents six major problems: First, high cost, requiring significant time from doctors and extensive retrieval resources from engineers; second, poor timeliness, with long data integration cycles that fail to meet the need for real-time data retrieval during follow-ups; third, low accuracy, with errors and omissions easily occurring during manual entry and copy-paste processes; fourth, low data dimensionality, only able to integrate structured data, making it difficult to effectively integrate unstructured data such as imaging reports and doctors' medical records; fifth, lack of traceability, with scattered data sources making it difficult to quickly locate the original data source when data issues arise; and sixth, difficulty in privacy protection, with scattered local documents prone to leakage of patient privacy data due to device loss or lax access control.

[0017] (ii) Lack of ability to analyze large-sample multimodal data Existing solutions lack intelligent analysis capabilities, resulting in the underutilization of the value of medical data and presenting three major limitations: First, the analysis efficiency is low, as manual data analysis by doctors is time-consuming and labor-intensive, making it difficult to handle a large number of follow-up patients; second, the analysis accuracy is limited, relying on personal experience and judgment, making it susceptible to subjective factors and unable to accurately capture subtle trends in data changes; third, there is no risk warning mechanism, which cannot identify and issue warnings in real time for abnormal patient data, such as abnormal immunosuppressant concentrations or sudden changes in renal function indicators, posing potential medical risks and failing to enable large-sample data mining to support clinical research and treatment plan optimization.

[0018] (III) Shortcomings of existing technology: Delayed doctor-patient collaboration and low management efficiency. The existing doctor-patient communication and follow-up management model has significant barriers: First, patient data collection is untimely and incomplete, lacking a systematic channel for collecting patient home health data (symptoms, signs), making it impossible to monitor patient dynamics in real time; second, doctor-patient interaction is one-way, only supporting patients to transmit information to the medical end, while the medical end cannot proactively and accurately push personalized health guidance, nor can it respond to patient needs in real time; third, remote management capabilities are insufficient, doctors cannot conduct real-time dynamic monitoring of patients, resulting in the inability to intervene in time when patients have emergencies, and the efficiency of remote follow-up management is low. Summary of the Invention

[0019] The technical problem this invention aims to solve is that the current medical system lacks an intelligent management system specifically for kidney transplantation, which leads to prominent issues such as difficulty in integrating medical data, superficial analysis, and weak doctor-patient collaboration. This severely restricts the accuracy and timeliness of follow-up management for kidney transplant patients and fails to meet the clinical needs for large-sample data mining and remote intelligent management of patients. Therefore, building an intelligent management system adapted to kidney transplantation has become an urgent need in the industry.

[0020] To address the aforementioned technical problems, the present invention discloses a single-disease follow-up database system for kidney transplantation based on a hospital intranet environment. The system is characterized by comprising a data integration layer, an intelligent analysis layer, and a doctor-patient collaboration layer, wherein: The data integration layer, as the underlying data support, is responsible for connecting with the hospital's multi-port medical systems, completing data collection and centralized storage, and transmitting structured / unstructured data to the intelligent analysis layer in real time through API interfaces; The intelligent analysis layer serves as the core processing unit, equipped with an AI algorithm engine. It receives data transmitted from the data integration layer, completes cleaning, integration, and analysis, and pushes the analysis results to the PC client of the data integration layer and the mobile app of the doctor-patient linkage layer. Through the AI-driven data cleaning-integration-analysis-early warning process, it realizes intelligent interpretation and risk prediction of kidney transplant patient data. The doctor-patient collaboration layer serves as an interactive terminal, enabling visualization of doctor-patient data and two-way interaction through a mobile app. Health data collected from the patient is encrypted and transmitted to the intelligent analysis layer for processing, while the doctor's treatment suggestions and responses reach the patient in real time through this layer.

[0021] Preferably, the data integration layer includes: The data acquisition interface interfaces with various target ports to convert data output from different systems into a unified format, thereby eliminating data format barriers. The data classification, collection, and structured processing module uses OCR and NLP technologies to transform unstructured data into searchable structured information while preserving the original data format, thus meeting data analysis needs and ensuring data traceability. The cloud-based centralized data storage and encryption protection module elevates data storage risk from "device-level" to "cloud-based professional protection level," while access control ensures the controllability of data access, complying with the requirements of the "Medical Data Security Guidelines."

[0022] Preferably, the data acquisition interface supports HL7 and DICOM protocols.

[0023] Preferably, the data classification, acquisition, and structuring module classifies the acquired data into "structured data" and "unstructured data": For structured data: Field information is directly extracted through the data acquisition interface and stored in a MySQL relational database, and an associated index of "Patient ID-Data Type-Collection Time-Source Port" is established; For unstructured data: OCR technology is used to extract text content, and semantic segmentation and keyword annotation are performed through the NLP module. The text information is stored in a MongoDB non-relational database, and the image data is compressed and encrypted before being stored on a cloud server. At the same time, it is associated with the corresponding patient ID and structured data index.

[0024] Preferably, the cloud-based centralized data storage and encryption protection module transmits all data integrated by the data classification, collection, and structured processing module to the cloud-based monitoring center. Data transmission is protected by SSL / TLS encryption, and data storage is protected by AES-256 encryption. Role-based access control is also implemented: doctors can only view data of patients under their care, administrators have only system configuration permissions, and data access logs are retained in real time for traceability.

[0025] Preferably, the intelligent analysis layer includes: The multimodal data cleaning and integration algorithm module employs a rule-based and machine learning-based data cleaning model to handle missing values, outliers, and duplicate values ​​in the integrated data. Specifically: For missing values: for critical data, a "fill-in with the mean of data from patients with the same disease at the same time + manual confirmation by doctors" approach is used; for non-critical data, it is marked as "missing data" and retained. For outliers: data deviating from the normal range is identified using the Z-score algorithm and marked as "abnormalities to be verified," and true abnormal data is screened in conjunction with clinical rules. For duplicate values: duplicate data is automatically deleted based on a triple deduplication rule of "patient ID + data type + collection timestamp."

[0026] The AI ​​analysis model specifically designed for kidney transplantation is a fusion analysis model based on gradient boosting tree and recurrent neural network. It captures the time-series features of patient data through the RNN model and combines the isolated forest algorithm to achieve real-time early warning of abnormal data. The warning threshold is adapted to different stages after kidney transplantation. The analysis results push module ensures that doctors receive key risk information first and patients receive concise and easy-to-understand prompts through tiered push notifications, achieving rapid access from "abnormal data to doctors to patients" and avoiding delays in processing.

[0027] Preferably, the kidney transplant-specific AI analysis model supports large-sample cluster analysis to uncover the diagnosis and treatment patterns of different subgroups of patients, providing data support for clinical research, and has the following three main analytical functions: Patient full-cycle health report generation: The model integrates patient medical history, treatment chain, laboratory tests, medication records, comorbidities and other data to automatically generate a structured health report, including "condition change trend chart", "medication effect assessment" and "complication risk score"; Large-sample population characteristic analysis: Cluster analysis is performed on large-sample data stored in the cloud to uncover the diagnosis and treatment patterns of different subgroups of patients. The results can support clinical research and optimization of treatment guidelines. Real-time abnormal data early warning: The abnormal detection model is trained based on the isolated forest algorithm. The patient's real-time test indicators and vital signs data are input, and a kidney transplant-specific early warning threshold is set. When the data triggers the threshold, the model automatically marks the risk level.

[0028] Preferably, after the AI ​​analysis is completed by the kidney transplant-specific AI analysis model, the analysis result push module pushes the AI ​​analysis results through the following path: Doctor's side: The PC client displays a pop-up alert, simultaneously showing details of abnormal data, risk level, and preliminary treatment suggestions; health reports are automatically stored in the patient's electronic file, and doctors can edit and supplement them online; Patient side: Warning notifications are pushed through the mobile app of the doctor-patient collaboration layer, and a simplified version of the health report is updated simultaneously to the mobile app.

[0029] Preferably, the doctor-patient linkage layer includes: The patient-side dual-channel health data collection module guides patients to report data in a standardized manner through standardized templates, while also using AI to automatically capture and supplement passive data, ensuring the comprehensiveness and timeliness of patients' home health data and providing data support for remote management. In the mobile app, the patient-side dual-channel health data collection module provides "active reporting + AI automatic capture" dual-channel collection functionality. Proactive reporting channel: The mobile app has a built-in standardized data collection template, allowing patients to manually enter symptoms, signs, medication information, and upload photos of symptoms. The system automatically generates a "patient health log". AI-automated data capture channel: Access to health-related data from the patient's mobile phone through mobile app authorization, combined with abnormal test data pushed by the intelligent analysis layer, AI automatically captures key information and adds it to the health log without requiring manual operation by the patient; The AI-powered intelligent interaction and tiered response module uses AI to handle routine consultations, freeing up doctors' time; and through tiered response, it ensures that critical situations are handled first, achieving an efficient model of "AI solving routine problems, doctors intervening in complex problems, and rapid response to emergencies." The patient data visualization and collaborative management module integrates data through visualization, enabling patients to clearly understand their own health status and improve compliance; it also allows doctors to quickly grasp the dynamics of patients throughout their entire life cycle, assisting in diagnosis and treatment decisions, and achieving collaborative management between doctors and patients based on the same set of data.

[0030] The patient data visualization and collaborative management module utilizes both mobile app and PC client to synchronize data visualization functions, including: Patient side: Key indicator changes, health logs, and doctor's advice are displayed in chart form, providing an intuitive view of one's own health status; On the doctor's end: The PC client integrates patient data from all terminals with home data collected by the mobile app to generate a "360° view of the patient," which includes treatment history, real-time indicators, risk warnings, health logs, etc. It supports doctors to annotate treatment opinions online and update them synchronously to the patient's mobile app.

[0031] Preferably, the AI ​​intelligent interaction and hierarchical response module is implemented via a mobile app that integrates an NLP-based AI interaction module, thereby achieving: Intelligent response to patient inquiries: Patients input their inquiries through the App, and the AI ​​module performs semantic understanding based on the kidney transplant clinical knowledge base to generate standardized responses; for complex questions that cannot be answered, the system automatically marks them as "requiring a doctor's reply" and prompts the patient to wait for the doctor's processing. Data anomaly tiered response: Based on the risk level pushed by the intelligent analysis layer, the AI ​​module initiates a tiered response, including: For mild risks: The mobile app automatically pushes health advice without the need for immediate doctor intervention; For moderate risk: The mobile app pushes notifications and automatically sends reminders to the doctor's PC client. The doctor can quickly reply to the patient through the app, adjust the medication or arrange a follow-up examination. For severe risks: The mobile app immediately pushes an emergency alert, simultaneously triggering dual reminders on the doctor's PC client and mobile SMS. Doctors can initiate real-time voice / video calls through the mobile app to assess the patient's condition and, if necessary, guide the patient to seek emergency medical treatment.

[0032] The technical solution disclosed in this invention can solve the following technical problems existing in the prior art: 1) Solve the problem of integrating medical data from multiple ports for a single disease: Integrate the diagnosis and treatment information of kidney transplant patients through information technology, and integrate medical data from more than ten ports such as outpatient system, inpatient system, emergency system, HIS system, electronic medical record, laboratory system, imaging system, prescription system, medical order system, and surgical system into a unified PC client for centralized management, and build a kidney transplant big data cloud monitoring center. 2) Solving the problem of large-sample multimodal medical data analysis: Through AI technology, the integrated multimodal medical data is cleaned, integrated and analyzed to form a complete health report and intelligent analysis for kidney transplant patients, including medical history, diagnosis and treatment chain, comorbidities and complications, test results, examinations, medications, etc. Abnormal medical data of patients are reported in real time and intelligently and accurately sent to the responsible doctor and the patient for timely processing; 3) Solving the problem of barriers to doctor-patient collaboration: Establish an AI doctor-patient collaboration system through the App to realize the visualization of medical data for both doctors and patients, enable patients to actively report symptoms, signs and discomfort and realize the dual-channel data collection function of AI automatic capture, realize the AI ​​intelligent response and linkage function of the medical end, report the abnormal status of patients in real time and intelligently, and realize the remote, real-time and intelligent management of patients.

[0033] This invention, through the collaborative operation of its designed three-layer technical architecture, achieves a comprehensive solution to the three core problems of existing technologies, and has the following technical effects: 1) Data integration level: Enables automatic and real-time integration of multi-port medical data of kidney transplant patients, improving data integration efficiency by 90%, accuracy rate to 99.5%, and significantly improving traceability and privacy security; 2) Data analysis level: AI-driven analysis models enable automatic generation of full-cycle health reports for patients, large-sample population feature mining, and real-time anomaly warning, reducing doctors' data analysis workload by 80% and shortening the response time for abnormal data to the minute level; 3) At the level of doctor-patient collaboration: Dual-channel data collection and AI intelligent interaction enable remote management of patients throughout the entire life cycle, improve doctor-patient communication efficiency by 70%, shorten the intervention time for severe risks to within 10 minutes, and completely solve the pain points of single-disease management of kidney transplantation.

[0034] Compared with the prior art solutions pointed out in the background section, the advantages of the present invention are shown in the table below.

[0035]

[0036] As shown in the table above, existing technical solutions lack specialized architecture design and intelligent technical support for kidney transplantation as a single disease, which fails to meet the clinical needs for efficient data management, accurate analysis, and real-time doctor-patient interaction. In contrast, this invention, through its innovative technical architecture, precisely addresses the core deficiencies of existing solutions and possesses significant technical advantages and application value. Attached Figure Description

[0037] Figure 1 The schematic diagram illustrates the overall architecture of the present invention; Figure 2 The structure of the data integration layer is illustrated; Figure 3 The structure of the intelligent analysis layer is illustrated. Figure 4 This illustrates the structure of the doctor-patient collaboration layer. Detailed Implementation

[0038] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0039] The purpose of this invention is to address the difficulties in integrating multi-port medical data for a single disease, analyzing large-sample multimodal medical data, and the lack of efficient medical communication technologies that traditional technologies cannot solve. By using a new technical architecture, functional design, and technical solution, an AI-based doctor-patient linkage system for kidney transplantation is constructed to achieve patient medical data integration, intelligent analysis, and efficient doctor-patient linkage, thereby solving this pain point in the management of kidney transplantation as a single disease.

[0040] To address these issues, this invention discloses a single-disease follow-up database system for kidney transplant patients based on a hospital intranet environment. This system addresses the core problems of data integration difficulties, superficial analysis, and weak doctor-patient collaboration in existing single-disease management of kidney transplant patients. It constructs a three-layer technical architecture: a data integration layer, an intelligent analysis layer, and a doctor-patient collaboration layer. Through information technology, AI algorithms, and mobile application development, it achieves unified integration, intelligent analysis, and efficient doctor-patient collaboration of kidney transplant patient medical data. The following section, in conjunction with the technical architecture and the functions of each module, elaborates on the implementation logic, core technical features, and corresponding technical effects of the technical solution.

[0041] The overall technical architecture of this invention is based on cloud + terminal + AI engine, and is divided into three layers: data integration layer, intelligent analysis layer, and doctor-patient linkage layer. Each layer independently implements specific functions, and at the same time, through bidirectional interaction via data interface, a complete closed loop for intelligent management of kidney transplantation as a single disease is formed.

[0042] 1) Data Integration Layer: As the underlying data support, it is responsible for connecting with the hospital's multi-port medical systems, completing data collection and centralized storage, and transmitting structured / unstructured data in real time with the intelligent analysis layer through API interfaces.

[0043] In a preferred embodiment of the present invention, the core hardware used in the data integration layer includes a cloud server (Alibaba Cloud / Huawei Cloud), a hospital local server, and a PC client, while the core software / technology used includes a data integration middleware, a database management system (MySQL+MongoDB), and a data encryption module.

[0044] Addressing the issues of "scattered data across multiple ports and low integration efficiency" in existing technologies, the data integration layer utilizes "interface standardization + data classification and collection + centralized cloud storage" to achieve automatic and efficient integration of medical data from more than ten ports for kidney transplant patients. Specifically: To address the issue of data incompatibility caused by inconsistent interface protocols across systems in existing technologies, the data integration layer provides a unified standard data acquisition interface (RESTful API) for over ten target ports, including outpatient systems, inpatient systems, HIS systems, electronic medical records, and laboratory systems. Access to data from each system is granted through authorization from the hospital's information department. This standardized interface converts output data from different systems into JSON format (structured data) and Base64 encoding (unstructured data, such as images), eliminating data format barriers and enabling "one-time connection, continuous data collection," replacing the traditional manual cross-system retrieval method.

[0045] In one preferred embodiment of the present invention, the data acquisition interface (RESTful API) supports HL7 (Medical Information Exchange Standard) and DICOM (Medical Imaging Standard) protocols to ensure compatible transmission of data in different formats.

[0046] By designing a data acquisition interface (RESTful API), the data acquisition cycle has been shortened from "manual integration for 1-3 weeks" to "real-time automatic acquisition," improving integration efficiency by more than 90% while avoiding data omissions and errors caused by manual operation.

[0047] To address the problem that existing technologies cannot effectively integrate unstructured data, resulting in incomplete data dimensions, a data classification, acquisition, and structured processing module was designed in the data integration layer. This module uses OCR+NLP technology to transform unstructured data into searchable structured information while preserving the original data format. This satisfies data analysis needs and ensures data traceability (original images / reports can be retrieved via index).

[0048] The data classification, acquisition, and structuring module classifies the acquired data into "structured data" and "unstructured data." For structured data such as test results, medication records, and surgery times: field information is directly extracted through the data acquisition interface (RESTful API), stored in a MySQL relational database, and an associated index of "patient ID-data type-acquisition time-source port" is established; For unstructured data such as imaging reports, doctors' medical records, and pathological slide images: OCR (Optical Character Recognition) technology is used to extract text content, and NLP module is used for semantic word segmentation and keyword annotation (such as "rejection reaction" and "elevated creatinine"). The text information is stored in a MongoDB non-relational database, and the image data is compressed and encrypted before being stored on a cloud server, while being associated with the corresponding patient ID and structured data index.

[0049] Through the data classification, collection and structured processing module, the data dimension coverage is improved from "structured data only (about 60%)" in the existing technology to "all types of data (100%)", and the data traceability time is shortened from "manual investigation for 4-6 days" to "system retrieval within 10 seconds".

[0050] To address the privacy concerns arising from the dispersed storage of local documents in existing technologies, a cloud-based centralized data storage and encryption protection module was designed within the data integration layer. Through centralized cloud storage and encryption technology, the risk of data storage is elevated from "device-level" to "cloud-based professional protection level." Meanwhile, access control ensures the controllability of data access, complying with the requirements of the "Guidelines for Medical Data Security."

[0051] All data, after being integrated through the data classification, collection, and structured processing modules, is transmitted to a cloud-based centralized data storage and encryption protection module running in the cloud monitoring center (using Alibaba Cloud's dedicated medical cloud, which meets the Level 3 Information Security Protection Standard). The module employs dual protection: "transmission encryption (SSL / TLS protocol) + storage encryption (AES-256 algorithm)". At the same time, role-based access control (RBAC) is set up: doctors can only view the data of patients under their care, administrators only have system configuration permissions, and data access logs are retained in real time for traceability.

[0052] By designing a cloud-based centralized data storage and encryption protection module, the risk of patient privacy data leakage is reduced by more than 95%, data storage stability reaches 99.99%, and data corruption caused by the loss of local documents is avoided.

[0053] The data integration process implemented through the data integration layer includes the following steps: System initialization: Complete the connection and permission configuration of each medical port and data acquisition interface (RESTful API), and set the data acquisition frequency. In this embodiment, a preferred implementation is to collect data in real time: laboratory / emergency data; and collect data at regular intervals: outpatient / inpatient data, once every 30 minutes.

[0054] Automatic data collection: The data classification, collection and structured processing module captures data from each port at a preset frequency through the data collection interface (RESTful API) and automatically converts it into a standard format.

[0055] Classification and processing: Structured data obtained by the data classification, collection and structuring processing module is directly stored in the database and indexed, while unstructured data is processed by OCR+NLP and then stored in association. Data synchronization: The cloud database is synchronized in real time with the doctor's PC client. Doctors can access all patient data with one click through the PC client without switching systems.

[0056] (ii) Intelligent Analysis Layer: As the core processing unit, it is equipped with an AI algorithm engine, receives data transmitted from the data integration layer, completes cleaning, integration, and analysis, and pushes the analysis results (health reports, abnormal warnings) to the PC client (doctor's end) of the data integration layer and the mobile app (both doctor and patient ends) of the doctor-patient linkage layer. Addressing the issues of existing technologies relying on manual data analysis and lacking intelligent warnings, this layer achieves intelligent interpretation and risk prediction of kidney transplant patient data through AI-driven data cleaning-integration-analysis-warning process.

[0057] In one preferred embodiment of the present invention, the core hardware used in the intelligent analysis layer includes an AI computing server (GPU cluster), and the core software / technology used includes a machine learning algorithm library (TensorFlow / PyTorch), a natural language processing (NLP) module, and an anomaly detection model.

[0058] To address the issue of poor data quality and inability to support in-depth analysis due to the lack of standardized cleaning processes in existing technologies, a multimodal data cleaning and integration algorithm module is designed in the intelligent analysis layer. The algorithm automatically processes data noise and combines it with clinical rules for kidney transplantation (such as differences in the normal range of indicators at different postoperative stages) to ensure the clinical applicability of the cleaned data and lay the foundation for subsequent analysis.

[0059] The multimodal data cleaning and integration algorithm module employs a rule-based and machine learning-based data cleaning model to handle missing values, outliers, and duplicate values ​​in the integrated data. For missing values: For critical data such as test indicators, the "fill in the mean of data from patients with the same disease at the same time + manual confirmation by doctors" approach is adopted; for non-critical data such as descriptions of minor symptoms, they are marked as "data missing" and then retained. For outliers: Data that deviates from the normal range (such as immunosuppressant concentration exceeding the clinical threshold) is identified using the Z-score algorithm and marked as "abnormal to be verified". True abnormal data is then screened in conjunction with clinical rules (such as whether it is a short-term fluctuation after medication). For duplicate values: Based on the triple deduplication rule of "patient ID + data type + collection timestamp", duplicate data is automatically deleted.

[0060] By designing a multimodal data cleaning and integration algorithm module, the data cleaning efficiency was improved from "manual processing of 1 patient / 2 hours" to "AI processing of 100 patients / 10 minutes", and the data accuracy was improved from "manual processing of about 85%" to "more than 99.5%".

[0061] To address the issue that existing technologies lack disease-specific analysis models, resulting in generalized but less targeted analysis results, a kidney transplant-specific AI analysis model was designed in the intelligent analysis layer. This model uses clinical data from 50,000 kidney transplant patients as a training set and incorporates kidney transplant clinical pathways (such as postoperative follow-up time points and key monitoring indicators) to make the analysis results more aligned with clinical needs.

[0062] The AI ​​analysis model specifically designed for kidney transplantation is a fusion analysis model based on gradient boosting tree (XGBoost) and recurrent neural network (RNN). The RNN model can capture the time-series characteristics of patient data (such as the dynamic trend of indicator changes), which can more accurately predict changes in the condition compared with traditional static analysis. By capturing the time-series characteristics of data (such as the dynamic changes of serum creatinine) through RNN, combined with the isolated forest algorithm, it can realize real-time early warning of abnormal data. The early warning threshold is adapted to different stages after kidney transplantation (such as triggering an early warning when serum creatinine rises by more than 20% 3 months after the operation).

[0063] The dedicated AI analysis model for kidney transplantation supports large-sample cluster analysis, enabling the discovery of treatment patterns in different subgroups of patients and providing data support for clinical research. Based on the characteristics of kidney transplant patient data, the dedicated AI analysis model can: generate a full-cycle health report including disease trend charts, medication assessments, and complication risk scores; perform cluster analysis on large-sample data to uncover population characteristics; and, based on the isolated forest algorithm and kidney transplantation-specific thresholds, detect abnormal data in real time and classify risk levels. Specifically, it has the following three main analytical functions: 1. Generation of full-cycle health reports for patients: The model integrates data such as patient medical history, treatment chain, laboratory tests, medication records, and comorbidities to automatically generate structured health reports, including "condition change trend graph" (such as the serum creatinine change curve in the past 6 months), "medication effect evaluation" (such as the concentration control effect of a certain immunosuppressant on the patient), and "complication risk score" (calculated based on postoperative time, index fluctuations, etc. to determine the risk of rejection / infection).

[0064] 2. Large-sample population characteristic analysis: Cluster analysis is performed on large-sample data (over 100,000 patient data in total) stored in the cloud to explore the diagnosis and treatment patterns of different subgroups of patients (such as the difference in complications between elderly patients and younger patients, and the comparison of the efficacy of different immunosuppressive regimens). The results can support clinical research and optimization of treatment guidelines.

[0065] 3. Real-time abnormal data early warning: Train an abnormality detection model (based on the isolated forest algorithm), input the patient's real-time test indicators (such as serum creatinine, cyclosporine concentration) and vital signs data (blood pressure, weight), set a kidney transplant-specific early warning threshold (such as a sudden increase in serum creatinine >20% 3 months after surgery), and when the data triggers the threshold, the model automatically marks the risk level (mild / moderate / severe).

[0066] By designing a dedicated AI analysis model for kidney transplantation, the time for doctors to generate patient analysis reports has been shortened from "4 hours / case for manual processing" to "5 minutes / case for automatic AI generation," and the response time for abnormal data has been shortened from "several hours for manual discovery" to "real-time early warning (≤1 minute)." The large-sample analysis capability fills the gap in existing technology and provides data support for clinical research.

[0067] To address the issue that existing technologies lack proactive push mechanisms and rely on doctors manually reviewing data, an analysis result push module is designed within the intelligent analysis layer. Through tiered push notifications, this ensures that doctors receive key risk information first, while patients receive concise and easy-to-understand prompts, achieving rapid access from "abnormal data" to "doctors" and "patients," thus avoiding delays in processing.

[0068] After AI analysis is completed using a kidney transplant-specific AI analysis model, the AI ​​analysis results will be pushed through the following paths: Doctor's side: The PC client displays a pop-up alert, simultaneously showing details of abnormal data, risk level, and preliminary treatment suggestions (generated based on clinical guidelines); health reports are automatically stored in the patient's electronic record, and doctors can edit and supplement them online; For patients: Warning notifications are pushed through the doctor-patient collaboration layer App (only displaying simple prompts that patients can understand, such as "Your immunosuppressant concentration is abnormal, please contact your doctor in time"), and a simplified version of the health report (hiding technical terms and retaining key conclusions and recommendations) is updated to the App simultaneously.

[0069] (iii) Doctor-patient interaction layer: As an interactive terminal, it realizes the visualization of doctor-patient data and two-way interaction through a mobile app. The health data collected by the patient is encrypted and transmitted to the intelligent analysis layer for processing. The doctor's diagnosis and treatment suggestions and responses reach the patient in real time through this layer.

[0070] To address the issues of "one-way communication between doctors and patients and lagging remote management" in existing technologies, the doctor-patient collaboration layer adopts an AI-driven doctor-patient collaboration management solution. It constructs an efficient doctor-patient collaboration system through dual-channel data collection via mobile app, AI intelligent interaction, and real-time linkage response technology.

[0071] To address the issues of existing technologies that only support one-way patient reporting or messaging, resulting in unsystematic and untimely data collection, a dual-channel health data collection module for patients is designed at the doctor-patient collaboration layer. This module guides patients to report data in a standardized manner through standardized templates, while AI automatically captures and supplements passive data to ensure the comprehensiveness and timeliness of patients' home health data, providing data support for remote management.

[0072] To implement a dual-channel health data collection module for patients, a dual-channel data collection function of "proactive reporting + AI automatic capture" was designed in the mobile app: Proactive reporting channel: The mobile app has a built-in standardized data collection template, allowing patients to manually enter symptoms (such as "fatigue" and "decreased urine output"), signs (blood pressure, weight, body temperature, and automatic import via Bluetooth connection to smart devices), medication information (whether a dose was missed, adverse reactions), and upload photos of symptoms (such as skin rashes). The system automatically generates a "patient health log". AI-automated data capture channel: Access to health-related data (such as step count and sleep duration) from the patient's mobile phone through App authorization, combined with abnormal test data pushed by the intelligent analysis layer, AI automatically captures key information (such as "sudden drop in step count in the past 3 days" and "abnormally high serum creatinine") and adds it to the health log without manual operation by the patient.

[0073] By designing a dual-channel health data collection module for patients, the coverage of patient health data collection has been increased from "offline follow-up period (about 10%)" to "full cycle (more than 90%)", and the standardization of data collection has been improved by 80%, avoiding information bias caused by ambiguous patient descriptions.

[0074] To address the technical issues of existing doctor-patient interaction technologies relying on manual responses from doctors, which are inefficient and unable to handle a large volume of inquiries, an AI-powered intelligent interaction and tiered response module is designed at the doctor-patient collaboration layer. AI handles routine inquiries, freeing up doctors' time; tiered response ensures that critical situations are handled first, achieving an efficient model of "AI solving routine problems, doctors intervening in complex problems, and rapid response to emergency problems."

[0075] The AI-powered intelligent interaction and tiered response module integrates an NLP-based AI interaction module into the app, enabling two core functions: First, intelligent response to patient inquiries: Patients can input their inquiries through the App (such as "What should I do if I forgot to take my immunosuppressant today?"). The AI ​​module performs semantic understanding based on the kidney transplant clinical knowledge base (including medication guidelines and frequently asked questions) and generates standardized responses. For complex questions that cannot be answered, the system automatically marks them as "requires a doctor's reply" and prompts the patient to wait for the doctor's processing.

[0076] Second, tiered response to data anomalies: Based on the risk level pushed by the intelligent analysis layer, the AI ​​module initiates a tiered response: For mild risks (such as slight blood pressure fluctuations): the app automatically pushes health advice (such as "Pay attention to rest and monitor blood pressure changes") without the need for immediate doctor intervention; For moderate risk (such as mildly abnormal immunosuppressant concentration): The App pushes a notification and automatically sends a reminder to the doctor's PC client. The doctor can quickly reply to the patient through the App, adjust the medication or arrange a follow-up examination. For severe risks (such as suspected rejection symptoms + abnormal indicators): The App will immediately push an emergency alert, simultaneously triggering dual reminders on the doctor's PC client and mobile SMS. Doctors can initiate real-time voice / video calls through the App to assess the patient's condition and, if necessary, guide the patient to seek emergency medical treatment.

[0077] By designing AI-powered intelligent interaction and tiered response modules, the response time for routine patient consultations has been reduced from "several hours" to "seconds," the average daily time doctors spend handling doctor-patient interactions has been reduced by 60%, and the intervention response time for severe risk situations has been reduced from "several hours" to "within 10 minutes," thus reducing the medical risks associated with remote management.

[0078] To address the problem of fragmented data in existing technologies, which prevents both doctors and patients from intuitively obtaining a complete picture of the data, a doctor-patient data visualization and collaborative management module is designed at the doctor-patient collaboration layer. By visually integrating data, patients can clearly understand their own health status and improve compliance; doctors can quickly grasp the dynamics of patients throughout their entire life cycle, assisting in diagnosis and treatment decisions, and achieving collaborative management between doctors and patients based on the same set of data.

[0079] The patient data visualization and collaborative management module utilizes the synchronization of the App and PC client to achieve data visualization functions, including: For patients: Key indicator changes (such as blood pressure curve over the past month, changes in immunosuppressant concentration), health logs, and doctor's advice are displayed in chart form, providing an intuitive view of their health status.

[0080] Doctor's side: The PC client integrates patient data from all terminals with home data collected by the App to generate a "360° view of the patient", which includes treatment history, real-time indicators, risk warnings, health logs, etc. It supports doctors to annotate treatment opinions online and update them synchronously to the patient's App.

[0081] The aforementioned single-disease follow-up database system for kidney transplantation can be directly deployed in kidney transplantation centers at all levels of hospitals. It connects to existing medical systems through standardized interfaces, without requiring large-scale modifications to hospital infrastructure, and has good compatibility and scalability.

[0082] The following example, using a kidney transplant center of a tertiary hospital as a case study, details the hardware configuration, software deployment, operation process, and implementation effects of the system, verifying the feasibility and practicality of the technical solution of this invention.

[0083] The kidney transplant center of this top-tier hospital performs approximately 300 kidney transplant surgeries annually and currently has over 2,000 kidney transplant patients under follow-up care. Its daily operations involve managing patient data across multiple scenarios, including outpatient, inpatient, and emergency care, and involve various medical systems such as the hospital's outpatient system, inpatient system, HIS system, electronic medical record system, laboratory system (biochemistry / immunology), imaging system (CT / ultrasound), prescription system, and surgical system.

[0084] Prior to implementation, the center faced several issues: data integration relied on manual Excel entry and cross-system retrieval (integrating single-patient data took 2 hours); data analysis depended on physician experience (abnormal indicator detection was delayed by more than 4 hours); and doctor-patient collaboration relied solely on telephone follow-ups (patient home data collection coverage was less than 15%). Based on these issues, the present invention deployed a three-layer architecture system—"data integration layer - intelligent analysis layer - doctor-patient collaboration layer"—to achieve intelligent management of kidney transplantation as a single disease.

[0085] The hardware configuration is shown in the table below.

[0086] The software deployment is shown in the table below.

[0087] Taking the follow-up management process of a kidney transplant patient (ID: PT-2024001) 3 months post-surgery at the center as an example, the collaborative operation logic of each level of the system is illustrated: Step 1: Automated integration of multi-port medical data (data integration layer) (1) Data collection trigger: When the patient arrives at the hospital on the same day for a routine postoperative outpatient follow-up, completes tests such as blood routine, serum creatinine, and immunosuppressant concentration (cyclosporine), takes kidney ultrasound images, and the doctor completes the outpatient medical record, the system automatically triggers data collection.

[0088] (2) Multi-port data connection: The data integration middleware uses a standardized API interface to capture medical records from the outpatient system, disease descriptions from the electronic medical record system, test indicators (serum creatinine 89μmol / L, cyclosporine concentration 180ng / mL) from the laboratory system, and ultrasound reports and image data from the imaging system in real time.

[0089] (3) Data classification and processing: - Structured data (test indicators, consultation time, medication records) are directly linked to the patient ID (PT-2024001) and stored in a MySQL database, creating an index of "patient ID-test item-collection time-source port".

[0090] Unstructured data (ultrasound report text, doctor's medical record) is extracted into text using OCR, and the NLP module identifies keywords (such as "normal blood flow in transplanted kidney" and "no obvious signs of rejection") and stores it in the MongoDB database; ultrasound images are encrypted using Base64 encoding and stored on the cloud server, linked to the patient ID and test data index.

[0091] (4) Data synchronization: The integrated data is synchronized to the doctor's PC client and the cloud monitoring center in real time. Doctors can open the client to view the full data of the patient's current follow-up examination without switching systems.

[0092] Step 2: AI Intelligent Analysis and Anomaly Warning (Intelligent Analysis Layer) (1) Data cleaning: After receiving the integrated data, the AI ​​algorithm engine processes it through the built-in cleaning model: - The patient's current serum creatinine level (89 μmol / L) was found to be different from the previous level (75 μmol / L 2 months post-surgery), but it was within the normal range (53-106 μmol / L) 3 months post-kidney transplantation and was therefore considered a normal fluctuation, requiring no abnormality marker.

[0093] - The cyclosporine concentration (180 ng / mL) was verified to be within the therapeutic window (150-250 ng / mL), and the data were correct.

[0094] (2) Full-cycle analysis: The fusion analysis model integrates all data from the patient within 3 months post-surgery to generate a health report. - Disease trend chart: Displays the dynamic change curves of key indicators such as serum creatinine and cyclosporine concentration, and marks the fluctuation of indicators at each time point after surgery.

[0095] - Medication assessment: Based on the stable cyclosporine concentration within the therapeutic window, it was determined that "the current immunosuppressive regimen is effective and no dose adjustment is required".

[0096] - Complication risk score: Combined with the patient's age, postoperative recovery status and index fluctuations, a rejection risk score of 0.3 (maximum score 1.0, <0.5 is low risk) is calculated.

[0097] (3) Results push: Health reports are automatically stored in the patient's electronic records and pushed to the doctor's PC client (full professional version) and the patient's App (simplified version, including the prompt "The indicators of this re-examination are normal, continue to take the medication according to the original plan").

[0098] Step 3: Patient Home Management and Doctor-Patient Collaboration (Doctor-Patient Collaboration Layer) (1) Patient data collection: After returning home, patients manage their home health through the App: - Proactive reporting: Enter your blood pressure (120 / 80 mmHg) and weight (65 kg) daily through the App, and upload your medication check-in record ("I have taken cyclosporine today").

[0099] - AI-automated data collection: After app authorization, the system automatically collects the patient's daily step count (5000 steps) and sleep duration (7 hours), and combines this with cloud-synchronized test data to generate a daily health log.

[0100] (2) Handling of sudden abnormalities: On the 10th day after surgery, the patient experienced fatigue. He entered "fatigue, slightly reduced urine output" in the "symptom report" module of the App and uploaded a photo of his urine.

[0101] - AI preliminary assessment: The app's built-in NLP module identifies symptom keywords and, combined with the patient's post-operative time (3 months), preliminarily determines "signs of mild rejection need to be monitored" and marks it as "moderate risk".

[0102] - Tiered response trigger: The system immediately pushes an early warning to the doctor's PC client (including symptom details and recent changes in the patient's indicators), and sends an SMS reminder to the responsible doctor; after the doctor opens the PC client to view the patient's full-cycle data, he initiates a voice call to the patient through the App to guide him to come to the hospital for a follow-up examination the next day.

[0103] (3) Linkage of follow-up results: The patient returned to the hospital for a follow-up examination the next day. The serum creatinine level rose to 130 μmol / L. After the system integrated the data, the AI ​​analysis determined that there was a "mild rejection reaction" and pushed it to the doctor's end. After the doctor adjusted the immunosuppressive regimen, the medication adjustment suggestions were synchronized to the patient's end through the App. The patient could view the adjusted medication list and precautions.

[0104] The table below shows the results of three months of operation after the kidney transplant center of this hospital implemented the technical solution of this invention, fully verifying the practicality and advanced nature of the solution:

[0105] This embodiment verifies the feasibility of the invention's technical solution through actual deployment in a tertiary hospital's kidney transplant center. The system, through standardized data integration, kidney transplant-specific AI analysis, and AI-driven doctor-patient collaboration, thoroughly addresses the core pain points of existing technologies, significantly improving the efficiency and accuracy of kidney transplant patient management. Furthermore, the solution is deployed based on existing hospital infrastructure, requiring no large-scale modifications, and possesses good compatibility and scalability, making it widely applicable to kidney transplant centers at all levels of hospitals and other single-disease medical management scenarios.

[0106] The three core problems that this invention needs to solve have alternative solutions with different technical approaches, as detailed below: 1. Alternative solutions to the problem of "multi-port data integration for a single disease". Alternative Solution 1: Customized development based on the hospital's existing data platform -Technical approach: Relying on the general medical data platform already built in some large hospitals, a customized kidney transplant single-disease data access module is developed to screen and extract scattered outpatient, inpatient, and laboratory data within the platform to form a dedicated dataset, which is then synchronized to the doctor's PC client.

[0107] - Implementation logic: Utilize the existing data platform interface resources of the hospital, without the need to develop separate multi-port standardized interfaces. Only screening rules and structured processing algorithms need to be developed for the data characteristics of kidney transplantation diseases (such as key indicators and data sources).

[0108] Alternative Option 2: Procurement of Third-Party Medical Data Integration Tools -Technical approach: Purchase mature third-party medical data integration tools (such as the data integration module of IBM Watson Health, or the integration kits of domestic medical information vendors), connect to the hospital's multi-port system through the tool's built-in interface adaptation capabilities to complete data collection and integration, and then develop a secondary development function for the data classification and storage of kidney transplant patients based on the tool.

[0109] 2. Alternative solutions for "large sample multimodal data analysis problems" Alternative Solution 1: Disease-based matching based on a general medical AI analysis platform -Technical approach: Use a general medical AI analysis platform (such as Tencent AI Medical Imaging or Alibaba Health AI Medical Platform) to import the integrated kidney transplant patient data according to the platform's data format requirements. Based on the platform's existing data analysis models (such as general anomaly detection models and chronic disease management analysis models), data cleaning, analysis, and early warning are achieved by adjusting model parameters (such as adapting to the threshold of key indicators for kidney transplantation).

[0110] Alternative Option 2: Development of Physician-Led Semi-Automated Data Analysis Tools -Technical Approach: Develop lightweight, semi-automated analysis tools (such as Python-based data analysis plugins) and integrate them into doctors' PC clients. The tool has built-in basic data cleaning functions (such as duplicate value removal and missing value marking). Doctors manually upload data files and select analysis dimensions (such as "creatinine changes in the past 3 months" and "immunosuppressant concentration trends"). The tool automatically generates basic statistical charts, and doctors make judgments on abnormal data based on their clinical experience.

[0111] 3. Alternative solutions to the problem of "barriers in doctor-patient communication". Alternative Solution 1: Upgrade the functionality of the existing general hospital follow-up app -Technical approach: Add a kidney transplant-specific module to the general patient follow-up app already deployed in the hospital (such as a follow-up system covering multiple departments), develop symptom / sign entry templates and key indicator visualization functions, and upgrade the message push mechanism (such as allowing doctors to manually mark abnormal data and push reminders) to achieve basic doctor-patient linkage.

[0112] Alternative Option 2: Customized Partnership with a Third-Party Health Management App - Technical approach: Collaborate with established third-party health management apps (such as Ping An Health and Chunyu Doctor) to customize exclusive service modules for kidney transplant patients, and open up some data interfaces of the hospital (such as test results and medication records) to the app. Patients can view data and submit their health status through the app, and doctors can receive information and reply through the app's backend. This linkage is achieved by leveraging the user base and communication capabilities of third-party platforms.

[0113] While the alternative solutions can partially achieve the objectives of the invention, they differ significantly from the present invention in terms of core performance, adaptability, and security. A detailed comparison is shown in the table below:

[0114] In summary, none of the alternative solutions can fully realize the core value of this invention. The fundamental reason is that the alternative solutions are all based on "secondary modification of a general system" or "single-function splicing," lacking a "full-process integrated architecture design" specifically for kidney transplantation. This is reflected in the following three points: The disease-specific core technologies of this invention are irreplaceable: The kidney transplant-specific AI analysis model (trained with data from 5,000 cases) and single-disease data integration rules (covering multiple interfaces and unstructured data processing) are features that general-purpose systems / tools cannot provide. General-purpose systems can only achieve "universal functions" and cannot accurately match the diagnostic and treatment characteristics and follow-up needs of kidney transplant patients.

[0115] The three-layer architecture collaborative closed loop cannot be replaced: The "data integration-intelligent analysis-doctor-patient linkage" three-layer architecture of this invention forms a closed loop of two-way data flow (such as patient App data → intelligent analysis → doctor's warning → doctor's reply → patient App synchronization). The alternative solutions are mostly "independent module splicing" (such as the data platform and general AI platform needing manual data transmission, and the AI ​​analysis results not being automatically linked with the App), which cannot achieve full-process automation and real-time performance.

[0116] The medical-grade security and adaptability are irreplaceable: This invention complies with medical data security standards (Level 3 Information Security Protection, AES-256 encryption) from the initial architecture design stage, and can be directly connected to existing systems of hospitals at all levels without large-scale modification; the general systems of alternative solutions are difficult to meet the requirements of medical scenarios in terms of permission management and data encryption, while third-party platforms pose risks to data ownership and privacy compliance.

[0117] In summary, while alternative solutions exist for achieving the objectives of this invention, such as general system modifications and third-party tool procurement, these solutions only partially address the existing technical problems and cannot match the effectiveness of this invention in terms of disease adaptability, intelligence level, linkage efficiency, and security. This invention, through its innovative approach of "single-disease-specific design + three-layer closed-loop architecture + medical-grade security," forms an irreplaceable technological advantage and represents the optimal technical solution for addressing the pain points of intelligent management of single-disease kidney transplantation.

Claims

1. A single-disease follow-up database system for kidney transplantation based on a hospital intranet environment, characterized in that, The system comprises a data integration layer, an intelligent analysis layer, and a doctor-patient interaction layer. The data integration layer, serving as the underlying data support, connects to the hospital's multi-port medical systems, completing data collection and centralized storage. It transmits structured and unstructured data to the intelligent analysis layer in real time via API interfaces. The intelligent analysis layer, the core processing unit, is equipped with an AI algorithm engine. It receives data from the data integration layer, cleans, integrates, and analyzes it, pushing the analysis results to the PC client of the data integration layer and the mobile app of the doctor-patient interaction layer. Through AI-driven data cleaning, integration, analysis, and early warning processes, it enables intelligent interpretation and risk prediction of kidney transplant patient data. The doctor-patient interaction layer, as the interactive terminal, enables visualization and two-way interaction of doctor-patient data via a mobile app. Health data collected from the patient is encrypted and transmitted to the intelligent analysis layer for processing. Doctors' treatment suggestions and responses are delivered to patients in real time through this layer.

2. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 1, characterized in that, The data integration layer includes: a data acquisition interface that interfaces with various target ports to convert data output from different systems into a unified format, eliminating data format barriers; a data classification, acquisition, and structuring module that uses OCR and NLP technologies to transform unstructured data into searchable structured information while preserving the original data format, meeting data analysis needs and ensuring data traceability; and a cloud-based centralized data storage and encryption protection module that elevates data storage risk from "device-level" to "cloud-based professional protection level," while access control ensures the controllability of data access, complying with the requirements of the "Medical Data Security Guidelines." 3. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 2, characterized in that, The data acquisition interface supports HL7 and DICOM protocols.

4. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 2, characterized in that, The data classification, acquisition, and structuring processing module classifies the acquired data into "structured data" and "unstructured data": For structured data, field information is directly extracted through the data acquisition interface and stored in a MySQL relational database, establishing an association index of "patient ID-data type-acquisition time-source port"; For unstructured data, OCR technology is used to extract text content, and semantic segmentation and keyword annotation are performed through an NLP module. The text information is stored in a MongoDB non-relational database, and image data is compressed and encrypted before being stored on a cloud server, while simultaneously associating the corresponding patient ID with the structured data index.

5. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 2, characterized in that, The cloud-based centralized data storage and encryption protection module transmits all data integrated by the data classification, collection, and structured processing module to the cloud-based monitoring center. Data transmission is protected by SSL / TLS encryption, and data storage is protected by AES-256 encryption. Role-based access control is also implemented: doctors can only view data of patients under their care, administrators have only system configuration permissions, and data access logs are retained in real time for traceability.

6. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 1, characterized in that, The intelligent analysis layer includes: a multimodal data cleaning and integration algorithm module, which employs a data cleaning model combining rules and machine learning to process missing values, outliers, and duplicate values ​​in the integrated data. Specifically: for missing values: for critical data, a "fill-in with the mean of data from patients with the same disease at the same time + manual confirmation by doctors" approach is used; for non-critical data, it is marked as "missing data" and retained. For outliers: data deviating from the normal range is identified using the Z-score algorithm and marked as "abnormal to be verified," and true abnormal data is filtered using clinical rules. For duplicate values: duplicate data is automatically deleted based on a triple deduplication rule of "patient ID + data type + collection timestamp." A kidney transplant-specific AI analysis model is a fusion analysis model based on gradient boosting trees and recurrent neural networks. It captures the time-series features of patient data using an RNN model and combines it with an isolated forest algorithm to achieve real-time early warning of abnormal data, with warning thresholds adapted to different stages after kidney transplantation. An analysis result push module ensures that doctors receive key risk information first, and patients receive concise and easy-to-understand prompts through tiered push notifications, achieving rapid access from "abnormal data - doctor - patient" and avoiding delays in processing.

7. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 6, characterized in that, The kidney transplant-specific AI analysis model supports large-sample cluster analysis to uncover treatment patterns in different subgroups of patients, providing data support for clinical research. It has three main analytical functions:

1. Generation of full-cycle patient health reports: The model integrates patient medical history, treatment chain, laboratory tests, medication records, comorbidities, and other data to automatically generate structured health reports, including "condition change trend charts," "medication efficacy assessments," and "complication risk scores." 2. Large-sample population characteristic analysis: Cluster analysis is performed on large-sample data stored in the cloud to uncover treatment patterns in different subgroups of patients. The results can support clinical research and optimization of treatment guidelines. Real-time abnormal data early warning: The abnormal detection model is trained based on the isolated forest algorithm. The patient's real-time test indicators and vital signs data are input, and a kidney transplant-specific early warning threshold is set. When the data triggers the threshold, the model automatically marks the risk level.

8. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 6, characterized in that, After the AI ​​analysis is completed by the kidney transplant-specific AI analysis model, the analysis result push module pushes the AI ​​analysis results through the following path: Doctor's end: The PC client pops up an alert prompt, and displays the details of abnormal data, risk level and preliminary treatment suggestions at the same time; Health reports are automatically stored in the patient's electronic records and can be edited and supplemented online by doctors; on the patient's end: alert notifications are pushed through the mobile app of the doctor-patient linkage layer, and a simplified version of the health report is updated synchronously to the mobile app.

9. A single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 1, characterized in that, The patient-doctor collaboration layer includes: a dual-channel health data collection module for patients, which guides patients to report data in a standardized manner through standardized templates, while automatically capturing and supplementing passive data with AI to ensure the comprehensiveness and timeliness of patients' home health data, providing data support for remote management. In the mobile app, the dual-channel health data collection module for patients provides a dual-channel collection function of "active reporting + AI automatic capture": Active reporting channel: The mobile app has a built-in standardized data collection template, allowing patients to manually enter symptoms, signs, medication information, and upload symptom photos; the system automatically generates a "patient health log"; AI automatic capture channel: Accessing health-related data from the patient's mobile phone through authorization via the mobile app, combined with abnormal test results pushed by the intelligent analysis layer, AI automatically captures key information and supplements it to the health log, without requiring manual operation by the patient; AI intelligent interaction and tiered response module: Handling routine consultations with AI, freeing up doctors' time; Through a tiered response system, critical situations are prioritized for handling, achieving an efficient model of "AI solving routine problems, doctor intervention for complex problems, and rapid response to emergencies." The patient data visualization and collaborative management module integrates data through visualization, allowing patients to clearly understand their health status and improving compliance. It enables doctors to quickly grasp the patient's dynamics throughout the entire process, assisting in treatment decisions and achieving collaborative management between doctors and patients based on the same set of data. The patient data visualization and collaborative management module utilizes mobile apps and PC clients to synchronously realize data visualization functions, including: Patient side: displaying key indicator changes, health logs, and doctor suggestions in chart form, intuitively presenting their own health status; Doctor side: the PC client integrates patient data from all terminals with home data collected by the mobile app to generate a "patient 360° view," including treatment history, real-time indicators, risk warnings, health logs, etc., supporting doctors to annotate treatment opinions online, which are synchronously updated to the patient's mobile app.

10. The single-disease follow-up database system for kidney transplantation based on a hospital intranet environment as described in claim 1, characterized in that, The AI ​​intelligent interaction and hierarchical response module is equipped with an NLP-based AI interaction module through a mobile app, enabling: intelligent response to patient inquiries: patients input their inquiries through the app, and the AI ​​module performs semantic understanding based on the kidney transplant clinical knowledge base to generate standardized responses; For complex questions that cannot be answered, the system automatically marks them as "requires a doctor's reply" and prompts the patient to wait for the doctor's intervention. Data anomaly tiered response: Based on the risk level pushed by the intelligent analysis layer, the AI ​​module initiates a tiered response, including: For mild risks: The mobile app automatically pushes health advice without immediate doctor intervention; for moderate risks: The mobile app pushes alerts and automatically sends reminders to the doctor's PC client, allowing the doctor to quickly respond to the patient, adjust medication, or schedule follow-up appointments; for severe risks: The mobile app immediately pushes emergency alerts, simultaneously triggering dual reminders on the doctor's PC client and via SMS, allowing the doctor to initiate real-time voice / video calls to assess the patient's condition and, if necessary, guide the patient to seek emergency medical attention.

Images