Interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram

Abstract

The present invention discloses an interpretable classification method for multi-stage electrocardiogram-based cardio-cerebral health monitoring, belonging to the field of early disease screening. The method comprises the following steps: establishing a prospective cohort based on inclusion / exclusion criteria, with the prospective cohort serving as a reference population; collecting multi-stage electrocardiogram data from the reference population to establish a reference sample of high-quality electrocardiogram data; obtaining multi-stage electrocardiogram variables based on the reference sample; creating an individual profile of the reference population based on the multi-stage electrocardiogram variables to obtain a reference space; calculating the dissimilarity between the individual to be classified and the reference population to obtain an individual profile of the individual to be classified; and projecting the individual profile of the individual to be classified onto the reference space to determine the classification result. The present invention achieves robust classification of patient status by mapping subtle changes in electrocardiogram variables to the population level.

Description

Technical Field

[0001] The present invention belongs to the technical field of early disease screening, and in particular relates to an explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiograms. Background Art

[0002] In the fast-paced, high-pressure environment of modern society, individuals are experiencing a significant increase in mental stress. Long-term chronic stress not only poses a challenge to mental health but also has profound impacts on physical health. Under the influence of chronic stress, the regulation of the autonomic nervous system may be disrupted, leading to cardiovascular dysfunction such as elevated blood pressure. This long-term stress on the cardiovascular system may increase the risk of myocardial ischemia and may even trigger acute myocardial infarction, imposing a serious health and economic burden on the individual patient, society, and their families.

[0003] Currently, society lacks sufficient awareness and attention to individual emotional regulation and mental health. Public understanding of mental health is often limited to extreme mental illnesses, while overlooking the potential impact of chronic stress on emotions and behavior. This neglect can exacerbate problems and, in some cases, even develop into serious mental health crises. In recent years, the concepts of the "heart-brain axis" and the "brain-heart axis" have gradually revealed the bidirectional regulatory relationship between the heart and the brain, emphasizing the importance of focusing on both mental and heart health in research and practice.

[0004] The electrocardiogram (ECG) is a convenient and economical health monitoring tool, widely used in various fields from primary health care to home health monitoring. It is a preferred screening tool for assessing physical and mental health. Despite this, current ECG-based algorithms mainly focus on screening for diseases such as arrhythmias, while ignoring the ECG's potential for mental health. Heart rate variability (HRV), an ECG-derived indicator, is a biomarker reflecting the regulation of the autonomic nervous system. It has been widely confirmed in the literature that it can reflect an individual's mental state. The widely used waveform characteristics of the ECG can reflect cardiac activity. In summary, the ECG is a powerful tool for long-term monitoring of a patient's physical and mental health.

[0005] However, the application of ECG in the field of cardio-cerebral health monitoring still faces the following difficulties: (1) Most existing ECG-based studies do not record the corresponding mental state of the subjects, and the research faces the problem of insufficient data; (2) There are large differences in individual responses to mental stress stimuli, and ECG signals during mental stress are more susceptible to environmental and patient noise, which increases the difficulty of robust variable analysis.

[0006] Therefore, establishing a corresponding reference cohort and developing a robust ECG analysis method that is insensitive to interference and bias are crucial for cardio-cerebral health monitoring. Summary of the Invention

[0007] In order to solve the above technical problems, the present invention proposes an interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram to solve the problems existing in the above-mentioned prior art.

[0008] To achieve the above objectives, the present invention provides an interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiograms, comprising:

[0009] Establish a prospective cohort based on the inclusion and exclusion criteria, and the prospective cohort serves as the reference population;

[0010] Collecting multi-stage electrocardiogram data of the reference population to establish a reference sample of high-quality electrocardiogram data;

[0011] obtaining multi-stage electrocardiogram variables based on the reference sample;

[0012] creating individual profiles of a reference population based on the multi-stage electrocardiogram variables to obtain a reference space;

[0013] Calculating the dissimilarity between the individual to be classified and the reference population to obtain an individual profile of the individual to be classified;

[0014] The individual profile of the individual to be classified is projected onto the reference space to determine a classification result.

[0015] Optionally, the reference population includes patients with angina pectoris, subjects with high mental stress, and a healthy control group.

[0016] Optionally, the process of creating a reference sample of high-quality ECG data includes:

[0017] The reference population was continuously monitored to obtain electrocardiograms;

[0018] Extracting label information and performing noise reduction on the electrocardiogram to obtain standardized electrocardiogram data;

[0019] A reference sample of high-quality electrocardiogram data of a reference population is obtained based on the standardized electrocardiogram data.

[0020] Optionally, the calculation process of obtaining multi-stage electrocardiogram variables based on the reference sample includes:

[0021] dividing the reference sample into a plurality of stages, wherein the plurality of stages include: a rest period, a stress period, and a recovery period;

[0022] The ECG variables of each stage were calculated independently to obtain several within-stage variables;

[0023] By calculating the values ​​between consecutive stages through directed subtraction, we can obtain several inter-stage difference variables;

[0024] Based on several intra-stage variables and inter-stage difference variables, 88 electrocardiogram-based explanatory variables were obtained.

[0025] Optionally, the process of creating an individual profile of a reference population based on the multi-stage electrocardiogram variables and classifying the individuals to be classified based on the individual profile of the reference population includes:

[0026] Based on the multi-stage electrocardiogram variables of the reference population, the corresponding variable set is calculated for each stage and the variable set is vectorized to obtain several individual vectors;

[0027] Calculate the dissimilarity between any two individual vectors to generate an individual profile file for each member of the reference population, and construct a reference space based on the individual profile files of several members;

[0028] Calculating the dissimilarity between the individual to be classified and the reference population to obtain an individual profile of the individual to be classified;

[0029] Projecting the individual profile of the individual to be classified into the reference space to determine a classification result;

[0030] Among them, the Pearson distance coefficient is used to quantify the dissimilarity between any two individual vectors. The expression for quantifying the dissimilarity between any two individual vectors is:

[0031] d(x i ,x j )=1-r(x i ,x j )

[0032] Where, represents the i-th individual, d(x i , x j ) represents individual x i and x j The distance between them, r(x i , x j ) represents the Pearson correlation coefficient.

[0033] Optionally, the process of projecting the individual profile of the individual to be classified into the reference space to obtain a classification result includes:

[0034] Projecting the individual profile of the individual to be classified into a specific reference space constructed from the individual profiles of the reference cohort;

[0035] The category to which the individual to be classified belongs is determined by minimizing the mean distance method, and the constructed effect is visualized using the multidimensional scaling projection method.

[0036] The present invention also provides a computer terminal device, comprising:

[0037] one or more processors;

[0038] a memory, coupled to the processor, for storing one or more programs;

[0039] When the one or more programs are executed by the one or more processors, the one or more processors implement an explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram.

[0040] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, an explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram is implemented.

[0041] Compared with the prior art, the present invention has the following advantages and technical effects:

[0042] This invention converts microscopic changes in multiple variables during patient electrocardiogram (ECG) monitoring into population-level distance metrics. This approach significantly reduces data perturbations caused by individual differences in responses to mental stimulation, as well as the impact of noise during data acquisition. By mapping subtle changes in ECG variables to the population level, robust classification of patient status is achieved. This interference-resistant method provides a solid technical foundation for applying the technology of this invention to everyday wearable devices such as smart bracelets and smart watches. By integrating this method into these devices, real-time monitoring and analysis of user health status can be achieved, providing users with timely health warnings and recommendations.

[0043] By mapping subtle changes in multiple ECG variables across multiple phases to the population level, this method significantly reduces the impact of data bias (individual heterogeneity) and data noise (instability in signal acquisition) on classification results, providing a new, interpretable classification method for cardiovascular and cerebrovascular health monitoring. This method not only improves classification accuracy but also enhances the interpretability of results, providing a powerful tool for early disease screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0045] Figure 1 Data collection for the reference queue of the embodiment of the present invention;

[0046] Figure 2 ECG variable computer profile algorithm construction for an embodiment of the present invention;

[0047] Figure 3 It is the construction of the overview algorithm and visualization of the effect of the embodiment of the present invention;

[0048] Figure 4 It is an iterative method for the reference space in the overview algorithm of an embodiment of the present invention;

[0049] Figure 5 This is the verification and application of the overview algorithm of the embodiment of the present invention. DETAILED DESCRIPTION

[0050] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0051] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0052] Example 1

[0053] like Figure 1 As shown, this embodiment provides an explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram, including the following steps:

[0054] This example uses multi-stage ECG data, based on group distance analysis, to capture individual responses to varying mental stressors. This data includes statistically significant HRV and waveform variables during rest, stimulation, and recovery periods, as shown in Table 1. By integrating these variables and transforming them into a set of features reflecting individual differences, this example constructs a unique individual profile.

[0055] Table 1

[0056]

[0057]

[0058]

[0059]

[0060] The core of this method lies in mapping microscopic variations in ECG variables to macroscopic population "distances," thereby quantifying individual physiological and psychological states at the population level. Furthermore, by comparing individual feature profiles with a reference space constructed from reference samples, this method enables preliminary classification of individual mental states and the presence of myocardial ischemia. This approach effectively reduces the influence of individual physiological variable fluctuations and improves classification accuracy.

[0061] As a specific implementation of this example, the method for constructing this characteristic profile relies on establishing a well-represented reference sample of stress-induced myocardial ischemia. To ensure that this method can be widely applied to the assessment of mental health status and the early warning of myocardial ischemia, this example constructed a diverse cohort that included a normal control group, stressed patients assessed using multiple stress scales, and patients with stress-induced myocardial ischemia.

[0062] To ensure broad sample representation and reduce data heterogeneity, this example employed strict age matching at enrollment. All cohorts used standardized screening tests to ensure consistent data collection. Subjects were assigned to a specific time period between 7:00 and 8:00 a.m. to receive mental stimulation and complete electrocardiogram (ECG) data collection simultaneously. This arrangement was designed to minimize the impact of circadian rhythms on the data and ensure consistency and comparability of the collected data.

[0063] Validation on a test cohort demonstrated the effectiveness of this method. This example provides a reference dataset and a robust monitoring algorithm, offering new tools for monitoring the physical and mental health of the population. These findings are expected to be integrated into wearable devices such as wristbands and smartwatches, enabling real-time monitoring of individual health status and timely early warning, thereby improving people's quality of life and health.

[0064] As a specific implementation of this embodiment, the inclusion and exclusion standards adopted in this embodiment are as follows:

[0065] Inclusion criteria: 1) women aged 18 to 75 years; 2) chest pain or other angina symptoms; 3) no obstructive coronary artery disease (coronary artery stenosis less than 50%); 4) self-reported depression and other problems.

[0066] Exclusion criteria: 1) chest pain caused by diseases other than the cardiovascular system; 2) complications of other serious diseases, such as pulmonary embolism, serious arrhythmia or aortic dissection; 3) complications of severe valvular disease; 4) New York Heart Association functional class IV; 5) myocardial infarction with coronary artery obstruction in the past month; 6) apical ballooning syndrome in the past month; 7) substance abuse, such as alcohol and / or illicit drugs, use of antidepressants and / or anxiolytics within 4 weeks before study enrollment; 8) participation in other drug trials in the past 3 months; complications of severe psychiatric illness; 9) current pregnancy or breastfeeding; 10) current use of postmenopausal hormone therapy and current use of psychotropic drugs.

[0067] Potential participants were identified by the clinical care team and initial discussions were held with potential participants, and those who expressed interest in participating in the study were asked to provide written consent.

[0068] As a specific implementation method of this embodiment, the following steps are included: establishing a prospective cohort based on the inclusion and exclusion criteria, where the prospective cohort is a reference population; collecting multi-stage electrocardiogram data of the reference population to establish a reference sample of high-quality electrocardiogram data; calculating based on the reference sample to obtain multi-stage electrocardiogram variables; creating an individual profile of the reference population based on the multi-stage electrocardiogram variables to obtain a reference space; calculating the dissimilarity between the individual to be classified and the reference population to obtain the individual profile of the individual to be classified; and projecting the individual profile of the individual to be classified onto the reference space to determine the classification result.

[0069] (1) Furthermore, the cohort was constructed, including the following three groups of people:

[0070] a) Subjects aged 18-75 years with angina pectoris but without organic heart changes.

[0071] b) Subjects aged 18-75 years with higher than normal levels of mental stress (For psychological assessment, the Hospital Anxiety and Depression Scale (HADS) was used to measure depressive symptoms, which contains different subscales for assessing anxiety (HADS-A) and depression (HADS-D). Mood status was assessed using the Positive and Negative Affect Scale (PANAS), which distinguishes between positive emotions (PANAS-P) and negative emotions (PANAS-N). The State-Trait Anxiety Inventory (STAI) provides insights into current (state) anxiety (STAI-S) and general (trait) anxiety (STAI-T) levels. In addition, participants completed the Perceived Stress Scale (PSS), the Life Events Scale (LES), and the Post-Traumatic Stress Disorder (PTSD) Checklist-Civilian Version (PCL-C) to provide a comprehensive overview of their psychological state.)

[0072] c) Healthy controls age-matched to a and b.

[0073] (2) Furthermore, the data collection process includes: using a standard 12-lead ECG system to continuously monitor the reference population to obtain an ECG; performing signal filtering and label information extraction on the ECG to obtain ECG features; performing standardization processing on the ECG features to obtain standardized ECG features; and obtaining a reference sample of high-quality ECG data of the reference population based on the ECG features.

[0074] Furthermore, this embodiment uses a standard 12-lead electrocardiogram system and Tim software provided by a Beijing, China company for continuous monitoring. The system has 16-bit accuracy and a sampling frequency of 500 Hz. The electrocardiogram recording begins with the participant in a resting state and continues for an additional six minutes after the completion of three mental stress environments (Stroop test, angry recall speech, and timed mental arithmetic). Figure 1The MedEx MECG-200 ECG analysis system was used to filter the signal and extract signature information. Noise removal from power supply interference, baseline drift, and muscle contraction was meticulously performed using two median filters (200 ms and 600 ms) and Daubechies wavelets. These extracted features were normalized using z-score standardization. Myocardial ischemia-related signatures were determined by three experienced clinicians based on an SDS ≥ 3 on PET / CT imaging during the mental stimulation period (clinical gold standard).

[0075] (3) Furthermore, the multi-stage ECG variable calculation process includes: dividing the reference sample into several stages, including: rest period, stress period and recovery period, independently calculating the ECG variables of each stage to obtain several intra-stage variables; obtaining several inter-stage difference variables through directional subtraction of consecutive inter-stage values; and obtaining 88 ECG-based interpretable variables based on several intra-stage variables and inter-stage difference variables.

[0076] Furthermore, for the ECG variables, a detailed literature review was conducted to identify 88 ECG-based explanatory variables, including ECG waveform variables and heart rate variability variables. The ECG recordings were exported and segmented into three different phases corresponding to the timeline of the mental stress task: rest period (6 minutes), stress period (12 minutes), and recovery period (6 minutes), as shown in Figure 2. Figure 2 ECG variables were calculated independently for each phase, resulting in a total of 264 within-phase variables (88 per phase). Furthermore, another set of between-phase difference variables was derived by performing directional subtraction of values ​​between consecutive phases (stress-rest, recovery-stress, and rest-recovery), for a total of 264 variables (88 in each direction).

[0077] (4) Further, individual profile algorithm description

[0078] This example proposes a distance-based approach to construct inter-group differentiation features using ECG data, called “individual profiles.” The approach consists of two main steps:

[0079] First, a specific reference space is established for the three populations mentioned above. Based on the multi-stage ECG variables of the reference cohort, and by calculating the dissimilarity between the individual and other individuals in the multi-stage ECG variables, an individual profile file of each member is generated. The projection of individual profile files of the same category forms the reference space of the category. In the verification data of this embodiment, the number of categories is 3. Secondly, by evaluating the differences between the individuals to be classified and the members of the reference cohort, profile files of the individuals to be classified are created. These profile files are then projected into the constructed reference space, and the classification results are determined by their positions in the reference space, such as Figure 2 shown.

[0080] In the profile construction, from the data set D = {x1, x2, ..., x N}, where N represents the size of the dataset and {1, 2, ..., C} represents the patient labels. Based on the statistically significant ECG variables from within-stage, between-stage, and aggregate sets, a vector containing the within-stage and between-stage variables was created for each individual. The differences between these vectors were then quantified using the Pearson distance coefficient, calculated as:

[0081] d(x i ,x j )=1-r(x i ,x j )

[0082] in represents the i-th patient, c i represents the label of the i-th patient, d(x i , x j )

[0083] Represents patient x i and x j The distance between them, r(x i , x j ) represents the Pearson correlation coefficient, which is defined as follows:

[0084]

[0085] in represents the vth variable of the i-th patient, V represents the length of the reduced variable, and Represents patient x i and x j The mean over all variables.

[0086] In this paper, the research cohort is used as the dataset D. Using all individual profile files of the research cohort, the interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram establishes a specific reference space through MDS projection, and then performs the classification task on the test cohort. For a new sample x t The goal of this embodiment is to determine its k The labels in the (k=1, 2, ..., C) classes are classified by minimizing the mean distance, which is calculated from the distances between samples.

[0087]

[0088] Among them, C * Represents a new sample x * Category, N kIndicates category c k The number of samples in d(x i ,x t ) represents the sample x i with x t The distance between them.

[0089] Figure 3 The details of the profile algorithm are presented, including the integration of intra-stage variables and inter-stage variables. During the research process, the interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiograms performed two-dimensional MDS projections on the profiles constructed by intra-stage and inter-stage variables and the fusion of the two, respectively. It was proved that the profile constructed by combining intra-stage and inter-stage variables has a better effect of narrowing the distance within the class and widening the distance between classes. Figure 3 Figure C shows that using only the raw ECG variables results in randomly scattered data points, with no clear grouping. However, when applying intra-stage or inter-stage individual profiles, samples in the same group cluster more closely, while samples from different groups become more separated. This clearer grouping effect is even more evident in the projection of the individual profiles constructed using the integrated vectors. Table 2 shows the performance of the individual profile method on the test set (n = 53).

[0090] Table 2

[0091]

[0092] like Figure 4 As shown in Figure 2, the actual application process of individual profiles is as follows: (i) the reference sample size is expanded through screening tests; (ii) the reference space is iteratively optimized until the centroids of the corresponding categories no longer change significantly (reach a certain threshold), indicating that a basic stable state is reached; (iii) the new individual profile files are projected into the reference space for preliminary classification based on the group centroid distance.

[0093] Furthermore, this example further evaluated the robustness of the analysis method. First, this example cut and spliced ​​data from the study cohort. The enhanced dataset was divided into 20 subsets. By gradually increasing the sample size, it was confirmed that when the sample size was large enough, the deviation of the group center achieved stability, such as Figure 5 As shown in the figure, part A is a schematic diagram of the reference space iteration, and part B is the Control, MSIMI(-) and MSIMI(+) category centers. During the data iteration process, the movement gradually decreases and reaches stability. This confirms the scalability of the present invention: when the number of people in the reference space is increased, the present invention will have higher robustness; and by incorporating more common cardiovascular and cerebrovascular diseases, the types of diseases in the reference space can be increased, expanding the scope of application of the method, thereby enabling daily monitoring of multiple common cardiovascular and cerebrovascular diseases.

[0094] The present invention collects electrocardiogram (ECG) data from subjects under stress under strictly controlled experimental conditions. This step effectively fills the data gap in existing research and provides valuable raw data for subsequent analysis.

[0095] Furthermore, as a specific embodiment of this embodiment, the present invention aims to promote early screening or routine monitoring of cardiocerebral diseases, making its applicability to single-lead data of wearable devices crucial. Therefore, this embodiment calculated the Spearman rank correlation coefficient between the single-lead curve and the 12-lead curve of the test cohort and evaluated the performance of each electrocardiogram lead. As shown in Table 3, although the Spearman rank correlation coefficient of the single-lead curve shows a slight change compared with the 12-lead curve, it does not affect the classification results. This shows that the analysis method may be applicable to single-lead data from wearable devices.

[0096] Table 3

[0097]

[0098] Wherein, ρ represents the Spearman rank correlation coefficient, Sen represents sensitivity, Spe represents specificity, and Acc represents accuracy.

[0099] Example 2

[0100] This embodiment further provides a computer terminal device, including:

[0101] one or more processors;

[0102] a memory, coupled to the processor, for storing one or more programs;

[0103] When one or more programs are executed by one or more processors, the one or more processors implement an explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram.

[0104] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, an explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram is implemented.

[0105] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. An interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram, characterized by: The following steps are involved: Establish a prospective cohort based on the inclusion and exclusion criteria, and the prospective cohort serves as the reference population; Collecting multi-stage electrocardiogram data of the reference population to establish a reference sample of high-quality electrocardiogram data; obtaining multi-stage electrocardiogram variables based on the reference sample; The calculation process for obtaining multi-stage electrocardiogram variables based on a reference sample includes: dividing the reference sample into a plurality of stages, wherein the plurality of stages include a resting period, a stress period, and a recovery period; independently calculating the electrocardiogram variables of each stage to obtain a plurality of intra-stage variables; calculating the values ​​between consecutive stages by directed subtraction to obtain a plurality of inter-stage difference variables; and obtaining 88 electrocardiogram-based explanatory variables based on the plurality of intra-stage variables and the inter-stage difference variables. creating individual profiles of a reference population based on the multi-stage electrocardiogram variables to obtain a reference space; Calculating the dissimilarity between the individual to be classified and the reference population to obtain an individual profile of the individual to be classified; The process of classifying the individuals to be classified includes: calculating a corresponding variable set for each stage based on multi-stage electrocardiogram variables of a reference population, and vectorizing the variable set to obtain a plurality of individual vectors; calculating the dissimilarity between any two individual vectors to generate an individual profile file for each member of the reference population, and constructing a reference space based on the individual profile files of the plurality of members; calculating the dissimilarity between the individual to be classified and the reference population to obtain an individual profile of the individual to be classified; and projecting the individual profile of the individual to be classified onto the reference space to determine a classification result. Among them, the Pearson distance coefficient is used to quantify the dissimilarity between any two individual vectors. The expression for quantifying the dissimilarity between any two individual vectors is: d(x i ,x j )=1-r(x i ,x j ) Where, represents the i-th individual, r(x i , x j ) represents individual x i and x j The distance between them, r(x i , x j ) represents the Pearson correlation coefficient; The individual profile of the individual to be classified is projected onto the reference space to determine a classification result; wherein the process of obtaining the classification result includes: projecting the individual profile file of the individual to be classified onto a specific reference space constructed by the individual profiles of the reference cohort; determining the category to which the individual to be classified belongs by minimizing the mean distance, and visualizing the constructed effect using a multidimensional scaling projection method.

2. The interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram according to claim 1 is characterized in that: The reference population includes patients with angina pectoris, subjects with high mental stress, and a healthy control group.

3. The interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram according to claim 1 is characterized in that: The process of creating a reference sample of high-quality ECG data includes: The reference population was continuously monitored to obtain electrocardiograms; Extracting label information and performing noise reduction on the electrocardiogram to obtain standardized electrocardiogram data; A reference sample of high-quality electrocardiogram data of a reference population is obtained based on the standardized electrocardiogram data.

4. A computer terminal device, characterized in that: include: one or more processors; a memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the explainable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram as described in any one of claims 1-3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the interpretable classification method for cardio-cerebral health monitoring based on multi-stage electrocardiogram as described in any one of claims 1 to 3 is implemented.

Images