Double-disease ai risk prediction system and method based on multi-source data fusion

Abstract

The application discloses a kind of double disease AI risk prediction system and method based on multi-source data fusion, belong to artificial intelligence medical risk assessment technical field;The system includes: data acquisition module, data preprocessing module, double disease parallel prediction module, explainability attribution module and output module, data preprocessing module is connected with data acquisition module, double disease parallel prediction module is connected with data preprocessing module, explainability attribution module is connected with double disease parallel prediction module, output module is connected with double disease parallel prediction module;The application completes the identification, typing, short-term crisis early warning and long-term prognosis prediction of two core diseases in parallel by unified architecture, and outputs the explainable risk attribution result, solves the problem that ACS and AHF evaluation system are independent in prior art, static linear model cannot capture time evolution characteristics, single disease AI model lacks double disease parallel reasoning ability and explainability is insufficient.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence medical risk assessment technology, and in particular to a dual-disease AI risk prediction system and method based on multi-source data fusion. Background Technology

[0002] Cardiovascular emergencies are the leading cause of sudden hospital death, intensive care unit transfer, and short-term mortality. Among these, acute coronary syndrome (ACS) and acute heart failure (AHF) are the two most common, highest-risk, and fastest-changing core diseases. In emergency and inpatient settings, they often present with overlapping symptoms such as chest pain and dyspnea, making clinical differentiation difficult. Misdiagnosis, missed diagnosis, and inaccurate risk stratification have long been problems.

[0003] Currently, clinical assessment of ACS mainly relies on scoring systems such as GRACE and TIMI, while assessment of AHF uses scoring systems such as ADHERE and EHF (e.g., MAGGIC, EHMRG30-ST). These scoring systems are independent, their data are not interoperable, and their models are incompatible. Physicians must collect indicators separately, score multiple times, and interpret data multiple times, resulting in fragmented processes and low efficiency. Furthermore, these scoring systems are all static linear models, reflecting only a single point of risk at the moment of admission. They cannot utilize dynamic time-series data such as heart rate, blood pressure, blood oxygen, myocardial markers, renal function, and respiratory status to capture the trend of disease progression, and their ability to identify occult deterioration and subclinical high-risk states is insufficient.

[0004] Existing medical AI models are mostly built for single diseases, lacking a unified architecture that simultaneously covers both ACS and AHF with intelligent prediction. Furthermore, most models are "black box" models, lacking risk attribution and interpretability, making it difficult to meet the actual clinical requirements for reliable AI. Currently, there is no system, domestically or internationally, capable of simultaneously achieving intelligent management of the entire process of ACS and AHF identification, automatic classification, short-term in-hospital crisis early warning, and long-term prognosis prediction.

[0005] In summary, there is an urgent clinical need for an intelligent risk prediction technology for cardiovascular emergencies and critical illnesses that employs a unified architecture, parallel reasoning for both diseases, and an interpretable attribution mechanism, in order to achieve simultaneous intelligent assessment of the two core diseases, acute coronary syndrome (ACS) and acute heart failure (AHF). Summary of the Invention

[0006] The technical problem to be solved by this invention is: addressing the issues in existing technologies where the ACS and AHF assessment systems are independent, static linear models cannot capture temporal evolution characteristics, and single-disease AI models lack parallel reasoning capabilities for dual diseases and have insufficient interpretability. This invention provides a dual-disease AI risk prediction system and method based on multi-source data fusion, which completes the identification, classification, short-term crisis warning, and long-term prognosis prediction of two core diseases in parallel through a unified architecture, and outputs interpretable risk attribution results.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a dual-disease AI risk prediction system based on multi-source data fusion, comprising: The data acquisition module is used to simultaneously acquire and output multi-source time-series clinical data of patients; A data preprocessing module, connected to the data acquisition module, is used to preprocess the multi-source time-series clinical data acquired by the data acquisition module to obtain shared feature data for both diseases. A dual-disease parallel prediction module is connected to the data preprocessing module; the dual-disease parallel prediction module includes an ACS prediction branch and an AHF prediction branch, the ACS prediction branch and the AHF prediction branch respectively execute their respective prediction tasks in parallel based on the shared feature data of the two diseases, and output ACS prediction results and AHF prediction results; An interpretability attribution module, connected to the dual-disease parallel prediction module, is used to output the corresponding risk contribution weights for the ACS prediction results and the AHF prediction results respectively, so as to realize independent attribution interpretation of the dual-disease prediction results. And an output module, connected to the dual-disease parallel prediction module, is used to generate dual-disease risk reports and perform graded early warning pushes based on the ACS prediction results and the AHF prediction results using a unified risk stratification standard.

[0008] Compared with the prior art, the advantages of the present invention are as follows: (I) Unified architecture to break the fragmentation of the assessment system. This invention collects multi-source time-series clinical data of ACS and AHF in a unified manner through a data acquisition module, and realizes dual-branch parallel inference under the same architecture through a dual-disease parallel prediction module. This changes the current situation where multiple independent scoring tools need to be used in clinical practice, and solves the problem of fragmentation of the assessment system.

[0009] (II) Dynamic temporal modeling to accurately capture the evolution trend of the disease. This invention uses a data preprocessing module to perform temporal alignment of multi-source time-series clinical data, and a dual-disease parallel prediction module to perform dynamic modeling based on temporal features, overcoming the shortcomings of traditional static linear scoring models that cannot capture temporal fluctuation features.

[0010] (III) Parallel reasoning for dual diseases to achieve integrated intelligent assessment. This invention utilizes the ACS prediction branch and AHF prediction branch in the dual-disease parallel prediction module to execute their respective prediction tasks in parallel based on the same shared feature data of the two diseases, thereby completing the prediction and assessment of the two diseases under the same architecture.

[0011] (iv) Interpretable attribution, enabling source tracing and interpretation of prediction results. This invention outputs the corresponding risk contribution weights for ACS prediction results and AHF prediction results respectively through an interpretable attribution module, enabling independent source tracing and interpretation of the prediction results for both diseases.

[0012] (v) Unified risk stratification to adapt to clinical decision-making in multiple scenarios. This invention generates dual-disease risk reports and provides tiered early warning pushes through the output module using a unified risk stratification standard.

[0013] Furthermore, the ACS prediction branch and the AHF prediction branch are implemented based on the Transformer backbone network.

[0014] Furthermore, the ACS prediction branch is used to perform the following tasks: screening for cardiogenic chest pain, identifying true and false ACS, distinguishing STEMI / NSTEMI / unstable angina subtypes, predicting short-term in-hospital risk, and predicting long-term MACE events; the AHF prediction branch is used to perform the following tasks: differentiating between cardiogenic / pulmonary dyspnea, classifying HFrEF / HFmrEF / HFpEF heart failure phenotypes, providing early warning of in-hospital crisis, and predicting long-term readmission / mortality risk.

[0015] Furthermore, the in-hospital short-term risk prediction includes the risk of sudden death, the risk of malignant arrhythmia, and the risk of cardiogenic shock, while the long-term MACE event prediction includes the prediction of MACE events 1 year after surgery, the prediction of re-infarction, and the prediction of all-cause mortality.

[0016] Furthermore, the in-hospital crisis early warning includes respiratory failure early warning, acute kidney injury early warning, cardiogenic shock early warning, ICU transfer early warning, and in-hospital death early warning; the long-term readmission / mortality risk prediction includes readmission prediction 3 months after discharge, readmission prediction 6 months after discharge, and all-cause mortality prediction.

[0017] Furthermore, the multi-source time-series clinical data includes at least two of the following data types: demographic information, medical history and risk factors, dynamic time-series vital signs data, biochemical biomarker data, high-dimensional electrocardiogram data, cardiac structural parameter data, and diagnosis and treatment time-series data.

[0018] Specifically, the dynamic time-series vital sign data includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, and urine output data; the biochemical marker data includes myocardial markers, renal function indicators, inflammatory markers, electrolytes, lactate, BNP / NT-proBNP, and troponin data; the high-dimensional electrocardiogram feature data includes 12-lead electrocardiogram features and heart rhythm variability features; the cardiac structural parameter data includes echocardiographic EF value and cardiac function classification data; and the diagnosis and treatment time-series data includes medication time-series records.

[0019] Furthermore, the preprocessing includes temporal alignment, missing value imputation, noise removal, standardization and normalization, high-dimensional feature screening, and collinearity removal.

[0020] Specifically, the temporal alignment is used to align clinical data from different sources and with different sampling frequencies to a unified time axis; the missing value imputation is used to adaptively impute missing clinical data during the data acquisition process; the noise removal is used to filter out abnormal fluctuations during the clinical data acquisition process; the standardization and normalization is used to convert clinical data of different dimensions to a unified scale; the high-dimensional feature screening is used to extract prediction-related features from the clinical data; and the collinearity removal is used to remove redundant information between clinical data.

[0021] Furthermore, the unified risk stratification standard divides risks into five levels: extremely low risk, low risk, medium risk, high risk, and extremely high risk, which is applicable to unified emergency triage, monitoring level adjustment, and critical care transfer decisions for dual diseases.

[0022] Specifically, "very low risk" refers to patients whose current condition is stable and the risk of adverse events in the short term is extremely low; "low risk" refers to patients with minor risks who can maintain their current level of monitoring; "intermediate risk" refers to patients whose condition may worsen and require increased monitoring frequency; "high risk" refers to patients whose condition is unstable and who may experience a crisis in the short term, requiring consideration of ICU transfer or escalation of treatment; and "very high risk" refers to patients who are in a pre-crisis state or whose condition continues to deteriorate, requiring immediate clinical intervention.

[0023] Furthermore, the system is deployed on the hospital's intranet and connects to the HIS / LIS / ECG / monitoring system to realize pre-hospital emergency care, emergency triage, cardiology inpatient care, CCU intensive care, and discharge follow-up. It covers the entire scenario from pre-hospital care to follow-up, providing intelligent closed-loop risk management for dual diseases and covering the complete management chain of cardiovascular acute and critical illnesses from screening, treatment, monitoring to long-term prognosis.

[0024] Secondly, this invention also provides a dual-disease AI risk prediction method based on multi-source data fusion, employing the aforementioned system and including the following steps: Step 1: Obtain multi-source time-series clinical data of patients; Step 2: Preprocess the multi-source time-series clinical data to obtain shared feature data for both diseases; Step 3: Input the shared feature data of the two diseases into the parallel prediction module for the two diseases, and execute their respective prediction tasks in parallel through the ACS prediction branch and the AHF prediction branch, respectively, and output the ACS prediction results and the AHF prediction results. Step 4: Generate corresponding risk attribution explanations for the ACS prediction results and the AHF prediction results respectively; and generate dual-disease risk reports based on the ACS prediction results and the AHF prediction results using a unified risk stratification standard and push out graded early warnings. Attached Figure Description

[0025] Figure 1 This is a block diagram of the overall architecture of the dual-disease AI risk prediction system based on multi-source data fusion of the present invention; Figure 2 This is a flowchart of the dual-disease AI risk prediction method based on multi-source data fusion according to the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the invention and not to limit it. Those skilled in the art will recognize that the invention can be practiced without some of these specific details. The following description of the embodiments is merely to provide a better understanding of the invention by illustrating examples of it.

[0027] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0028] Example 1: This example provides a dual-disease AI risk prediction system based on multi-source data fusion. Figure 1 The overall architecture block diagram of the system in this embodiment is shown.

[0029] like Figure 1 As shown, the system in this embodiment includes a data acquisition module, a data preprocessing module, a dual-disease parallel prediction module, an interpretability attribution module, and an output module.

[0030] The data acquisition module is used to simultaneously collect multi-source time-series clinical data from patients. This multi-source time-series clinical data includes demographic information, past medical history and risk factors, dynamic time-series vital signs data, biochemical biomarker data, high-dimensional electrocardiogram (ECG) characteristic data, cardiac structural parameter data, and treatment time-series data. Specifically, dynamic time-series vital signs data includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, and urine output; biochemical biomarker data includes cardiac biomarkers, renal function indicators, inflammatory markers, electrolytes, lactate, BNP / NT-proBNP, and troponin data; high-dimensional ECG characteristic data includes 12-lead ECG characteristics and heart rate variability; cardiac structural parameter data includes echocardiographic ejection fraction (EF) values ​​and cardiac function classification; and treatment time-series data includes medication administration time-series records.

[0031] The data preprocessing module is connected to the data acquisition module and is used to preprocess the multi-source time-series clinical data acquired by the data acquisition module. Preprocessing includes time alignment, missing value imputation, noise removal, standardization and normalization, high-dimensional feature selection, and collinearity removal. Time alignment aligns clinical data from different sources and with different sampling frequencies to a unified time axis; missing value imputation adaptively interpolates missing clinical data during data acquisition; noise removal filters out abnormal fluctuations during clinical data acquisition; standardization and normalization converts clinical data of different dimensions to a unified scale; high-dimensional feature selection extracts prediction-related features from the clinical data; and collinearity removal removes redundant information between different clinical data sets.

[0032] The dual-disease parallel prediction module is connected to the data preprocessing module. The dual-disease parallel prediction module includes an ACS prediction branch and an AHF prediction branch, both implemented using a Transformer backbone network. The ACS and AHF prediction branches execute their respective prediction tasks in parallel based on shared feature data from both diseases, outputting ACS and AHF prediction results respectively.

[0033] The ACS prediction branch is used to perform the following tasks: screening for cardiogenic chest pain, differentiating between true and false ACS, identifying STEMI / NSTEMI / unstable angina subtypes, predicting short-term in-hospital risks, and predicting long-term MACE events. Short-term in-hospital risk prediction includes the risk of sudden cardiac death, malignant arrhythmias, and cardiogenic shock. Long-term MACE event prediction includes predicting MACE events 1 year post-surgery, predicting re-infarction, and predicting all-cause mortality. The AHF prediction branch is used to perform the following tasks: differentiating between cardiogenic and pulmonary dyspnea, classifying heart failure phenotypes (HFrEF / HFmrEF / HFpEF), providing early warning of in-hospital crises, and predicting long-term readmission / mortality risks. In-hospital crisis early warning includes warnings of respiratory failure, acute kidney injury, cardiogenic shock, ICU transfer, and in-hospital mortality. Long-term readmission / mortality risk prediction includes predictions of readmission 3 months after discharge, readmission 6 months after discharge, and predicting all-cause mortality.

[0034] The interpretability attribution module is connected to the dual-disease parallel prediction module to output the corresponding risk contribution weights for the ACS prediction results and AHF prediction results respectively, so as to realize independent attribution interpretation of the dual-disease prediction results.

[0035] The output module connects to the dual-disease parallel prediction module to generate dual-disease risk reports and provide tiered early warnings based on ACS and AHF prediction results using a unified risk stratification standard. The unified risk stratification standard classifies risks into five levels: very low, low, intermediate, high, and very high risk, applicable to unified emergency triage, monitoring level adjustment, and critical care transfer decisions for dual-disease cases.

[0036] The system in this embodiment is deployed on the hospital's intranet and connects to the HIS / LIS / ECG / monitoring system to achieve intelligent closed-loop risk management for both diseases across all scenarios, including pre-hospital emergency care, emergency triage, cardiology inpatient care, CCU intensive care, and discharge follow-up.

[0037] Example 2: This example provides a dual-disease AI risk prediction method based on multi-source data fusion, using the system of Example 1. Figure 2 The flowchart of the method in this embodiment is shown.

[0038] like Figure 2 As shown, the method in this embodiment includes the following steps: Step 1: Acquire the patient's multi-source time-series clinical data through the data acquisition module. This multi-source time-series clinical data includes demographic information, past medical history and risk factors, dynamic time-series vital signs data, biochemical biomarker data, high-dimensional ECG data, cardiac structural parameters data, and treatment time-series data. Specifically, dynamic time-series vital signs data includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, and urine output; biochemical biomarker data includes cardiac biomarkers, renal function indicators, inflammatory markers, electrolytes, lactate, BNP / NT-proBNP, and troponin data; high-dimensional ECG data includes 12-lead ECG characteristics and heart rate variability; cardiac structural parameters data includes echocardiographic ejection fraction (EF) and cardiac function classification; and treatment time-series data includes medication administration records.

[0039] Step 2: The data preprocessing module performs time-series alignment, missing value imputation, noise removal, standardization and normalization, high-dimensional feature screening, and collinearity removal on the multi-source time-series clinical data to obtain shared feature data for both diseases. Specifically, time-series alignment aligns clinical data from different sources and sampling frequencies to a unified time axis; missing value imputation adaptively fills in missing clinical data during data acquisition; noise removal filters out abnormal fluctuations during clinical data acquisition; standardization and normalization transforms clinical data of different dimensions to a unified scale; high-dimensional feature screening extracts prediction-related features from the clinical data; and collinearity removal removes redundant information between different clinical data sets.

[0040] Step 3: Input the shared feature data of the two diseases into the parallel prediction module for the two diseases, and execute their respective prediction tasks in parallel through the ACS prediction branch and the AHF prediction branch, and output the ACS prediction results and the AHF prediction results.

[0041] The ACS prediction branch performs the following tasks: screening for cardiogenic chest pain, identifying true and false ACS, determining STEMI / NSTEMI / unstable angina subtypes, predicting short-term in-hospital risks, and predicting long-term MACE events. Short-term in-hospital risk prediction includes the risk of sudden cardiac death, the risk of malignant arrhythmias, and the risk of cardiogenic shock. Long-term MACE event prediction includes predicting MACE events 1 year post-surgery, predicting re-infarction, and predicting all-cause mortality.

[0042] The AHF prediction branch performs the following tasks: differentiation of cardiogenic / pulmonary dyspnea, HFrEF / HFmrEF / HFpEF heart failure phenotype classification, in-hospital crisis warning, and long-term readmission / mortality risk prediction. In-hospital crisis warning includes warnings of respiratory failure, acute kidney injury, cardiogenic shock, ICU transfer, and in-hospital mortality; long-term readmission / mortality risk prediction includes predictions of readmission 3 months after discharge, readmission 6 months after discharge, and all-cause mortality.

[0043] Step 4: The interpretability attribution module generates corresponding risk attribution explanations for the ACS and AHF prediction results respectively; and the output module generates dual-disease risk reports based on the ACS and AHF prediction results using a unified risk stratification standard and sends tiered early warnings. The unified risk stratification standard classifies risks into five levels: very low risk, low risk, intermediate risk, high risk, and very high risk.

[0044] Example 3: This example provides a practical application scenario for a dual-disease AI risk prediction system based on multi-source data fusion.

[0045] In emergency triage scenarios, when a patient arrives at the hospital, the system automatically retrieves multi-source time-series clinical data from the HIS / LIS / ECG / monitoring system via the data acquisition module. After processing by the data preprocessing module, the data is input into the dual-disease parallel prediction module. This module simultaneously outputs ACS and AHF prediction results, including the identification, classification, short-term risk, and long-term prognosis predictions for both diseases. The interpretability attribution module outputs the risk contribution weights for ACS and AHF respectively. The output module generates a dual-disease risk report based on a unified five-level risk stratification standard and pushes it to the healthcare terminal to assist emergency physicians in making rapid triage decisions.

[0046] In cardiology wards and CCU intensive care settings, the system continuously monitors patients’ dynamic time-series clinical data, updates risk prediction results and risk stratification levels in real time, and automatically pushes early warning information to medical staff terminals when the risk level rises to high risk or very high risk, reminding medical staff to adjust treatment plans in a timely manner or consider ICU transfer.

[0047] In the scenario of discharge follow-up, the system predicts the risk of readmission and all-cause mortality 3 months and 6 months after discharge based on the patient's clinical data during hospitalization and the patient's condition at discharge, providing a basis for developing individualized follow-up plans.

Claims

1. A dual-disease AI risk prediction system based on multi-source data fusion, characterized in that, include: The data acquisition module is used to simultaneously collect multi-source time-series clinical data from patients; The data preprocessing module is used to preprocess the multi-source time-series clinical data to obtain shared feature data for the two diseases; The dual-disease parallel prediction module includes an ACS prediction branch and an AHF prediction branch. The ACS prediction branch and the AHF prediction branch respectively execute their respective prediction tasks in parallel based on the shared feature data of the two diseases, and output ACS prediction results and AHF prediction results. The interpretability attribution module is used to output the corresponding risk contribution weights for the ACS prediction results and the AHF prediction results respectively, so as to realize the independent source attribution interpretation of the prediction results of the two diseases. And an output module, used to generate dual-disease risk reports and push graded early warnings based on the ACS prediction results and the AHF prediction results using a unified risk stratification standard.

2. The system according to claim 1, characterized in that, The ACS prediction branch and the AHF prediction branch are implemented based on the Transformer backbone network.

3. The system according to claim 1, characterized in that, The ACS prediction branch is used to perform the following tasks: screening for cardiogenic chest pain, identifying true and false ACS, distinguishing STEMI / NSTEMI / unstable angina subtypes, predicting short-term in-hospital risk, and predicting long-term MACE events; the AHF prediction branch is used to perform the following tasks: differentiating between cardiogenic and pulmonary dyspnea, classifying HFrEF / HFmrEF / HFpEF heart failure phenotypes, providing early warning of in-hospital crisis, and predicting long-term readmission / mortality risk.

4. The system according to claim 3, characterized in that, The in-hospital short-term risk prediction includes the risk of sudden death, the risk of malignant arrhythmia, and the risk of cardiogenic shock. The long-term MACE event prediction includes the prediction of MACE events 1 year after surgery, the prediction of re-infarction, and the prediction of all-cause mortality.

5. The system according to claim 3, characterized in that, The in-hospital crisis early warning includes respiratory failure early warning, acute kidney injury early warning, cardiogenic shock early warning, ICU transfer early warning, and in-hospital death early warning; the long-term readmission / mortality risk prediction includes readmission prediction 3 months after discharge, readmission prediction 6 months after discharge, and all-cause mortality prediction.

6. The system according to claim 1, characterized in that, The multi-source time-series clinical data includes at least two of the following data types: demographic information, medical history and risk factors, dynamic time-series vital signs data, biochemical biomarker data, high-dimensional electrocardiogram (ECG) characteristic data, cardiac structural parameter data, and treatment time-series data. The dynamic time-series vital signs data includes heart rate, blood pressure, blood oxygen saturation, respiratory rate, and urine output data. The biochemical biomarker data includes myocardial biomarkers, renal function indicators, inflammatory markers, electrolytes, lactate, BNP / NT-proBNP, and troponin data. The high-dimensional ECG characteristic data includes 12-lead ECG characteristics and arrhythmia characteristics. The cardiac structural parameter data includes echocardiographic EF values ​​and cardiac function classification data. The treatment time-series data includes medication time-series records.

7. The system according to claim 1, characterized in that, The preprocessing includes temporal alignment, missing value imputation, noise removal, standardization and normalization, high-dimensional feature screening, and collinearity removal.

8. The system according to claim 1, characterized in that, The unified risk stratification standard classifies risks into five levels: very low risk, low risk, medium risk, high risk, and very high risk. It is applicable to unified emergency triage, monitoring level adjustment, and critical care transfer decisions for dual diseases.

9. The system according to claim 1, characterized in that, The system is deployed on the hospital's intranet and connects to the HIS / LIS / ECG / monitoring system to achieve intelligent closed-loop risk management for both diseases across all scenarios, including pre-hospital emergency care, emergency triage, cardiology inpatient care, CCU intensive care, and discharge follow-up.

10. A dual-disease AI risk prediction method based on multi-source data fusion, characterized in that, The system according to any one of claims 1 to 9 comprises the following steps: Step 1: Obtain multi-source time-series clinical data of patients; Step 2: Preprocess the multi-source time-series clinical data to obtain shared feature data for both diseases; Step 3: Input the shared feature data of the two diseases into the parallel prediction module for the two diseases, and execute their respective prediction tasks in parallel through the ACS prediction branch and the AHF prediction branch, respectively, and output the ACS prediction results and the AHF prediction results. Step 4: Generate corresponding risk attribution explanations for the ACS prediction results and the AHF prediction results respectively; and generate dual-disease risk reports based on the ACS prediction results and the AHF prediction results using a unified risk stratification standard and push out graded early warnings.

Images