Cardiovascular acute and severe case rapid diagnosis model establishment method
By establishing a rapid diagnostic model for multimodal data integration, using deep convolutional neural networks and other machine learning models to diagnose cardiovascular acute diseases, solving the problems of lack of data and insufficient algorithm accuracy in the existing technology, and achieving more accurate diagnosis and higher prediction capabilities.
Patent Information
- Application Number
- CN202411899872.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-23
- Publication Date
- 2025-04-15
AI Technical Summary
The prior art faces the problems of lack of data and insufficient algorithm accuracy in the diagnosis of acute cardiovascular diseases, resulting in insufficient diagnosis and limited prediction capabilities.
By establishing a rapid diagnostic model integrating multimodal data, using three-dimensional deep convolutional neural networks and other machine learning models, biochemical data, CT data, electrocardiogram data, etc., combined with transcriptome sequencing and non-targeted metabolomics analysis, the diagnostic-related features are screened out, and a multi-model machine learning model is established to improve the accuracy and prediction ability of diagnosis.
It has achieved the integration of multi-dimensional and multi-modal data models for cardiovascular acute and severe cases, improved the accuracy and prediction ability of diagnosis, helped doctors to formulate more effective treatment plans, and reduced the mortality rate of patients.
Smart Images
Figure CN119833151A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical monitoring technology, and in particular to a method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases. Background Art
[0002] Cardiovascular disease (CVD) is the leading cause of death worldwide, especially acute cardiovascular emergencies such as acute myocardial infarction, which lead to a large number of deaths. The number of deaths from CVD is expected to increase further by 2030. To meet this challenge, timely and accurate diagnosis and prediction are essential, and machine learning technology has shown great potential in the early detection and diagnosis of cardiovascular emergencies. By analyzing electrocardiogram (ECG) signals through machine learning, abnormalities of cardiovascular emergencies can be accurately identified, but current research faces the challenges of lack of data and insufficient algorithm accuracy. To this end, it is necessary to combine multimodal data and multi-center databases to develop more comprehensive and accurate AI models to improve diagnostic capabilities and help doctors develop treatment plans, thereby reducing mortality.
[0003] In addition to ECG-based diagnosis, detection based on cardiac biomarkers is also a research focus for myocardial infarction diagnosis. For example, the University of Augsburg in Germany studied 92 inflammatory biomarkers and found their important role in acute myocardial infarction and mortality. The University of Science and Technology of China has developed a combined strategy combining biomarkers and machine learning, which is superior to traditional clinical testing. Shandong University has proposed a new method to promote early detection and prognosis evaluation of AMI by improving the detection sensitivity of C5 protein.
[0004] Although there has been progress in the detection and prevention of acute myocardial infarction, most of the research uses a single data source or database, which leads to insufficient data diversity and inability to fully utilize the multimodal physiological indicators of patients. In addition, the combined use of electrocardiogram (ECG) and biomarkers has not been widely used in the prevention of acute myocardial infarction, and the existing models are relatively simple.
[0005] To this end, the present invention aims to provide a method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases to solve the above-mentioned problems. Summary of the invention
[0006] The purpose of the present invention is to solve the above problems and provide a method for establishing a rapid diagnosis model for cardiovascular emergencies and severe diseases. Through a three-dimensional deep convolutional neural network and other machine learning models, biochemical data, CT data, electrocardiogram data, gene sequencing and other related data are vectorized, and image features and clinical data features related to the diagnosis of cardiovascular emergencies and severe diseases are found and modeled. The quantified metabolites, gene expression, clinical information, and electrocardiogram / CT image features are summarized, and multiple models are further used for feature screening. The model is established through multi-model machine learning to explore whether the accuracy, specificity and sensitivity of the integrated multi-dimensional and multi-modal data model are more accurate than those of a single modality, and to determine the final diagnosis and prediction prognosis model.
[0007] In order to achieve the above object, the technical solution of the present invention is as follows:
[0008] The present invention provides a method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases, the method comprising the following steps:
[0009] S1. Establish a clinical database of cardiovascular emergencies: Based on the clinical electronic medical records of Jiangxi Provincial People's Hospital, systematically extract and collect cases of cardiovascular emergencies. With the informed consent of patients and their families, collect detailed clinical histories, CT imaging data at various stages of disease development, disease outcomes and their prognosis of approximately 500 patients with cardiovascular emergencies. In addition, in order to conduct a control study, it is planned to collect CT imaging data and basic health information of 50 to 100 healthy people according to gender, age and other factors. Screen and correct records with missing or erroneous information in clinical medical records to ensure the accuracy and reliability of the data for subsequent analysis and research.
[0010] S2. Collect and analyze samples of cardiovascular emergencies: extract biological samples of patients with cardiovascular emergencies in the hospital and monocytes PBMC, serum and plasma of patients' peripheral blood as the experimental group from the database, and extract biological samples of healthy volunteers and monocytes PBMC, serum and plasma of patients' peripheral blood as the control group;
[0011] After transcriptome sequencing of peripheral blood PBMCs in different groups, the deconvolution method was used to further determine whether the composition ratio of PBMCs and differential cell populations were associated with the characteristics of different diseases based on quality control, differential genes, enrichment pathways and GSEA analysis. Non-targeted metabolomics sequencing was performed on the serum of different groups. Based on quality control, principal component analysis, differential metabolites and related pathway enrichment analysis, the Gene-Metabolites analysis method of MetaboAnalyst was further used to conduct a joint analysis of differential genes and metabolites at the pathway level to determine the association between the acute transcriptional and metabolic phenotype changes of different cardiovascular emergencies and the characteristic attributes of different diseases.
[0012] S3. Establish a rapid diagnosis model for cardiovascular emergencies and severe diseases:
[0013] The rapid diagnosis model for acute and severe cardiovascular diseases is as follows: using electrocardiogram and / or imaging to confirm the diagnosis of acute and severe cardiovascular disease patients; selecting the top 200 differentially expressed genes based on the results of transcriptome sequencing analysis and non-targeted metabolomics sequencing analysis, performing feature screening, and determining the final diagnosis and prognosis prediction model;
[0014] Or use a three-dimensional deep convolutional neural network to vectorize the imaging data of critically ill cardiovascular patients, extract image features, and determine the final diagnosis and prognosis prediction model.
[0015] The final diagnosis of patients with acute cardiovascular disease was determined using gold standards (such as electrocardiograms and angiography), and the outcome information was filled into the data grid. Based on the results of transcriptome sequencing analysis and non-targeted metabolomics sequencing analysis, the top 200 differentially expressed genes were selected for feature screening based on multiple machine learning models (including random forests, XGBoost, logistic regression, support vector machines, etc.), and then modeling was performed for metabolome and transcriptome results to find characteristic metabolite-gene combinations. The accuracy, sensitivity, and specificity were further calculated through multi-multiplication orthogonal experiments. 100 newly admitted chest pain patients were selected for model validation, and the accuracy, sensitivity, and specificity of the validation group were further calculated based on their final diagnosis results.
[0016] For the biochemical data, CT data, and electrocardiogram data of 500 patients, it is planned to vectorize the relevant data through three-dimensional deep convolutional neural networks and other machine learning models to find image features and clinical data features related to the diagnosis of critically ill cardiovascular patients. Among them, the vectorized metabolites, gene expression, clinical information, and electrocardiogram / CT image features are summarized and feature screening is performed. The model is established through multi-model machine learning to explore whether the accuracy, specificity, and sensitivity of the multi-dimensional and multi-modal data model integration are accurate, and then the final diagnosis and prognosis prediction model is determined.
[0017] Compared with the prior art, this solution has the following beneficial effects:
[0018] By applying machine learning algorithms, the present invention can accurately detect anomalies in electrocardiogram (ECG) signals, thereby effectively identifying anomalies related to cardiovascular emergencies and severe diseases. By combining multimodal data and multi-center databases for in-depth research, a comprehensive and optimized AI model method and system can be developed that can accurately and quickly diagnose and treat cardiovascular emergencies and severe diseases. This can not only improve the diagnostic accuracy and predictive ability of cardiovascular emergencies and severe diseases, but also help doctors better formulate treatment plans and reduce patients' mortality rates. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 It is a flow chart of establishing a rapid diagnosis model for acute and severe cardiovascular diseases in an embodiment of the present invention. DETAILED DESCRIPTION
[0020] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solution of the present invention will be further described in detail below in conjunction with the embodiments of the present invention and the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0021] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below in conjunction with the embodiments.
[0022] Example:
[0023] 1. Construction of a clinical database for cardiovascular emergencies
[0024] 1.1 Based on the clinical electronic medical records of Jiangxi Provincial People's Hospital, systematically extract and collect cases of cardiovascular emergencies. With the informed consent of patients and their families, collect detailed clinical histories, CT imaging data at various stages of disease development, disease outcomes and their prognosis of approximately 500 patients with cardiovascular emergencies (collection is expected to be completed from 2024 to 2025). In addition, in order to conduct a control study, it is planned to collect CT imaging data and basic health information of 50 to 100 healthy people according to factors such as gender and age.
[0025] 1.2 Screen and correct missing or erroneous records in clinical medical records to ensure the accuracy and reliability of the data for subsequent analysis and research.
[0026] 2. Collection and preliminary testing and analysis of clinical samples of cardiovascular emergencies
[0027] 2.1 In this study, we will conduct detailed biological sample collection on approximately 500 selected patients. This includes extracting mononuclear cells (PBMC) from the peripheral blood of these patients, as well as separating serum and plasma. In addition, 100 healthy volunteers will be added as a normal control group for the needs of the control study.
[0028] 2.2 After transcriptome sequencing of different groups of peripheral blood PBMCs, based on quality control, differential genes, enriched pathways and GSEA analysis, the deconvolution method was used to further determine whether the composition ratio of PBMCs and differential cell populations were related to the characteristics of different diseases.
[0029] 2.3 Non-targeted metabolomics sequencing was performed on the sera of different groups. On the basis of quality control, principal component analysis, differential metabolites and related pathway enrichment analysis, the Gene-Metabolites analysis method of MetaboAnalyst was further used to conduct a joint analysis of differential genes and metabolites at the pathway level to determine the association between the acute transcriptional and metabolic phenotypic changes of different cardiovascular emergencies and the characteristic attributes of different diseases.
[0030] 3. Establishment of AI-based rapid diagnosis model for cardiovascular emergencies
[0031] 3.1 After the final diagnosis of acute cardiovascular patients is determined using the gold standard (such as electrocardiogram, angiography), the outcome information is filled into the data grid. According to the results of transcriptome sequencing analysis and non-targeted metabolomics sequencing analysis, the top 200 differentially expressed genes are selected for feature screening based on multiple machine learning models (including random forest, XGBoost, logistic regression, support vector machine, etc.), and then modeling is performed for metabolome and transcriptome results to find characteristic metabolite-gene combinations. Further, the accuracy, sensitivity and specificity are calculated through multi-multiplication orthogonal experiments. 100 newly admitted chest pain patients are selected for model validation, and the accuracy, sensitivity and specificity of the validation group are further calculated based on their final diagnosis results.
[0032] 3.2 Follow up the prognosis of 500 patients with acute myocardial infarction, group them according to the prognosis of patients with different risk levels, use multiple machine learning methods to develop a prediction model pointing to prognosis, and select 50 patients with different prognoses for each disease for further verification to determine the accuracy, sensitivity and specificity of the model in the test set.
[0033] 3.3 For the biochemical data, CT data, and electrocardiogram data of 500 patients, it is planned to vectorize the relevant data through a three-dimensional deep convolutional neural network and other machine learning models, find the image features and clinical data features related to the diagnosis of critically ill cardiovascular patients, build a multimodal model, and compare it with the model in 3.1.
[0034] 3.4 Summarize the quantized metabolites, gene expression, clinical information, and ECG / CT image features in 3.1 and 3.3, and further use multiple models for feature screening. Through multi-model machine learning, explore whether the accuracy, specificity, and sensitivity of the multi-dimensional and multi-modal data model integration are more accurate than those of a single modality, and determine the final diagnosis and prognosis prediction model.
[0035] 3.5 Apply the diagnosis and prognosis prediction model to build an AI digital doctor system with the diagnosis and prognosis prediction model as the core. The system can realize real-time interactive functions. Patients only need to describe their condition through voice to get instant diagnosis results and professional advice on the prognosis of the condition. Improve the efficiency of medical diagnosis and provide patients with convenient and fast medical services.
[0036] The above specific embodiments are merely explanations of the present invention and are not limitations of the present invention. After reading this specification, those skilled in the art may make modifications to the embodiments without any creative contribution as needed. However, such modifications are protected by the patent law as long as they are within the scope of the claims of the present invention.
Claims
1. A method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases, characterized by: The method comprises the following steps: S1. Establish a clinical database for cardiovascular emergencies and severe diseases; S2. Collect and analyze samples of cardiovascular emergencies; S3. Establish a rapid diagnosis model for acute and severe cardiovascular diseases.
2. A method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases as claimed in claim 1, characterized in that: The S1 step specifically includes: collecting hospital cases and extracting cardiovascular emergency and severe cases to establish a database.
3. A method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases as claimed in claim 1, characterized in that: The step S2 specifically includes: extracting biological samples of hospital patients with acute and severe cardiovascular diseases and monocytes PBMC, serum and plasma of peripheral blood of patients as the experimental group from the database, and extracting biological samples of healthy volunteers and monocytes PBMC, serum and plasma of peripheral blood of patients as the control group; The transcriptome sequencing and analysis of peripheral blood PBMCs in the experimental and control groups were performed, and the non-targeted metabolomics sequencing and analysis of serum were performed.
4. A method for establishing a rapid diagnosis model for acute and severe cardiovascular diseases as claimed in claim 1, characterized in that: The rapid diagnosis model for cardiovascular emergency and severe disease in step S3 is: using electrocardiogram and / or imaging to confirm the diagnosis of cardiovascular emergency and severe disease patients; selecting the top 200 differentially expressed genes based on the transcriptome sequencing analysis results and the non-targeted metabolomics sequencing analysis results, performing feature screening, and determining the final diagnosis and prognosis prediction model; Or use a three-dimensional deep convolutional neural network to vectorize the imaging data of critically ill cardiovascular patients, extract image features, and determine the final diagnosis and prognosis prediction model.