Electrocardiosignal positioning method, diagnosis device and heart sound and electrocardio detection system

Abstract

The invention belongs to the technical field of intelligent medical treatment, and particularly relates to an electrocardiosignal positioning method, a diagnosis device and a heart sound and electrocardio detection system. The electrocardiosignal positioning method comprises the following steps: acquiring a plurality of original lead electrocardiosignals; screening out N first lead electrocardiosignals from the plurality of original lead electrocardiosignals, wherein N is a positive integer; determining an R wave peak point position of the N first lead electrocardiosignals in combination with the N first lead electrocardiosignals; n electrocardiosignal data segments with a first preset time length are intercepted from the N first lead electrocardiosignals on the basis of the determined R wave peak value point position, the first preset time length is larger than T and smaller than zeta T, and T represents the duration of a heart movement cycle; and inputting the N electrocardiosignal data segments to a pre-trained positioning model to obtain a plurality of waveform positioning point positions of the N electrocardiosignal data segments. According to the invention, high-precision waveform positioning can be rapidly carried out, and an accurate waveform positioning point position is obtained.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, and in particular to a method for locating electrocardiogram (ECG) signals, a diagnostic device, and a heart sound and ECG detection system. Background Technology

[0002] Electrocardiogram (ECG) signals are an important basis for the clinical diagnosis of cardiovascular diseases. Figure 2 This diagram illustrates the periodic waveform of an electrocardiogram (ECG) signal. The accuracy of waveform localization points directly affects the reliability of medical analyses such as heart rate calculation and arrhythmia detection. Waveform localization points include the P-wave start point, P-wave end point, QRS wave start point (also called Q-wave start point), QRS wave end point (also called S-wave end point), T-wave start point, and T-wave end point. Traditional waveform localization relies heavily on physician experience, making it difficult to guarantee diagnostic accuracy and efficiency. Therefore, related technologies provide the following automatic ECG waveform localization methods that do not rely on manual intervention:

[0003] (1) Threshold detection method: By setting a fixed threshold or sliding window to detect QRS complex, P wave and T wave, key point localization can be obtained. However, this method is sensitive to noise and is easily affected by baseline drift, electromyography interference, etc., which leads to false detection or missed detection. It has poor adaptability and its performance drops sharply in low signal-to-noise ratio (SNR) signals.

[0004] (2) Template matching method, which is based on predefined waveform templates to match ECG signals, but there are large differences in ECG morphology between individuals, and the template generalization ability is limited, making it difficult to cover ECG variations in pathological states (such as the disappearance of P waves in atrial fibrillation).

[0005] It is evident that existing methods for automatically locating key points in electrocardiogram signals without relying on human intervention suffer from technical problems such as noise sensitivity and poor generalization, leading to a high misdiagnosis rate in clinical applications. Summary of the Invention

[0006] This application aims to at least solve the technical problems existing in the prior art, and to provide a method for locating electrocardiogram signals, a diagnostic device, and a heart sound and electrocardiogram detection system.

[0007] In a first aspect, this application provides a method for locating electrocardiogram (ECG) signals, the method comprising: acquiring multiple raw lead ECG signals; selecting N first lead ECG signals from the multiple raw lead ECG signals, where N is a positive integer; determining the position of an R-wave peak point of the N first lead ECG signals by combining the N first lead ECG signals; based on the determined R-wave peak point position, extracting N ECG signal data segments of a first preset time length from the N first lead ECG signals, where T < first preset time length < ζT, T represents the duration of one cardiac cycle, ζ represents the extraction coefficient, 0 < ζ < 1; inputting the N ECG signal data segments into a pre-trained localization model to obtain multiple waveform localization point positions of the N ECG signal data segments.

[0008] Preferably, the step of selecting N first lead ECG signals from multiple original lead ECG signals includes: calculating the sample entropy and / or permutation entropy of each original lead ECG signal in the multiple original lead ECG signals, and determining the sample entropy threshold and / or permutation entropy threshold; selecting one or more original lead ECG signals from the multiple original lead ECG signals that meet the entropy value condition, wherein the entropy value condition includes that the sample entropy of the original lead ECG signal is less than the sample entropy threshold, and / or that the permutation entropy of the original lead ECG signal is less than the permutation entropy threshold; and obtaining N first lead ECG signals based on one or more original lead ECG signals that meet the entropy value condition.

[0009] Preferably, obtaining N first lead ECG signals based on one or more original lead ECG signals that satisfy the entropy value condition includes: performing principal component analysis on one or more original lead ECG signals that satisfy the entropy value condition, and determining N first lead ECG signals based on the principal component analysis results.

[0010] Preferably, determining the location of an R-wave peak point from the N first-lead ECG signals by combining the N first-lead ECG signals includes: dividing the N first-lead ECG signals into a reference lead ECG signal and N-1 candidate lead ECG signals; determining one or more reference R-wave peak points that satisfy the R-wave peak point condition on the reference lead ECG signal; and iterating through the one or more reference R-wave peak points one by one, with each iteration performing the following: if the corresponding point of the currently traversed reference R-wave peak point in one or more candidate lead ECG signals satisfies the R-wave peak point condition, then the currently traversed reference R-wave peak point is taken as the R-wave peak point, and the iteration ends; if the corresponding points of the currently traversed reference R-wave peak point in none of the N-1 candidate lead ECG signals satisfy the R-wave peak point condition, then the next reference R-wave peak point is traversed.

[0011] Preferably, the R-wave peak point condition includes: the amplitude of the currently traversed reference R-wave peak point is greater than the R-wave amplitude threshold; the average absolute value of the slope of the reference lead ECG signal within the determination time interval is greater than a preset slope threshold, wherein the determination time interval is obtained by extending a second preset time length forward and a third preset time length backward based on the currently traversed reference R-wave peak point.

[0012] Preferably, the localization model includes a cascaded CNN network and a fully connected layer.

[0013] Secondly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the electrocardiogram signal localization method described in the first aspect.

[0014] Thirdly, this application provides an electrocardiogram (ECG) signal localization device for implementing an ECG signal localization method provided in the first aspect, comprising: a signal acquisition module for acquiring multiple raw lead ECG signals; a filtering module for filtering N first lead ECG signals from the multiple raw lead ECG signals, where N is a positive integer; an R-wave peak point determination module for determining the position of an R-wave peak point of the N first lead ECG signals in conjunction with the N first lead ECG signals; a data truncation module for truncating N ECG signal data segments of a first preset time length from the N first lead ECG signals based on the determined R-wave peak point position, where T < first preset time length < ζT, T represents the duration of one cardiac cycle, ζ represents the truncation coefficient, and 0 < ζ < 1; and a localization module for inputting the N ECG signal data segments into a pre-trained localization model to obtain multiple waveform localization point positions of the N ECG signal data segments.

[0015] Fourthly, this application provides a diagnostic device, comprising: a signal input module for inputting multiple leads of electrocardiogram (ECG) signals and heart sound signals; an ECG localization module for obtaining multiple waveform localization point positions of N ECG signal data segments according to the steps of an ECG signal localization method provided in the first aspect of this application; an ECG feature extraction module for extracting multiple ECG features from the N ECG signal data segments based on the multiple waveform localization point positions; a heart sound feature extraction module for extracting multiple heart sound features from the heart sound signals; and a diagnostic module for inputting the multiple ECG features and the multiple heart sound features into a pre-trained diagnostic model to obtain a diagnostic result.

[0016] Fifthly, this application provides a heart sound and electrocardiogram (ECG) detection system, comprising: a signal acquisition unit connected to multiple ECG electrodes and a sound sensor, for acquiring multiple leads of ECG signals from the multiple ECG electrodes and for acquiring heart sound signals from the sound sensor; a diagnostic device provided in the fourth aspect of this application, the diagnostic device being connected to the signal acquisition unit; and a printer connected to the signal acquisition unit and the diagnostic device, for printing a heart sound and ECG, wherein the positions of multiple waveform positioning points are marked in the heart sound and ECG.

[0017] The beneficial technical effects of this application are as follows:

[0018] This application provides an ECG signal localization method, computer program product, and ECG signal localization device. First, it eliminates ECG signals with significant noise interference from multiple original leads, selecting N first-lead ECG signals with less noise. This reduces the impact of power line interference, baseline drift, and electromyographic interference on localization accuracy, while also reducing the amount of data processed subsequently, thus improving localization efficiency. Next, it combines the N first-lead ECG signals to determine the location of an R-wave peak point, fully utilizing the amplitude, slope, and phase correlation of multi-lead ECG signals. By leveraging the spatiotemporal consistency between leads, it ensures the accuracy of the R-wave peak point. This method is also applicable to ECG signal variations in pathological states such as atrial fibrillation, improving the generalization of the localization method. Finally, it segments the N ECG signal data and inputs these segments into a localization model to obtain multiple waveform localization point locations for each segment. The localization model quickly and accurately identifies characteristic waves of different morphologies, enabling rapid and high-precision waveform localization point location even in the presence of signal interference or abnormalities.

[0019] In addition to the aforementioned beneficial technical effects, the diagnostic device and cardiac sound / electrocardiogram detection system provided in this application also combine cardiac sound signals and electrocardiogram signals, inputting multiple electrocardiogram features and multiple cardiac sound features into the diagnostic model to obtain diagnostic results, which can improve diagnostic accuracy and reliability and reduce the misdiagnosis rate. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a preferred embodiment of the electrocardiogram signal localization method of the present invention;

[0021] Figure 2 This is a schematic diagram of the periodic waveform of an electrocardiogram (ECG) signal;

[0022] Figure 3 This is a schematic diagram of the positioning lines corresponding to the positions of multiple waveform positioning points in an example of the present invention;

[0023] Figure 4 This is a schematic diagram of the electrocardiogram signal localization device in a preferred embodiment of the present invention;

[0024] Figure 5 This is a schematic diagram of the structure of a diagnostic device in a preferred embodiment of the present invention;

[0025] Figure 6 This is a schematic diagram of the structure of a heart sound and electrocardiogram detection system according to a preferred embodiment of the present invention;

[0026] Figure 7 This is a schematic diagram of the 12-lead electrocardiogram signal waveform of a patient with atrial fibrillation in an example;

[0027] Figure 8 This is an example of a central heart rate electrocardiogram. Detailed Implementation

[0028] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0029] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0030] In the description of this invention, unless otherwise specified and limited, it should be noted that the terms "installation", "connection" and "linking" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two components. They can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0031] The execution entity of the ECG signal localization method disclosed in this invention includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in the embodiments of this application: a server, a terminal, etc. In other words, the ECG signal localization method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0032] This invention provides a method for locating electrocardiogram (ECG) signals. In a preferred embodiment, please see... Figure 1 The method includes:

[0033] Step S1: Acquire multiple raw lead ECG signals.

[0034] In this embodiment, depending on the lead configuration during the actual electrocardiogram (ECG) test, the multiple raw lead ECG signals are not limited to 3, 5, 12, or 15 raw lead ECG signals. The raw lead ECG signals can be obtained through a heart sound ECG detection system or the signal acquisition unit of an ECG detection system, and are digital signals.

[0035] For example, the original ECG signal is 12 leads. Figure 7 An example of a 12-lead electrocardiogram for a patient with atrial fibrillation is shown. Figure 8 An example of a heart sound electrocardiogram is shown. Figure 8 The top 12 channels are 12-lead electrocardiograms. As you can see, some of the leads in the 12-lead series have signals that are too flat. These leads cannot be used for waveform localization. If these leads are selected to implement the traditional single-lead localization method, localization will not be completed, or there will be a large localization error.

[0036] Step S2: Select N first lead ECG signals from multiple raw lead ECG signals, where N is a positive integer.

[0037] In this embodiment, noise is evaluated for each original lead ECG signal, and the original lead ECG signal with less noise is selected as the first lead ECG signal based on the noise evaluation results, so as to reduce the impact of noise on waveform localization.

[0038] Step S3: Combine the N first-lead ECG signals to determine the location of an R-wave peak point from the N first-lead ECG signals. The location of the R-wave peak point is the time of the R-wave peak point on the time axis.

[0039] In this embodiment, the electrocardiogram (ECG) waveform corresponding to one cardiac motion cycle is denoted as a cardiac motion cycle ECG waveform, and each first lead ECG signal is composed of multiple cardiac motion cycle ECG waveforms. Theoretically, each cardiac motion cycle ECG waveform has an R-wave peak point (see [link to ECG waveform description]). Figure 2 The amplitude and slope changes of the R-wave peak point are more significant compared to other waves. Therefore, locating the R-wave peak point first is more conducive to achieving accurate positioning of the entire waveform.

[0040] However, some pathological electrocardiograms may show reversed R waves (e.g., in patients with atrial fibrillation), and even some normal electrocardiograms may show flat and small fluctuations in certain leads (e.g., Figure 8 In this study, relying solely on a single lead signal for R-wave peak location may result in significant errors. Therefore, based on the spatiotemporal consistency of multiple lead ECG signals, this implementation method combines N first lead ECG signals to determine the location of an R-wave peak point from the N first lead ECG signals, and then uses the determined R-wave peak point location for subsequent waveform localization.

[0041] In this embodiment, the position of the R-wave peak point of any one cardiac cycle's ECG waveform is determined among multiple cardiac motion cycles of the ECG signal. Due to the spatiotemporal consistency of ECG waveforms between leads, if an ECG signal in a certain first lead is an R-wave peak point at a certain time position, then theoretically, there must be an R-wave peak point in the other N-1 first lead ECG signals at that time position.

[0042] Step S4: Based on the determined R-wave peak position, extract N ECG signal data segments of a first preset time length from the N first lead ECG signals, where T < first preset time length < ζT, T represents the duration of one cardiac cycle, ζ represents the truncation coefficient, and 0 < ζ < 1. The time intervals of the N ECG signal data segments are consistent.

[0043] In this embodiment, the duration T of the cardiac cycle can be set empirically. Alternatively, the patient's heart rate can be determined based on the multiple received raw ECG signals, and then the duration of the patient's cardiac cycle can be determined based on the patient's heart rate and used as T. Preferably, please see... Figure 2 Since the peak point of the R wave is not located at the center of the electrocardiogram waveform during the cardiac motion cycle, N electrocardiogram signal data segments are extracted by extending the peak point of the R wave to different lengths before and after the peak point, and the sum of the extended lengths before and after the peak point is the first preset time length.

[0044] Step S5: Input N ECG signal data segments into the pre-trained localization model to obtain multiple waveform localization point positions for the N ECG signal data segments. These multiple waveform localization point positions include the P-wave start point, P-wave end point, QRS wave start point (also called Q-wave start point), QRS wave end point (also called S-wave end point), T-wave start point, and T-wave end point, etc. Figure 3 The waveform location is marked from left to right with blue lines indicating the start and end points of the P-wave, QRS complex, T-wave, and T-wave. The waveform location may include only the time position of the waveform location, or it may include both the time position and the corresponding amplitude.

[0045] In this embodiment, preferably, to improve the positioning accuracy of the positioning model, after acquiring N ECG signal data segments, the time position of each sampling point in the N ECG signal data segments is modulated using the R-wave peak point position to obtain modulation data. The modulation data is then input into the positioning model to obtain multiple waveform positioning point positions. The N ECG signal data segments include a sub-segment of N first lead ECG signals.

[0046] For example, the shape of the N ECG signal data segments is: [N×L×2]. N represents the number of ECG signals in the first lead; L represents the first preset time length, i.e., including L sampling points; 2 represents the two dimensions of time position and amplitude information of each sampling point. Let t be the time position of the l-th sampling point in the sub-segment belonging to the n-th first lead ECG signal in the N ECG signal data segments. n,l Using the position of the R-wave peak point to determine t n,l Modulate, and t n,l Updated to t' n,l ,t' n,l =t R -t n,l The time position of each sampling point in N segments of ECG signal data is updated in the above manner to obtain modulated data. R The position of the R-wave peak point determined in step S3 is represented by n and l, where n and l are both positive integers, and n∈[1, N], l∈[1, L]. The modulated data and the N ECG signal data segments have the same shape.

[0047] In this embodiment, preferably, the localization model includes a cascaded CNN network and a fully connected layer. The CNN network extracts waveform features from N segments or modulated data of the electrocardiogram signal, and the fully connected layer regresses and maps the waveform features to multiple waveform localization points.

[0048] In this embodiment, the training process of the localization model includes:

[0049] Step 1: Acquire multiple multi-lead ECG signal data sets (from different patients and healthy individuals). Extract one or more N ECG signal data segments from each set. Plot the waveforms of each N ECG signal data segment. Based on the plotted waveforms, experts locate the R-wave peak position and multiple waveform positioning points. Obtain the modulation data corresponding to each N ECG signal data segment based on the expert-located R-wave peak position. Use each N ECG signal data segment or its corresponding modulation data as a sample data set. Use the multiple waveform positioning point positions (corresponding to time positions) located by the expert and the amplitude values ​​of each waveform positioning point in the N ECG signal data sets of that sample as the ground truth labels for that sample. Each waveform positioning point corresponds to N amplitude values ​​in the ground truth labels, thus completing the construction of one sample. Obtain the sample set using the above method.

[0050] Step 2: Build the network for the localization model and initialize the network parameters.

[0051] Step 3: Divide the sample set into a training set, a test set, and a validation set;

[0052] Step 4: Train the constructed localization model using the training set until the preset maximum number of training iterations is reached or the loss reaches the loss threshold, then stop training.

[0053] Step 5: Test and validate the trained localization model using the test set and validation set respectively. If the test and validation pass, the localization model training is complete. If the test and / or validation fail, adjust the training parameters such as the learning rate and return to step 4 to continue training.

[0054] In this embodiment, preferably, the loss function in step 4 is Loss:

[0055] Loss = α·Loss1 + β·Loss2

[0056] Where Loss1 represents the first loss component, and Loss1 includes the mean squared error loss L MSE and / or mean absolute error loss L MAE The first weighting coefficient α∈(0,1], preferably ranging from 0.6 to 0.8. The second weighting coefficient β∈(0,1], preferably ranging from 0.2 to 0.3.

[0057] Mean square error loss L MSE The mean square error between the predicted positions of multiple waveform positioning points of the sample output by the positioning model and the true value label of the sample is calculated using the following formula:

[0058]

[0059] Where Num represents the number of samples used in each training batch, i represents the index of the training sample in each training batch, and i is a positive integer; M represents the number of waveform localization points output by the localization model. In one example, M takes the value 6, and the 6 waveform localization point positions include the P-wave start point, P-wave end point, QRS wave start point (also called Q-wave start point), QRS wave end point (also called S-wave end point), T-wave start point, and T-wave end point. m represents the waveform localization point position index, m = 1, 2, ..., M. This represents the true position of the m-th waveform localization point in the true label of the i-th training sample in the current training batch, specifically the true time normalized to a dimensionless value. t i,m This represents the predicted position of the m-th waveform positioning point of the i-th training sample output by the positioning model in the current training batch, specifically the prediction time normalized to a dimensionless value.

[0060] Mean absolute error loss L MAE The formula for calculating the mean absolute error between the predicted positions of multiple waveform positioning points of the sample output by the positioning model and the true value label of the sample is as follows:

[0061]

[0062] Loss2 represents the second loss component, including waveform feature consistency loss L. feat Waveform feature consistency loss L feat The calculation formula is:

[0063]

[0064] in, This represents the true amplitude of the nth ECG signal at the mth waveform location point in the true label of the i-th training sample in the current training batch. This represents the predicted amplitude of the nth ECG signal of the i-th training sample output by the localization model in the current training batch at the predicted position of the m-th waveform localization point.

[0065] In this embodiment, the loss function Loss ensures localization accuracy through the first loss component. The second loss component prevents the localization model from relying solely on time patterns for prediction and ignoring the waveform's inherent characteristics. Furthermore, it simultaneously performs joint amplitude constraints on multiple leads of ECG signals, ensuring prediction accuracy.

[0066] In a preferred embodiment, step S2, which involves selecting N first-lead ECG signals from multiple original lead ECG signals, includes:

[0067] Step S21: Calculate the sample entropy and / or permutation entropy of each original lead ECG signal in multiple original lead ECG signals, and determine the sample entropy threshold and / or permutation entropy threshold.

[0068] Step S22: Select one or more original lead ECG signals from multiple original lead ECG signals that meet the entropy value condition. The entropy value condition includes that the sample entropy of the original lead ECG signal is less than the sample entropy threshold, and / or that the permutation entropy of the original lead ECG signal is less than the permutation entropy threshold.

[0069] Step S23: Obtain N first lead ECG signals based on one or more original lead ECG signals that satisfy the entropy value condition.

[0070] In this embodiment, the sample entropy threshold E τ1 for: in, δ1 represents the average sample entropy of multiple raw ECG leads, and K1 represents the standard deviation of the sample entropy of multiple raw ECG leads. K1 is the first coefficient, which is a positive integer, K1 = 1, 2, 3.

[0071] In this embodiment, the permutation entropy threshold E τ2 for: in, δ1 represents the average permutation entropy of multiple raw ECG leads, and δ2 represents the standard deviation of the permutation entropy of multiple raw ECG leads. K2 is the second coefficient, which is a positive integer, K2 = 1, 2, 3.

[0072] This implementation utilizes the principle that the higher the entropy value, the more disordered the signal (noise-dominated). It uses sample entropy and / or permutation entropy to quantify the regularity and complexity of the signal, and adaptively determines the sample entropy threshold and / or permutation entropy threshold based on the ECG signal. It automatically removes leads with abnormal entropy values ​​and retains only high-quality leads to participate in waveform localization point calculation.

[0073] In this embodiment, more preferably, step S23, obtaining N first lead ECG signals based on one or more original lead ECG signals that satisfy the entropy value condition, includes:

[0074] Principal component analysis (PCA) was performed on ECG signals from one or more original leads that met the entropy condition. Based on the PCA results, N first-lead ECG signals were determined. PCA was then performed on the original ECG signals from one or more original leads that met the entropy condition to extract principal components reflecting QRS complex characteristics and suppress interference components introduced by low-quality leads. In a mixed sample containing 30% low-quality leads, the waveform localization misclassification rate decreased from 21% to 6%, and the algorithm's robustness to noise was improved by 40%.

[0075] In a preferred embodiment, step S3, determining the location of an R-wave peak point from the N first-lead ECG signals by combining the N first-lead ECG signals, includes:

[0076] Step S31: Divide the N first lead ECG signals into one reference lead ECG signal and N-1 candidate lead ECG signals;

[0077] Step S32: Determine one or more reference R-wave peak points on the reference lead ECG signal that meet the R-wave peak point condition;

[0078] Step S33: Traverse one or more reference R-wave peak points one by one. Each traversal is as follows: If the corresponding point of the currently traversed reference R-wave peak point in one or more candidate lead ECG signals meets the R-wave peak point condition, then the currently traversed reference R-wave peak point is taken as the R-wave peak point, and the traversal ends; If the corresponding point of the currently traversed reference R-wave peak point in N-1 candidate lead ECG signals does not meet the R-wave peak point condition, further determine whether all reference R-wave peak points have been traversed. If all reference R-wave peak points have been traversed, then end the localization and prompt "ECG signal error". If there are still reference R-wave peak points that have not been traversed, then traverse the next untraversed reference R-wave peak point.

[0079] Preferably, the R-wave peak point conditions include:

[0080] (1) The amplitude of the current reference R-wave peak point is greater than the R-wave amplitude threshold; the R-wave amplitude threshold can be set based on experience.

[0081] (2) The average absolute value of the slope of the reference lead ECG signal within the determination time interval is greater than a preset slope threshold. The determination time interval is obtained by extending a second preset time length forward and a third preset time length backward based on the currently traversed reference R-wave peak point. The sum of the second preset time length and the third preset time length is the R-wave duration. The patient's heart rate is determined based on multiple original lead ECG signals, and then obtained by looking up a pre-constructed heart rate-second preset time length-third preset time length correspondence table based on the patient's heart rate.

[0082] In this embodiment, the amplitude and slope characteristics of the R-wave peak point are reflected through the R-wave peak point condition. Furthermore, the correlation of multiple leads is utilized; that is, the corresponding point of the currently traversed reference R-wave peak point in the ECG signal of more than one candidate lead satisfies the R-wave peak point condition. If the condition is not met, it indicates that the currently traversed reference R-wave peak point may not be the true R-wave peak point. It is evident that the amplitude, slope, and phase correlation of the multi-lead ECG signal exhibit spatiotemporal consistency, ensuring the accuracy of the determined R-wave peak point.

[0083] The present invention also discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the electrocardiogram signal localization method provided by the present invention. The computer program product should be understood as a software product that mainly implements its solution through a computer program, such as a program product integrated in the cloud or a software library.

[0084] This invention also discloses an electrocardiogram (ECG) signal localization device for implementing the ECG signal localization method provided by this invention. (See attached image.) Figure 4 In a preferred embodiment, the electrocardiogram signal localization device includes:

[0085] The signal acquisition module acquires multiple raw lead ECG signals;

[0086] The filtering module selects N first-lead ECG signals from multiple raw lead ECG signals, where N is a positive integer.

[0087] The R-wave peak point determination module combines N first-lead ECG signals to determine the location of one R-wave peak point in the N first-lead ECG signals;

[0088] The data interception module intercepts N ECG signal data segments of a first preset time length from N first lead ECG signals based on the determined R wave peak point position, where T < first preset time length < ζT, T represents the duration of one cardiac cycle, ζ represents the interception coefficient, and 0 < ζ < 1.

[0089] The positioning module takes N ECG signal data segments as input and feeds them into a pre-trained positioning model to obtain the positions of multiple waveform positioning points for the N ECG signal data segments.

[0090] In this embodiment, the signal acquisition module, the filtering module, the R-wave peak point determination module, the data interception module, and the positioning module correspond one-to-one with steps S1, S2, S3, S4, and S5 provided by the ECG signal positioning method of the present invention, and will not be described again here.

[0091] The present invention also discloses a diagnostic device, in a preferred embodiment, please see [link to preferred embodiment]. Figure 5 The diagnostic device includes:

[0092] The signal input module is used to input multiple leads of electrocardiogram signals and heart sound signals;

[0093] The ECG positioning module obtains the positions of multiple waveform positioning points for N ECG signal data segments according to the steps of the ECG signal positioning method provided by the present invention.

[0094] The ECG feature extraction module extracts multiple ECG features from N ECG signal data segments based on the positions of multiple waveform positioning points.

[0095] The heart sound feature extraction module is used to extract multiple heart sound features from the heart sound signal;

[0096] The diagnostic module takes multiple electrocardiogram features and multiple heart sound features as input to a pre-trained diagnostic model to obtain diagnostic results.

[0097] In this embodiment, the signal input module is preferably, but not limited to, a serial or parallel data interface connected to the signal acquisition unit.

[0098] In this embodiment, the ECG feature extraction module obtains ECG features such as heart rate, PR interval, QT interval, QRS axis, and maximum R wave amplitude based on the positions of multiple waveform positioning points in N ECG signal data segments.

[0099] In this embodiment, the heart sound feature extraction module is used to extract heart sound features such as heart sound intensity, frequency, and duration from the heart sound signal.

[0100] In this embodiment, the diagnostic model is preferably an MLP (Multilayer Perceptron) model. The diagnostic results include disease types such as arrhythmia, myocardial ischemia, valvular disease, atrial fibrillation, and normal status, as well as the confidence level for each disease type.

[0101] In this embodiment, a diagnostic sample set is constructed by collecting data from multiple cases. Each case includes the patient's heart sounds and electrocardiogram (ECG), and the doctor's diagnosis. Multiple ECG features and multiple heart sound features are extracted based on the ECG. The doctor's diagnosis includes the disease type. Each diagnostic sample in the sample set includes multiple ECG features and multiple heart sound features, and the sample label for each diagnostic sample is the disease type diagnosed by the doctor. The diagnostic sample set is then used to train an MLP (Multilayer Perceptron) model network to obtain a diagnostic model.

[0102] This invention also discloses a heart sound and electrocardiogram detection system. In a preferred embodiment, please see... Figure 6 The heart sound and electrocardiogram detection system includes:

[0103] The signal acquisition unit is connected to multiple ECG electrodes and a heart sound sensor, respectively, for acquiring multiple leads of ECG signals from the multiple ECG electrodes, and for acquiring heart sound signals from the heart sound sensor;

[0104] A network system connected to a signal acquisition device, including the diagnostic device described above provided by the present invention;

[0105] The printer, connected to the signal acquisition unit, is used to print heart sound electrocardiograms (ECGs). The ECGs mark the positions of multiple waveform positioning points. For example... Figure 7As shown, the markings are in the form of positioning lines.

[0106] In this embodiment, the signal acquisition unit can be a high-precision analog-to-digital converter module, with a sampling frequency not limited to 500Hz. It has multiple channels for acquiring analog signals output from multiple ECG electrodes and heart sound sensors. The network system includes a host and a display, and the diagnostic device can be deployed in the host. The host is connected to the display.

[0107] In the description of this specification, the references to terms such as "an embodiment," "some embodiments," "example," "specific example," "a implementation," "a preferred implementation," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0108] Although embodiments of the invention have been shown and described, those skilled in the art will understand that these embodiments can be adapted without departing from the principles and spirit of the invention.

[0109] With various changes, modifications, substitutions, and variations, the scope of this invention is defined by the claims and their equivalents.

[0110] Same thing limited.

Claims

1. A method for locating electrocardiogram (ECG) signals, characterized in that, The method includes: Acquire multiple raw lead ECG signals; N first-lead ECG signals are selected from multiple raw lead ECG signals, where N is a positive integer; The location of an R-wave peak point is determined by combining N first-lead ECG signals; Based on the determined R-wave peak position, N ECG signal data segments of a first preset time length are extracted from N first lead ECG signals, where T < first preset time length < ζT, T represents the duration of one cardiac cycle, ζ represents the extraction coefficient, and 0 < ζ < 1. Input N ECG signal data segments into a pre-trained localization model to obtain the positions of multiple waveform localization points for the N ECG signal data segments.

2. The method for locating electrocardiogram signals as described in claim 1, characterized in that, The process of selecting N first-lead ECG signals from multiple raw lead ECG signals includes: Calculate the sample entropy and / or permutation entropy of each original lead ECG signal in multiple original leads, and determine the sample entropy threshold and / or permutation entropy threshold; From multiple original lead ECG signals, one or more original lead ECG signals that meet the entropy value condition are selected. The entropy value condition includes that the sample entropy of the original lead ECG signal is less than the sample entropy threshold, and / or that the permutation entropy of the original lead ECG signal is less than the permutation entropy threshold. N first-lead ECG signals are obtained based on one or more original lead ECG signals that meet the entropy condition.

3. The method for locating electrocardiogram signals as described in claim 2, characterized in that, The process of obtaining N first-lead ECG signals based on one or more original lead ECG signals that satisfy the entropy condition includes: Principal component analysis is performed on one or more original lead ECG signals that meet the entropy condition, and N first lead ECG signals are determined based on the principal component analysis results.

4. The method for locating electrocardiogram signals as described in claim 1, characterized in that, The method of determining the location of an R-wave peak point from N first-lead ECG signals includes: The N first lead ECG signals are divided into one reference lead ECG signal and N-1 candidate lead ECG signals; Identify one or more reference R-wave peak points on the reference lead ECG signal that meet the R-wave peak point criteria; For one or more reference R-wave peak points, iterate through them one by one. Each iteration performs the following: if the corresponding point of the currently traversed reference R-wave peak point in one or more candidate leads of ECG signals meets the R-wave peak point condition, then the currently traversed reference R-wave peak point is taken as the R-wave peak point, and the iteration ends; if the corresponding points of the currently traversed reference R-wave peak point in N-1 candidate leads of ECG signals do not meet the R-wave peak point condition, then iterate through the next reference R-wave peak point.

5. The method for locating electrocardiogram signals as described in claim 4, characterized in that, The R-wave peak point conditions include: The amplitude of the current reference R-wave peak is greater than the R-wave amplitude threshold. The average absolute value of the slope of the reference lead ECG signal within the determination time interval is greater than a preset slope threshold. The determination time interval is obtained by extending forward by a second preset time length and backward by a third preset time length based on the current reference R wave peak point.

6. A method for locating electrocardiogram signals as described in any one of claims 1-5, characterized in that, The localization model includes a cascaded CNN network and fully connected layers.

7. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the electrocardiogram signal localization method according to any one of claims 1-6.

8. A cardiac signal localization device, used to implement the cardiac signal localization method according to any one of claims 1-6, characterized in that, include: The signal acquisition module acquires multiple raw lead ECG signals; The filtering module selects N first-lead ECG signals from multiple raw lead ECG signals, where N is a positive integer. The R-wave peak point determination module combines N first-lead ECG signals to determine the location of one R-wave peak point in the N first-lead ECG signals; The data interception module intercepts N ECG signal data segments of a first preset time length from N first lead ECG signals based on the determined R wave peak point position, where T < first preset time length < ζT, T represents the duration of one cardiac cycle, ζ represents the interception coefficient, and 0 < ζ < 1. The positioning module takes N ECG signal data segments as input and feeds them into a pre-trained positioning model to obtain the positions of multiple waveform positioning points for the N ECG signal data segments.

9. A diagnostic device, characterized in that, include: The signal input module is used to input multiple leads of electrocardiogram signals and heart sound signals; The ECG positioning module obtains the positions of multiple waveform positioning points for N ECG signal data segments according to the steps of an ECG signal positioning method according to any one of claims 1-6. The ECG feature extraction module extracts multiple ECG features from N ECG signal data segments based on the positions of multiple waveform positioning points. The heart sound feature extraction module is used to extract multiple heart sound features from the heart sound signal; The diagnostic module takes multiple electrocardiogram features and multiple heart sound features as input to a pre-trained diagnostic model to obtain diagnostic results.

10. A heart sound and electrocardiogram detection system, characterized in that, include: The signal acquisition unit is connected to multiple ECG electrodes and a sound sensor, respectively, for acquiring multiple leads of ECG signals from the multiple ECG electrodes, and for acquiring heart sound signals from the sound sensor; The diagnostic device as described in claim 9, wherein the diagnostic device is connected to a signal acquisition unit; A printer, connected to both the signal acquisition unit and the diagnostic device, is used to print a heart sound electrocardiogram (ECG), in which the positions of multiple waveform positioning points are marked.

Images