Cardiovascular disease risk assessment method and device, electronic equipment and medium

Abstract

The invention relates to the field of cardiovascular disease monitoring and risk assessment, in particular to a cardiovascular disease risk assessment method which is characterized by comprising the following steps: acquiring dynamic physiological feature data of a user; acquiring static physiological feature data of the user; performing feature extraction on the dynamic physiological feature data, performing normalization processing on the dynamic physiological feature data and the static physiological feature data, and generating structured input data; and inputting the structured input data into the pre-trained prediction model, and generating a cardiovascular disease risk assessment report, so that portable, accurate and easy-to-operate cardiovascular disease monitoring can be realized, the diagnosis cost is reduced, the portability is improved, and personalized cardiovascular disease risk assessment is realized.

Description

Technical Field

[0001] The present application relates to the field of cardiovascular disease monitoring and risk assessment, and in particular to a cardiovascular disease risk assessment method, device, electronic device and medium. Background Art

[0002] Cardiovascular disease (CVD) is a major global public health challenge, with persistently high morbidity and mortality rates. According to statistics, CVD causes approximately 17.9 million deaths annually, accounting for over 31% of all deaths worldwide. This problem is exacerbated by an aging population and changing lifestyles. As a common disease, CVD not only severely impacts patients' quality of life but also places a heavy economic burden on families and society. Early detection and risk assessment are considered key strategies for reducing CVD morbidity and mortality. While traditional medical treatments have achieved some success in treating acute symptoms, invasive procedures, lengthy consultations, and subtle symptoms hinder access to medical care. Early risk assessment and intervention are significantly inadequate, particularly in remote areas and developing countries, where uneven distribution of medical resources prevents many potential patients from receiving timely testing and guidance, leading to worsening of their condition. Therefore, the development of portable, accurate, and user-friendly cardiovascular disease monitoring tools is crucial.

[0003] Currently, there are many non-invasive devices on the market for personal health management, mainly including smart bracelets, watches, and electrocardiogram patches. These devices generally have basic vital sign monitoring functions, such as continuous tracking of heart rate, blood pressure fluctuations, and sleep status. Some high-end models can also capture more detailed physiological data through built-in sensors, such as blood oxygen concentration, stress levels, and even preliminary arrhythmia detection. However, the original design of these products is mostly focused on daily fitness assistance or chronic disease management, and their warning capabilities for complex cardiovascular diseases are limited. In addition, most consumer electronic devices rely on fixed algorithms to interpret the collected information, lacking flexibility, the ability to simultaneously collect multi-dimensional physiological parameters, and the ability to conduct in-depth analysis, making it difficult to meet the needs of professional medical levels.

[0004] The main problem with existing technologies is their inability to provide comprehensive and detailed risk assessment services. For one thing, most commercial health trackers only reflect immediate numerical changes, but fail to integrate historical trends and personal characteristics to provide a comprehensive assessment. Furthermore, even though some advanced devices can perform a certain degree of automated diagnosis, their lack of transparency and "black box" nature lead doctors and patients to question the results, reducing trust and willingness to adopt them. This urgently requires a new solution that can not only ensure the authenticity and timeliness of data collection, but also utilize advanced computational methods to achieve reliable risk quantification, and provide a clear and understandable explanatory framework to facilitate communication between doctors and patients.

[0005] The risk assessment methods in the existing technology have the following defects: they are unable to effectively integrate multi-dimensional physiological data such as dynamic physiological characteristic data and static physiological characteristic data, and the assessment results are incomplete. Especially when processing complex and diverse physiological characteristic data, existing means are difficult to combine historical and current situations while using intelligent AI for efficient model prediction to achieve reliable and comprehensive risk assessment. Summary of the Invention

[0006] In view of the above problems, the present application is proposed to provide a cardiovascular disease risk assessment method, device, electronic device and medium that overcome the above problems or at least partially solve the above problems.

[0007] According to one aspect of the present application, a cardiovascular disease risk assessment method is provided, which is characterized by including: obtaining dynamic physiological characteristic data of a user; obtaining static physiological characteristic data of the user; performing feature extraction on the dynamic physiological characteristic data, normalizing the dynamic physiological characteristic data and the static physiological characteristic data to generate structured input data; inputting the structured input data into a pre-trained prediction model to generate a cardiovascular disease risk assessment report.

[0008] Optionally, in the above method, the dynamic physiological characteristic data include electrocardiogram signals, pulse signals, blood pressure parameters, and blood oxygen saturation parameters, and the static physiological characteristic data include age, gender, chest pain type, blood sugar parameters, and cholesterol values.

[0009] Optionally, in the above method, feature extraction of dynamic physiological characteristic data includes extracting the QRS wave group, heart rate variability characteristics, and ST segment offset of the electrocardiogram signal, and extracting the conduction velocity, reflected wave, and amplitude of the pulse signal.

[0010] Optionally, in the above method, the prediction model is a Catboost prediction model.

[0011] Optionally, in the above method, the pre-training includes: dividing the pre-training data set into a training set and a test set in proportion, and calculating the score of the pre-trained prediction model based on the risk assessment parameters.

[0012] Optionally, in the above method, the risk assessment parameters include AUC, precision, recall and F1 score.

[0013] Optionally, in the above method, a feature contribution report is generated based on the SHAP algorithm.

[0014] According to a second aspect of the present application, a cardiovascular disease risk assessment device is characterized by comprising: The physiological characteristic data acquisition module is used to obtain dynamic physiological characteristic data; the human-computer interaction module is used to obtain the user's static physiological characteristic data; the structured data module is used to extract features from dynamic physiological characteristic data, normalize dynamic physiological characteristic data and static physiological characteristic data, and generate structured input data; the risk assessment module is used to input structured input data into a pre-trained prediction model to generate a cardiovascular disease risk assessment report.

[0015] Optionally, in the above device, the dynamic physiological characteristic data include electrocardiogram signals, pulse signals, blood pressure parameters, and blood oxygen saturation parameters, and the static physiological characteristic data include age, gender, chest pain type, blood sugar parameters, and cholesterol values.

[0016] Optionally, in the above-mentioned device, feature extraction of dynamic physiological characteristic data includes extracting the QRS wave group, heart rate variability characteristics, and ST segment offset of the electrocardiogram signal, and extracting the conduction velocity, reflected wave, and amplitude of the pulse signal.

[0017] Optionally, in the above device, the prediction model is a Catboost prediction model.

[0018] Optionally, in the above device, the pre-training includes: dividing the pre-training data set into a training set and a test set in proportion, and calculating the score of the pre-trained prediction model based on the risk assessment parameters.

[0019] Optionally, in the above device, the risk assessment parameters include AUC, precision, recall and F1 score.

[0020] Optionally, in the above device, a feature contribution report is generated based on the SHAP algorithm.

[0021] According to the third aspect of the present application, an electronic device is provided, characterized in that it includes a processor and a memory, the processor is coupled to the memory, and the processor is used to execute a computer program stored in the memory so that the electronic device performs the method according to any one of claims 1 to 5.

[0022] According to a fourth aspect of the present application, a computer-readable storage medium is provided, characterized in that it includes a computer program or instructions, which, when the computer program or instructions are run on a computer, enables the computer to execute the method according to any one of claims 1 to 5.

[0023] The beneficial effects of this application are: combining artificial intelligence with portable, non-invasive, user-friendly and low-cost monitoring devices, and performing computational assessments of cardiovascular diseases through predictive model algorithms, can reduce diagnostic costs, improve portability, improve the efficiency of medical resource utilization, reduce the disease burden caused by delayed treatment, promote the development of personalized medicine, and provide decision support for doctors and patients by providing scientific predictions and risk explanations, thereby improving the reliability and comprehensiveness of risk assessments.

[0024] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below and explained in detail with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic diagram of a cardiovascular disease risk assessment method according to an embodiment of the present application is shown.

[0026] Figure 2 A schematic diagram of the evaluation process of a cardiovascular disease risk assessment model according to an embodiment of the present application is shown.

[0027] Figure 3 A schematic structural diagram of a cardiovascular disease risk assessment device according to an embodiment of the present application is shown.

[0028] Figure 4 A schematic structural diagram of an electronic device according to an embodiment of the present application is shown.

[0029] Figure 5 A schematic structural diagram of a computer-readable storage medium according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, rather than all of the embodiments. It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0031] The concept of this application is to develop a non-invasive, user-friendly, and low-cost cardiovascular disease monitoring tool that combines artificial intelligence and portable hardware devices for early risk assessment and intervention, in order to improve the level of early diagnosis, improve treatment outcomes, and reduce misdiagnosis related to cardiovascular disease. It solves the high cost and inconvenience of traditional diagnostic methods through non-invasive means, improves the efficiency of medical resource utilization, reduces the disease burden caused by delayed treatment, and promotes the development of personalized medicine by providing scientific predictions and risk explanations, providing decision support for doctors and patients. This application achieves the effect of high-precision risk assessment through the collaborative work of multiple sensors and the combination of predictive model algorithms.

[0032] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0033] Figure 1 A cardiovascular disease risk assessment system according to an embodiment of the present application is shown, characterized in that it includes: Step S101: Acquire dynamic physiological characteristic data of the user; Cardiovascular disease is one of the most lethal diseases worldwide, with complex pathological mechanisms and significant dynamic changes in the disease state. Traditional diagnostic and treatment methods rely on discrete, single-dimensional physiological parameters, such as occasional monitoring of blood pressure and resting electrocardiograms. Such monitoring methods are difficult to capture early pathological signals and are also unable to obtain real-time disease fluctuations. For example, critical information such as transient episodes of myocardial ischemia and abnormal nocturnal blood pressure fluctuations can easily be missed, resulting in delayed intervention methods or inaccurate intervention plans. Therefore, the convenient and rapid acquisition of multimodal physiological data has become an effective means of monitoring cardiovascular disease.

[0034] Dynamic physiological data includes but is not limited to electrocardiogram signals, pulse signals, blood pressure parameters, blood oxygen saturation parameters, microcirculation parameters, and fatigue status. By acquiring dynamic physiological data, the user's physical condition can be recorded in real time.

[0035] ECG signals are acquired by detecting the potential difference created by cardiac electrical activity on the body surface. For example, this can be achieved through a standard 12-lead system, using 10 motors to cover the heart's three-dimensional electrical activity and a right-leg drive circuit to eliminate common-mode interference. Alternatively, a portable 3-lead recording system can be used. ECG signal acquisition can also be achieved using a portable ECG acquisition circuit. Pulse signals are acquired by detecting changes in pressure or volume caused by blood flow within blood vessels. This can be achieved through contact acquisition methods such as photoplethysmography, piezoelectric sensors, and pressure sensors, or through non-contact acquisition technologies such as wrist PPG, microwave sensing, and Doppler ultrasound for contactless monitoring. Blood pressure parameters can be monitored using cuff oscillometric methods, camera-based remote PPG, or pulse ECG signal-based inference. Blood oxygen saturation parameters can be monitored using contact photoelectric monitoring, such as transmissive or reflective blood oxygen detection, or through non-contact optical detection, such as camera-based remote PPG. By capturing changes in facial / hand skin color, the RGB signal is extracted and then monitored using algorithms. Microcirculatory parameters, including capillary blood flow, vascular density, blood flow velocity, and vascular permeability, can be monitored through spectral imaging and Doppler flowmetry. Fatigue status and other physiological parameters can be accurately assessed and predicted using multimodal sensors combined with AI algorithms.

[0036] Step S102: Obtain the user's static physiological characteristic data.

[0037] Static physiological data includes but is not limited to age, gender, chest pain type, blood sugar parameters, cholesterol values, work and rest patterns, medical history, and main symptoms. The user's age can be divided into age groups, chest pain types include typical angina, atypical angina, non-cardiac chest pain, etc. Blood sugar parameters include fasting blood sugar before meals, postprandial blood sugar, static blood sugar, etc. Static physiological characteristic data also includes the user's medical history and main symptoms. These are all static physiological characteristic data of the user. By acquiring static physiological data, the user's previous physical data can be recorded, thereby ensuring the multi-dimensional collection of the user's physiological characteristics and ensuring the multimodality of the data.

[0038] Step S103: extracting features from the dynamic physiological feature data, normalizing the dynamic physiological feature data and the static physiological feature data, and generating structured input data.

[0039] Feature extraction refers to extracting static data values from physiological signals. For example, it normalizes parameters such as electrocardiogram (ECG) signals, pulse signals, blood pressure parameters, blood oxygen saturation parameters, age, gender, chest pain type, blood sugar parameters, and cholesterol values. By establishing a structured model, physiological signals are aggregated into a structured data model to form a related structured dataset. This allows for the collection of user physiological data from multiple dimensions and the acquisition of a multimodal dataset. The more comprehensive the collection of user physiological data, the more scientific, multimodal, and multidimensional cardiovascular disease prediction results can be obtained based on this basic physiological data.

[0040] Feature extraction uses algorithms to extract information from the ECG signal, including the QRS complex, heart rate variability, ST segment deviation, pulse velocity, reflected waves, amplitude, maximum and minimum blood pressure values, oxygen saturation parameters, age range, chest pain type, fasting and postprandial blood glucose levels, and cholesterol levels. These extracted features undergo preliminary filtering, amplification, and noise reduction to generate dynamic and static physiological characteristic data for signal analysis. For example, fast Fourier transforms can be used to analyze the signal in the frequency domain, while wavelet transforms can be used for time-frequency analysis. The normalization module normalizes the dynamic and static physiological characteristic data to eliminate dimensional differences. Normalization can use methods such as min-max normalization or Z-score normalization, depending on the data distribution.

[0041] Step S104: Input the structured input data into the pre-trained prediction model to generate a cardiovascular disease risk assessment report.

[0042] After the prediction model is trained on a specific dataset, structured data is input into the prediction model to generate prediction results. Prediction models include logistic regression, support vector machine (SVM), K-nearest neighbor (KNN), deep neural network (DNN), random forest, decision tree, XGBoost, CatBoost, LightGBM, gradient boosting models, etc. The dataset can be selected from the Cleveland Clinic database for cardiovascular disease diagnosis. During pre-training, a variety of training methods can be selected to improve training results. For example, a symmetric tree structure can be selected to generate a decision tree, and overfitting can be suppressed by dynamically permuting the sample order in each round of gradient boosting iteration. GPU-accelerated computing can also be selected, and target optimization can be achieved through a combination of loss functions.

[0043] By inputting structured data into a pre-trained prediction model, the prediction model can give prediction results for the current structured data based on the pre-trained data set, thereby realizing the risk assessment of cardiovascular disease.

[0044] Depend on Figure 1As can be seen from the method shown, the beneficial effects of the present application are: through multimodal collection of user physiological signals, starting from the user's physiological basis, the user's cardiovascular condition is evaluated through a predictive model algorithm, and a risk assessment report is generated based on the user's personalization.

[0045] In some embodiments of the present application, the above method includes: dynamic physiological characteristic data includes electrocardiogram signals, pulse signals, blood pressure parameters, and blood oxygen saturation parameters; static physiological characteristic data includes age, gender, chest pain type, blood sugar parameters, and cholesterol values.

[0046] ECG signal acquisition uses the AD8232 as the core integrated circuit, which includes bioelectrodes, signal acquisition units, signal processing units, and signal output units. The bioelectrode circuit is connected to the signal acquisition unit using a bioelectrode connector. The signal acquisition unit transmits the collected bioelectrode signal to the signal processing unit. The signal processing unit amplifies, filters, and Fourier transforms the weak ECG signal to extract the amplitude and frequency of the ECG signal. After the ECG signal is restored through inverse Fourier transform, the ECG signal's amplitude, frequency, interval, QRS wave, slope and other parameter values are extracted through joint time-frequency analysis, specifically the maximum heart rate, average heart rate in an interval, PR interval, QT interval, ST segment data changes, and heart rate variability data.

[0047] ECG signal acquisition can also be achieved using other integrated circuit chips, such as ADS1299, MAX30003, etc. During the signal processing process, other parameters can also be combined for multi-parameter fusion analysis.

[0048] Physiological parameters such as pulse waveform, heart rate, blood oxygen saturation, microcirculation, reference blood pressure, and fatigue status are measured using the high-precision sensor MKS-5V45-HRV-FP. The built-in signal conditioning circuit and algorithm processing module directly output important vital sign indicators such as pulse waveform and heart rate values that have undergone preliminary processing. By receiving data packets and parsing the data in the data packets according to the parsing rules, the corresponding values of each parameter are obtained.

[0049] You can also consider using the MAX30102 chip in combination with an AD amplifier circuit to achieve joint monitoring of heart rate, pulse, and blood oxygen. The authenticity and accuracy of the obtained data can be guaranteed through the ADC converter and the corresponding compensation algorithm.

[0050] Users' static physiological data can be obtained through user complaints and classified into specific categories using the corresponding data format. For example, age can be divided into age ranges of 1-9, where 1 represents the 0-10 age range, 2 represents the 11-20 age range, and so on, with 9 representing over 90 years old. Chest pain types include typical angina, atypical angina, and non-cardiac chest pain, and can be divided into severity levels of 1-5, where 1 represents no obvious feeling, 2 represents mild pain, 3 represents paroxysmal pain, 4 represents severe pain, and 5 represents unbearable pain. Users can obtain blood sugar and cholesterol values through daily measurements, and define values within a reasonable range as normal, and values that are too high or too low as abnormal, with 1 representing normal, 0 representing too low, and 2 representing too high.

[0051] Through the above technical solutions, the multimodal collection of patients' physiological data is guaranteed, the comprehensive understanding of patients' physical conditions is improved, the reliability of data is improved, the accuracy of data analysis is improved, and the comprehensiveness of risk assessment system data is guaranteed.

[0052] In some embodiments of the present application, the above method includes: extracting features of dynamic physiological characteristic data, including extracting the QRS wave group, heart rate variability characteristics, and ST segment offset of the electrocardiogram signal, and extracting the conduction velocity, reflected wave, and amplitude of the pulse signal.

[0053] The QRS complex is a core component of the electrocardiogram (ECG) signal, reflecting the rapid depolarization of the ventricular myocardium. Its morphology, duration, and amplitude are crucial for diagnosing cardiovascular disease. The QRS complex consists of the Q wave, R wave, and S wave. The Q wave is the first downward wave, the R wave is the first upward wave, and the S wave is the downward wave following the R wave. Parameters such as the QRS complex duration, amplitude, Q wave depth, and R wave increment should be within normal ranges. Parameter values outside these ranges indicate different disease risks. For example, a QRS complex duration >110ms may indicate ventricular arrhythmia. Abnormal amplitude may indicate perimyocardial fluid, obesity, or myocardial damage. A Q wave width exceeding 30ms and a depth exceeding 1mm may indicate myocardial infarction or hypertrophic cardiomyopathy.

[0054] The heart rate variability feature is used to measure small fluctuations during the heartbeat interval. The standard deviation of the PR interval, the root mean square of the difference between adjacent PR intervals, the difference between adjacent PR intervals, and the standard deviation of the average PR interval in a unit time period can be calculated in the time domain. The power, low-frequency ratio, high-frequency ratio, and low-frequency-high-frequency ratio within a certain frequency range can also be calculated in the frequency domain.

[0055] ST segment deviation is an important indicator for assessing myocardial ischemia, myocardial infarction, or other cardiac pathological conditions. It reflects abnormal electrical activity of myocardial cells during ventricular repolarization. In the electrocardiogram (ECG) signal, the interval from the end of the QRS complex to the onset of the T wave corresponds to the rapid repolarization period. Normally, it is aligned with the PR segment isoelectric line, with no significant upward or downward deviation. When the upper edge of the ST segment is above the PR isoelectric line, also known as ST segment elevation, the slope is positive, suggesting acute myocardial infarction. When the upper edge of the ST segment is below the PR isoelectric line, also known as ST segment depression, the slope is negative, suggesting myocardial ischemia, unstable angina, etc.

[0056] Pulse signal parameters can be characterized by extracting their conduction velocity, reflected wave, and amplitude. The normal range for pulse signal conduction velocity is set, such as 5-8 m / s for the carotid-femoral artery. A velocity exceeding 10 m / s may indicate cardiovascular disease risk. The pulse reflected wave index is calculated based on the main systolic wave amplitude and the reflected wave amplitude. The ratio of the incremental reflected wave amplitude to the main wave amplitude reflects the pulse reflected wave augmentation index (AII). A normal value is below 80%. An elevated AII indicates an increased risk of arterial stiffness.

[0057] Through the above technical solution, the extraction of characteristic parameters of electrocardiogram signals and pulse signals is guaranteed. By calculating the characteristic parameters, the specific parameter values are associated with the corresponding cardiovascular disease risks. Through multimodal data collection, the prediction of cardiovascular disease risks is achieved.

[0058] In some embodiments of the present application, the above method includes: the prediction model is a CatBoost prediction model.

[0059] CatBoost (Categorical Boosting) is a gradient boosting decision tree algorithm developed by Yandex that can efficiently process complex data types.

[0060] The characteristics of CatBoost are automatic processing of categorical features without manual coding, automatic conversion of feature parameters into numerical features, automatic combination of features of multiple categories, mining of interactive relationships between features, and enrichment of feature dimensions.

[0061] Using the CatBoost prediction model for cardiovascular disease prediction includes data preparation, model training, and model evaluation.

[0062] Data preparation includes extracting dynamic physiological characteristic data, specifically including the time domain and frequency domain features of dynamic ECG signals, pulse signals, etc., such as extracting heart rate-specific features, ST segment slope, pulse wave conduction velocity, etc. from ECG signals; it also includes static physiological characteristic data, specifically including age, blood pressure, cholesterol, blood sugar, etc.; it also includes chief symptoms, specifically including the user's medical history, daily routine, living habits, etc., such as whether there has been a specific disease, whether there has been a lack of sleep for a long time, whether there is smoking or drinking, etc.

[0063] The UCI Cleveland database is used for model training. This database is a classic medical database, mainly used for predictive modeling of heart disease diagnosis. It is a commonly used benchmark dataset in the field of medical data analysis and machine learning.

[0064] The selected database contains 303 patient cardiac assessment instances, covering 14 attributes, including age, gender, chest pain type, cholesterol level, and resting electrocardiogram results. During initial data processing, the data needed to be cleaned to eliminate outliers or missing values. Six records with missing values were excluded, ultimately retaining 297 complete records. The dataset was generally balanced, with 53.87% of patient assessments being positive and the remaining 46.13% being negative.

[0065] To ensure the reliability of the study results, the research was narrowed down to 10 attributes that were compatible with the hardware and highly relevant to cardiac assessment. The attributes ultimately selected included age, sex, chest pain type, resting blood pressure, cholesterol level, fasting blood glucose, maximum heart rate, exercise-induced angina, ST slope, and post-exercise ST depression.

[0066] Model training process: from catboost import CatBoostClassifier, Pool import pandas as pd # Load data (here load a file named data.csv) data = pd.read_csv('data.csv') X = data.drop('target', axis=1) y = data['target'] #Specify the category feature column name, here you can select ten categories cat_features = ['gender','smoking','diabetes'] # Create a dataset (supports automatic processing of categories and missing values, and can choose to supplement or discard missing values) train_pool = Pool(X, y, cat_features=cat_features) # Initialize the model model = CatBoostClassifier( iterations=1000, # number of trees learning_rate=0.03, # learning rate depth=6, # tree depth loss_function='Logloss', # binary logarithmic loss eval_metric='AUC', # evaluation metric early_stopping_rounds=50, # Early stopping to prevent overfitting task_type='GPU'# Enable GPU acceleration ) # Training and validation (automatically dividing training / validation sets) model.fit(train_pool, plot=True) Model evaluation involves inputting the data to be evaluated into the training model. The trained model will then generate evaluation results based on the input data.

[0067] Through the aforementioned technical solutions, CatBoost's core advantages in cardiovascular disease prediction lie in its automated handling of categories and missing data, robustness to overfitting for small sample sizes, high-precision capture of complex feature relationships, and interpretability that meets clinical needs. Combined with its efficient GPU computing capabilities, CatBoost can be seamlessly integrated into hospital information systems, wearable devices, and remote monitoring platforms, enabling intelligent support from risk warning to personalized treatment, providing a reliable technical tool for the precise prevention and control of cardiovascular disease.

[0068] In some embodiments of the present application, the above method includes: pre-training includes: dividing the pre-training data set into a training set and a test set in proportion, and calculating the prediction model score after pre-training based on the risk assessment parameters.

[0069] The dataset for model training is split into a training set and a test set in an 8:2 ratio. 80% of the UCI data will be used to train the model to identify patterns in the data, while the test set will be used to evaluate the model's ability to handle unknown data after training. A large performance gap between the training and test sets may indicate that the model is "overfitting," meaning that the model has learned too much about the details of the training data, which negatively affects its performance on unfamiliar new data.

[0070] Through the above technical solution, the training set is used for pre-training, and then the pre-training results are tested using the test set to evaluate the performance gap between the training set and the test set. By adjusting the parameters of the training process, the model is avoided from over-learning, which ensures the rationality of the prediction model and improves the prediction accuracy of the risk assessment parameters.

[0071] In some embodiments of the present application, the above method includes: risk assessment parameters include AUC, precision, recall and F1 score.

[0072] AUC-ROC (Area Under the Receiver Operating Characteristic Curve) is a core indicator for evaluating the performance of binary classification models. The ROC curve is formed by plotting the true positive rate and false positive rate at different classification thresholds. AUC is the area under the ROC curve and ranges from [0,1]. The closer the value is to 1, the better the effect. This parameter does not depend on the classification threshold and can evaluate the overall performance of the model. It is suitable for unbalanced data and is insensitive to category distribution. It can be used to intuitively compare models. The AUC value can directly compare the advantages and disadvantages of different models.

[0073] Precision measures the reliability of the model's predictions for positive samples, that is, the proportion of true positive examples in the positive predictions.

[0074] Recall measures the model's coverage of positive samples, that is, the proportion of correct predictions among all positive samples.

[0075] The F1 score, the harmonic mean of precision and recall, comprehensively reflects the model performance.

[0076] Through the above technical solution, the model is evaluated from multiple angles such as AUC, precision, recall rate, and F1 score, which enables multi-dimensional evaluation of the model, ensures the comprehensiveness of the model, and improves the accuracy of cardiovascular disease risk assessment.

[0077] In some embodiments of the present application, the above method includes: generating a feature contribution report based on the SHAP algorithm.

[0078] SHAP (Shapley Additive exPlannations) is a model interpretation method based on game theory that can quantify the contribution of each feature to model predictions.

[0079] The impact of each feature on the predicted outcome for individual samples is shown on the Y-axis, sorted by their average contribution to the final outcome. The X-axis displays the SHAP value, reflecting the degree of influence each feature has on the predicted outcome. Positive values indicate a positive influence, indicating a higher risk of cardiovascular disease (CVD), while negative values indicate a negative influence, indicating a lower risk. Color represents the feature value, with red indicating a higher feature value and blue indicating a lower feature value (e.g., sex = 1 = red, sex = 0 = blue).

[0080] Visual displays help doctors understand users' risks and propose targeted, personalized interventions.

[0081] Through the above technical solutions, SHAP provides an intuitive explanation for cardiovascular disease risk assessment by quantifying feature contributions, assisting doctors and patients in understanding the decision-making basis of AI intelligent models, formulating personalized intervention plans for key risk factors, and improving the reliability of cardiovascular disease assessment.

[0082] In some embodiments of the present application, the above method includes: a communication connection for uploading data collected by the data collection module.

[0083] By integrating multiple high-performance sensors and advanced data processing technologies, comprehensive monitoring of cardiovascular disease patients or those at risk is achieved. Compared to traditional methods that rely solely on a single indicator, this approach can more objectively and accurately reflect an individual's true health status. Coupled with the support of supporting intelligent algorithms, the final risk assessment conclusions are both authoritative and humane, greatly facilitating doctors' development of personalized treatment plans while also increasing the general public's control over their own health. More importantly, this is accomplished through a portable and easy-to-maintain physical platform, making heart health monitoring convenient and fast.

[0084] Figure 2 A flow chart of a cardiovascular disease risk assessment system according to another embodiment of the present application is shown.

[0085] S1, data acquisition, after the physiological characteristic data acquisition module is started, it initializes each sensor and communication component, and prepares to receive data from the electrocardiogram acquisition unit and the multi-functional sensing unit. The human-computer interaction module interface receives the user's basic information, such as age, gender, chest pain type, blood sugar parameters, cholesterol, and main symptoms.

[0086] S2, data processing, the electrocardiogram acquisition unit collects the original electrocardiogram signal, and outputs it to the signal care unit after denoising, amplification, and filtering. The multifunctional sensing unit collects pulse waveform, heart rate value and other related vital signs data respectively, and outputs them to the signal processing unit for processing after pre-processing through the signal conditioning circuit. The signal processing unit performs time-frequency joint analysis on the input signal and then conducts risk assessment.

[0087] S3, risk assessment. The risk assessment module collects all physiological data and calls the CatBoost model pre-stored in the internal flash memory to perform deep mining calculations on the batch of data to generate prediction results for the probability of cardiovascular disease. The SHAP model is used to analyze the weight ratio of each key influencing factor in the prediction process.

[0088] S4, Report Feedback: The complete report document will be pushed to the cloud server at the specified address through the successfully configured communication component using the MQTT protocol for permanent archiving and backup, and the email notification mechanism will be triggered simultaneously to inform relevant personnel to view the latest inspection feedback details.

[0089] This implementation principle is based on standardized operating procedures to minimize human error and improve overall operational efficiency and service levels. A mature AI intelligent machine learning framework, a reliable and efficient automated diagnosis and treatment platform, enhances the efficiency and convenience of cardiovascular disease risk assessment and enables personalized intervention plans.

[0090] Figure 3 A schematic structural diagram of a cardiovascular disease risk assessment device according to an embodiment of the present application is shown. The device 300 includes a physiological characteristic data acquisition module 301, a human-computer interaction module 302, a structured data module 303, and a risk assessment module 304.

[0091] Physiological characteristic data acquisition module 301, used to obtain dynamic physiological characteristic data; The human-computer interaction module 302 is used to obtain static physiological characteristic data of the user; The structured data module 303 is used to extract features from the dynamic physiological characteristic data, normalize the dynamic physiological characteristic data and the static physiological characteristic data, and generate structured input data; The risk assessment module 304 is used to input the structured input data into the pre-trained prediction model to generate a cardiovascular disease risk assessment report.

[0092] Optionally, in the above device, the physiological characteristic data acquisition module 301 is used to collect dynamic physiological characteristic data including electrocardiogram signals, pulse signals, blood pressure parameters, and blood oxygen saturation parameters, and static physiological characteristic data including age, gender, chest pain type, blood sugar parameters, and cholesterol values.

[0093] Optionally, in the above device, the structured data module 303 is used to extract features from dynamic physiological characteristic data, including extracting the QRS complex, heart rate variability characteristics, and ST segment offset of the electrocardiogram signal, and extracting the conduction velocity, reflected wave, and amplitude of the pulse signal.

[0094] Optionally, in the above device, the prediction model used by the risk assessment module 304 is a Catboost prediction model.

[0095] Optionally, in the above device, the risk assessment module 304 is used for pre-training, including: dividing the pre-training data set into a training set and a test set in proportion, and calculating the score of the pre-trained prediction model based on the risk assessment parameters.

[0096] Optionally, in the above apparatus, the risk assessment module 304 is configured to assess risk using parameters including AUC, precision, recall, and F1 score.

[0097] Optionally, in the above device, the risk assessment module 304 is configured to generate a feature contribution report based on the SHAP algorithm.

[0098] In some embodiments of the present application, the above-mentioned device further includes: a communication unit, which is used to upload the collected data.

[0099] To ensure that the acquired information can be shared promptly with remote professionals or stored for future use, a communication component is incorporated into the design. This component is connected to the dynamic physiological signal acquisition module and relies on the popular MQTT protocol to package the various locally collected data and upload it to a specific remote server. You can choose to use a Wi-Fi network for high-speed interconnection over a large area, or you can use Bluetooth technology for convenient and fast data exchange over short distances. Both methods have their own advantages and can be flexibly determined based on the actual application scenario.

[0100] It should be noted that the above-mentioned cardiovascular disease risk assessment device can implement the above-mentioned cardiovascular disease risk assessment method one by one, and will not be described in detail here.

[0101] In summary, the beneficial effects of this application are: cardiovascular disease risk assessment is achieved through convenient equipment, diagnostic costs are reduced, and the convenience of early intervention for cardiovascular disease is improved.

[0102] Figure 4 1 shows a schematic structural diagram of an electronic device according to an embodiment of the present application. The electronic device 400 includes a processor 410 and a memory 420 arranged to store computer-executable instructions (computer-readable program codes).

[0103] The memory 420 may be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM, a hard disk, or a ROM. The memory 420 has a storage space 430 for storing computer-readable program code 431. For example, the storage space 430 for storing computer-readable program code may include various computer-readable program codes 431 for respectively implementing the various steps in the above method. The computer-readable program code 431 may be read from or written to one or more computer program products. These computer program products include program code carriers such as a hard disk, a compact disk (CD), a memory card, or a floppy disk. Such a computer program product is typically, for example, Figure 5 The computer-readable storage medium.

[0104] Figure 5 A schematic diagram of the structure of a computer-readable storage medium according to one embodiment of the present application is shown. The computer-readable storage medium 500 stores computer-readable program code 431 for executing the method steps according to the present application and can be read by the processor 410 of the electronic device 400. When the computer-readable program code 431 is executed by the electronic device 400, the electronic device 400 executes each step of the method described above. Specifically, the computer-readable program code 431 stored in the computer-readable storage medium can execute the method described in any of the above embodiments. The computer-readable program code 431 can be compressed in an appropriate form.

[0105] A computer-readable storage medium may be a tangible device that holds and stores instructions for use by an instruction execution device. A computer-readable storage medium may be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium may be a portable computer disk, a hard drive, a USB flash drive, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a rostrum random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, an optical disc, a magnetic disk, a mechanical encoding device, or any combination thereof.

[0106] It should be noted that the above embodiments are illustrative rather than limiting of the present invention, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claims.

[0107] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to which steps and units are clearly listed, but may include other steps or units that are not clearly listed that are inherent to these processes, methods, products or devices. The terms "first", "second" and the like are used to distinguish similar phenomena and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate for the embodiments of the invention described herein. The present application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim that lists several means, several of these means may be embodied by the same item of hardware.

Claims

1. A method for assessing cardiovascular disease risk, characterized in that: include: Obtaining dynamic physiological characteristic data of users; Obtain the user's static physiological characteristic data; performing feature extraction on the dynamic physiological characteristic data, normalizing the dynamic physiological characteristic data and the static physiological characteristic data to generate structured input data; The structured input data is input into a pre-trained prediction model to generate a cardiovascular disease risk assessment report.

2. The cardiovascular disease risk assessment method according to claim 1, characterized in that: The dynamic physiological characteristic data include electrocardiogram signals, pulse signals, blood pressure parameters, and blood oxygen saturation parameters; the static physiological characteristic data include age, gender, chest pain type, blood sugar parameters, and cholesterol values.

3. The cardiovascular disease risk assessment method according to claim 2, characterized in that: The feature extraction of the dynamic physiological characteristic data includes extracting the QRS wave group, heart rate variability characteristics, and ST segment offset of the electrocardiogram signal, and extracting the conduction velocity, reflected wave, and amplitude of the pulse signal.

4. The cardiovascular disease risk assessment method according to claim 3, characterized in that: The prediction model is a Catboost prediction model.

5. The cardiovascular disease risk assessment method according to claim 4, characterized in that: The pre-training includes: dividing the pre-training data set into a training set and a test set in proportion, and calculating the prediction model score after pre-training based on the risk assessment parameters.

6. The cardiovascular disease risk assessment method according to claim 5, characterized in that: The risk assessment parameters include AUC, precision, recall and F1 score.

7. The cardiovascular disease risk assessment method according to any one of claims 1 to 6, characterized in that: Generate feature contribution report based on SHAP algorithm.

8. A cardiovascular disease risk assessment device, characterized in that: include: Physiological characteristic data acquisition module, used to obtain dynamic physiological characteristic data; Human-computer interaction module. Used to obtain the user's static physiological characteristic data; a structured data module, configured to extract features from the dynamic physiological characteristic data, normalize the dynamic physiological characteristic data and the static physiological characteristic data, and generate structured input data; The risk assessment module is used to input the structured input data into a pre-trained prediction model to generate a cardiovascular disease risk assessment report.

9. An electronic device, characterized in that: The electronic device comprises a processor and a memory, wherein the processor is coupled to the memory, and the processor is configured to execute a computer program stored in the memory, so that the electronic device executes the method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that The method comprises a computer program or instructions, which, when executed on a computer, causes the computer to execute the method according to any one of claims 1 to 5.

Images