Electrocardiogram processing method and device based on wearable film sensor

Abstract

The embodiment of the invention provides an electrocardiogram processing method and device based on a wearable thin film sensor, the wearable thin film sensor comprises a main sensor and an auxiliary sensor, and the method comprises the steps that original electrocardiosignals of a single channel collected by the main sensor are obtained, and original electrocardiosignals of the single channel collected by the auxiliary sensor are obtained; the method comprises the following steps: synchronously acquiring an auxiliary reference signal which interferes with an original electrocardiosignal, performing interference correction on the original electrocardiosignal based on the auxiliary reference signal to obtain a target electrocardiosignal, performing spatial-temporal feature extraction on the target electrocardiosignal based on a pre-trained deep learning model to obtain spatial-temporal features, and outputting the spatial-temporal features. And performing lead mapping processing on the spatial-temporal characteristics to obtain a standard lead electrocardiogram. Therefore, the high-quality electrocardiogram is generated based on the single-channel electrocardiosignal, the accuracy of reconstructing the standard lead signal is improved, and the convenience and accessibility of electrocardiogram monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of medical engineering applications, and in particular to an electrocardiogram processing method and apparatus based on a wearable thin-film sensor. Background Technology

[0002] Electrocardiography (ECG) is an important tool for diagnosing heart disease. With the continuous rise in the incidence of heart disease, routine ECG monitoring is crucial for early screening and health management. Among existing ECG technologies, the standard lead system can provide multi-dimensional information on cardiac electrical activity, which is of great value in clinical diagnosis. However, this traditional ECG examination requires specialized equipment, fixed lead positions, and long monitoring times, which limits its convenience and long-term monitoring capabilities.

[0003] To improve the convenience and accessibility of ECG monitoring, wearable sensor devices have emerged, enabling lightweight, long-term, and even continuous monitoring. However, most current wearable sensor devices can only utilize ECG signals from a limited number of leads. Therefore, it is necessary to reconstruct standard lead signals from these limited lead data signals, but existing reconstruction methods have certain limitations and insufficient accuracy. Summary of the Invention

[0004] In view of the above problems, a method and apparatus for electrocardiogram processing based on a wearable thin-film sensor is proposed to overcome or at least partially solve the above problems, comprising: An electrocardiogram (ECG) processing method based on a wearable thin-film sensor, wherein the wearable thin-film sensor includes a main sensor and an auxiliary sensor, the method comprising: Acquire the raw ECG signal from a single channel obtained by the main sensor, and acquire the auxiliary reference signal that interferes with the raw ECG signal by an auxiliary sensor. Based on the auxiliary reference signal, the original electrocardiogram signal is subjected to interference correction to obtain the target electrocardiogram signal; Based on a pre-trained deep learning model, spatiotemporal features are extracted from the target electrocardiogram signal to obtain spatiotemporal features, and lead mapping is performed on the spatiotemporal features to obtain a standard lead electrocardiogram.

[0005] Optionally, the spatiotemporal features include local waveform features and long-range rhythm features, and the step of performing lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram includes: The local waveform features and the long-range rhythm features are spatiotemporally fused to obtain a three-dimensional feature tensor; the three-dimensional feature tensor is then mapped to the standard lead space to obtain a standard lead electrocardiogram.

[0006] Optionally, the wearable thin-film sensor is attached to the user's wrist, the auxiliary sensor is an accelerometer, and the step of performing interference correction on the original ECG signal based on the data from the auxiliary sensor to obtain the target ECG signal includes: Based on the auxiliary reference signal, the corresponding acceleration amplitude is obtained; The acceleration amplitude is filtered using a first preset frequency to obtain a motion noise signal; Based on the motion noise signal, motion artifacts in the original ECG signal are eliminated to obtain the target ECG signal.

[0007] Optionally, the wearable thin-film sensor is attached to the user's chest, the auxiliary sensor is a respiratory sensor, and the process of performing interference correction on the original electrocardiogram signal based on the data from the auxiliary sensor to obtain the target electrocardiogram signal includes: The auxiliary reference signal is filtered according to the second preset frequency to obtain the respiratory fluctuation signal; Based on the baseline fluctuation amplitude of the original electrocardiogram signal, the respiratory fluctuation signal is scaled to obtain a respiratory interference signal; Based on the respiratory interference signal, baseline drift correction is performed on the original electrocardiogram signal to obtain the target electrocardiogram signal.

[0008] Optionally, before performing signal correction on the original ECG signal based on the auxiliary reference signal to obtain the target ECG signal, the method further includes: Power frequency interference is detected on the original electrocardiogram signal to obtain the corresponding power frequency interference frequency; The original electrocardiogram (ECG) signal is filtered based on the notch filter corresponding to the power frequency interference frequency to obtain the original ECG signal after power frequency interference suppression.

[0009] Optionally, after denoising the original ECG signal based on the auxiliary sensor data to obtain the target ECG signal, the method further includes: By employing a sliding window mechanism, the quality of the target ECG signal is assessed, and the signal-to-noise ratio of the target ECG signal is determined. The signal-to-noise ratio is determined using a preset threshold, and if the determination result indicates that the target ECG signal is invalid, the original ECG signal is re-acquired.

[0010] Optionally, it also includes: The standard lead electrocardiogram is sent to a cloud server and a terminal display device; the cloud server is used to analyze the standard lead electrocardiogram, obtain the corresponding electrocardiogram analysis results, and send the electrocardiogram analysis results to the terminal display device.

[0011] An electrocardiogram (ECG) processing device based on a wearable thin-film sensor, wherein the wearable thin-film sensor includes a main sensor and an auxiliary sensor, and the device includes: The signal acquisition module is used to acquire the raw ECG signal of a single channel collected by the main sensor, and to acquire the auxiliary reference signal that interferes with the raw ECG signal by an auxiliary sensor. The signal correction module is used to perform interference correction on the original electrocardiogram signal based on the auxiliary reference signal to obtain the target electrocardiogram signal; The signal reconstruction module is used to extract spatiotemporal features from the target electrocardiogram signal based on a pre-trained deep learning model, obtain spatiotemporal features, and perform lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram.

[0012] The embodiments of the present invention have the following advantages: In embodiments of the present invention, a single-channel raw electrocardiogram (ECG) signal is acquired through the main sensor, and an auxiliary reference signal, simultaneously acquired through an auxiliary sensor, interferes with the raw ECG signal. Based on the auxiliary reference signal, interference correction is performed on the raw ECG signal to obtain a target ECG signal. Spatiotemporal features are extracted from the target ECG signal using a pre-trained deep learning model to obtain spatiotemporal features, and lead mapping processing is performed on these features to obtain a standard lead ECG. This achieves the generation of a high-quality ECG based on a single-channel ECG signal, thereby improving the accuracy of reconstructing the standard lead signal and enhancing the convenience and accessibility of ECG monitoring. Attached Figure Description

[0013] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart of the steps of an electrocardiogram processing method based on a wearable thin-film sensor provided in some embodiments of the present invention; Figure 2 This is a structural block diagram of an electrocardiogram processing device based on a wearable thin-film sensor provided in some embodiments of the present invention. Detailed Implementation

[0015] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0016] Reference Figure 1 The diagram illustrates a flowchart of an electrocardiogram (ECG) processing method based on a wearable thin-film sensor according to some embodiments of the present invention. The wearable thin-film sensor includes a main sensor and an auxiliary sensor, and the method may specifically include the following steps: Step 101: Acquire the raw ECG signal from a single channel obtained by the main sensor, and acquire an auxiliary reference signal that interferes with the raw ECG signal, simultaneously obtained by an auxiliary sensor. Currently, wearable sensor devices have emerged to improve the convenience and accessibility of ECG monitoring, enabling lightweight, long-term, and even continuous monitoring. However, most current wearable sensor devices can only utilize a limited number of ECG leads and are easily interfered with by other signals, resulting in insufficient accuracy in producing ECGs. In this application, the wearable thin-film sensor integrates both a main sensor and an auxiliary sensor.

[0017] In practical applications, the main sensor can utilize a novel piezoelectric composite material formed by thermally mixing piezoelectric ceramic powder and elastic polymer particles. This material possesses a high-strength, highly elastic, and flexible polymer structure, easily conforming to objects of any shape and capable of returning to its original state. The voltage signal generated by sensing deformation caused by external pressure in this polarized region can serve as a sensor with high sensitivity. This piezoelectric composite material also exhibits good flexibility and high elasticity, facilitating adhesion to various contact surfaces.

[0018] Specifically, after the wearable thin-film sensor is attached to the user's body surface, the raw ECG signal of a single channel, that is, the single-lead ECG signal, can be acquired through the main sensor. Then, the auxiliary reference signal that interferes with the raw ECG signal can be acquired through the auxiliary sensor.

[0019] Step 102: Based on the auxiliary reference signal, perform interference correction on the original ECG signal to obtain the target ECG signal.

[0020] After acquiring the raw ECG signal and auxiliary reference signal of a single lead, since the auxiliary reference signal is a physical signal that interferes with the raw ECG signal, the raw ECG signal can be corrected for interference by using the auxiliary reference signal to obtain the target ECG signal that filters out the interfering physical signal.

[0021] In some embodiments of this application, before performing signal correction on the original ECG signal based on the auxiliary reference signal to obtain the target ECG signal, the method further includes: Sub-step 11: Perform power frequency interference detection on the original electrocardiogram signal to obtain the corresponding power frequency interference frequency.

[0022] When acquiring electrocardiogram (ECG) signals, the electric / magnetic field of the power line may couple into the leads of the wearable thin-film sensor. This interference signal can be called power frequency interference, which is passively mixed into the original ECG signal. Therefore, it is necessary to detect power frequency interference in the original ECG signal to obtain the frequency of power frequency interference, i.e., the power frequency interference frequency.

[0023] In practical applications, the power frequency interference frequency may vary due to power grid standards (50Hz / 60Hz) or equipment aging (±2Hz drift). Therefore, when monitoring power frequency interference, a windowed Fourier transform can be performed on the electrocardiogram signal, with the window length set to an integer multiple of the power frequency period. The maximum spectral peak value can be searched in the 45-55Hz and / or 55-65Hz frequency bands. When the peak amplitude exceeds 10dB of the noise floor, the corresponding frequency is determined to be the power frequency interference frequency.

[0024] Sub-step 12: Based on the notch filter corresponding to the power frequency interference frequency, the original ECG signal is filtered to obtain the original ECG signal after suppressing the power frequency interference.

[0025] Specifically, after determining the detected power frequency interference frequency, a notch filter corresponding to the power frequency interference frequency can be used to filter the original ECG signal to suppress the power frequency interference. Zero-phase filtering can be used to filter the original ECG signal to avoid signal delay.

[0026] In some embodiments of this application, the wearable thin-film sensor is attached to the user's wrist, the auxiliary sensor is an accelerometer, and the step of performing interference correction on the original electrocardiogram signal based on the data from the auxiliary sensor to obtain the target electrocardiogram signal includes: Sub-step 21: Obtain the corresponding acceleration amplitude based on the auxiliary reference signal.

[0027] In a scenario involving ECG monitoring of a user, a wearable thin-film sensor can be attached to the user's wrist to acquire a single-channel ECG signal, i.e., a single-lead ECG signal. However, because the user's arm may shake or swing while walking, the acquisition of the single-channel ECG signal may be affected by motion noise. Therefore, in this detection scenario, an auxiliary sensor can be an accelerometer, and an auxiliary reference signal can be an acceleration sensor signal.

[0028] In practical applications, the acceleration sensor can be a triaxial accelerometer. The acceleration sensor signal acquired by the triaxial accelerometer can be normalized first, and then the instantaneous acceleration amplitude can be calculated by taking the square root of the sum of the squares of each component of the normalized acceleration sensor signal.

[0029] Sub-step 22 involves filtering the acceleration amplitude using a first preset frequency to obtain a motion noise signal.

[0030] After obtaining the acceleration amplitude, the acceleration amplitude can be bandpass filtered by a first preset frequency to extract the motion noise signal.

[0031] In practical applications, the first preset frequency can be a frequency range, with a lower limit of 0.1Hz to filter out slow posture changes (i.e., filter out non-motion artifact signals) and an upper limit of 10Hz to suppress high-frequency vibration noise (i.e., noise that exceeds the frequency band of the ECG signal). Bandpass filtering of the acceleration amplitude within this frequency range can retain motion components that overlap with the frequency band of the ECG signal, thereby obtaining the motion noise signal.

[0032] Sub-step 23: Based on the motion noise signal, the motion artifacts of the original ECG signal are eliminated to obtain the target ECG signal.

[0033] After acquiring the motion noise signal, it can be used as a reference input, and the original ECG signal as the main input. The weights are then iteratively updated using an adaptive filter, as shown in the following formula: e(t) = x(t) w(t)*n(t)(1) In formula (1), n(t) is the motion noise signal, w(t) is the dynamically adjusted adaptive filter coefficient, x(t) is the original ECG signal, and e(t) is the output target ECG signal.

[0034] In practical applications, the LMS algorithm (Least Mean Square Algorithm) can be used in adaptive filters.

[0035] In some embodiments of this application, the wearable thin-film sensor is attached to the user's chest, the auxiliary sensor is a respiratory sensor, and the step of performing interference correction on the original electrocardiogram signal based on the data from the auxiliary sensor to obtain the target electrocardiogram signal includes: Sub-step 31: Filter the auxiliary reference signal according to the second preset frequency to obtain the respiratory fluctuation signal.

[0036] In a scenario involving electrocardiogram (ECG) monitoring of a user, a wearable thin-film sensor can be attached to the user's chest to acquire a single-channel ECG signal, i.e., a single-lead ECG signal. However, because the user's chest may rise and fall due to breathing, the acquisition of a single-channel ECG signal may be affected by respiratory noise. Therefore, in this monitoring scenario, an auxiliary sensor can be a respiratory sensor, and the auxiliary reference signal can be the respiratory sensor signal.

[0037] In practical applications, the second preset frequency can be a frequency range, with a lower limit of 0.1Hz to filter out DC offset and ultra-low frequency noise, and an upper limit of 0.8Hz to cover the physiological respiratory frequency range. Thus, the respiratory sensor signal can be bandpass filtered from 0.1 to 0.8Hz according to the second preset frequency to obtain the respiratory fluctuation signal.

[0038] Sub-step 32: Based on the baseline fluctuation amplitude of the original electrocardiogram signal, the respiratory fluctuation signal is scaled to obtain a respiratory interference signal.

[0039] Specifically, the baseline fluctuation amplitude can be calculated based on the original electrocardiogram (ECG) signal. Then, the respiratory fluctuation signal is dynamically scaled to match the baseline fluctuation of the original ECG signal, resulting in a dynamically scaled respiratory fluctuation signal, i.e., a respiratory interference signal. This adapts the respiratory fluctuation signal to the actual impact level of the baseline drift of the original ECG signal, avoiding overcompensation or undercompensation in subsequent steps.

[0040] In practical applications, the low-frequency envelope of the original electrocardiogram (ECG) signal can be extracted by using a 0.5Hz low-pass filter to separate the slow-varying baseline fluctuation component caused by respiration or other physiological activities from the original ECG signal.

[0041] Sub-step 33: Based on the respiratory interference signal, perform baseline drift correction on the original ECG signal to obtain the target ECG signal.

[0042] After acquiring the respiratory interference signal, baseline drift correction can be performed on the original ECG signal based on the respiratory interference signal to obtain the target ECG signal.

[0043] Specifically, the respiratory interference signal can be directly subtracted from the original ECG signal, and the original ECG signal after baseline drift correction can be subjected to a 0.5Hz high-pass filter to eliminate residual drift, thereby obtaining the target ECG signal.

[0044] In some embodiments of this application, after denoising the original ECG signal based on the auxiliary sensor data to obtain the target ECG signal, the method further includes: Sub-step 41 involves using a sliding window mechanism to assess the quality of the target ECG signal and determine its signal-to-noise ratio.

[0045] After acquiring the target ECG signal, a sliding window mechanism can be used to assess the quality of the target ECG signal and determine its signal-to-noise ratio.

[0046] Specifically, the target ECG signal can be divided into overlapping sliding windows, for example, each window is 5 seconds long with a 50% overlap. Then, the QRS complex is detected within each window, and the R wave position is marked. Non-QRS segments, such as the TP interval, are extracted as noise regions, and the root mean square P1 (noise power) of the noise region is calculated. Furthermore, the peak value P2 (signal power) of the QRS complex amplitude can be calculated. Finally, based on the acquired noise power and signal power, the signal-to-noise ratio (SNR) of the target ECG signal is determined. This allows for dynamic evaluation of the target ECG signal quality, preventing the misuse of low-quality data.

[0047] The QRS complex reflects the changes in depolarization potential and time of the left and right ventricles. The first downward wave is the Q wave, the upward wave is the R wave, and the next downward wave is the S wave. The TP interval is the time period from the end of the T wave to the beginning of the next P wave in the electrocardiogram. It represents the electrical resting state of the heart during diastole and is an ideal range for assessing noise power.

[0048] Sub-step 42 involves using a preset threshold to determine the signal-to-noise ratio, and if the determination result indicates that the target ECG signal is invalid, re-acquiring the original ECG signal.

[0049] After acquiring the signal-to-noise ratio (SNR) of the target ECG signal, a threshold determination is performed. Specifically, an acceptable SNR threshold can be preset, such as an SNR less than 15dB. Then, the SNR is determined based on this preset threshold. If the determination result indicates that the acquired target ECG signal is invalid, the original ECG signal can be reacquired.

[0050] As an example, a specific determination method could be that if the signal-to-noise ratio of three consecutive windows is lower than a preset threshold, then the target ECG signal is determined to be invalid.

[0051] Step 103: Based on the pre-trained deep learning model, the spatiotemporal features of the target ECG signal are extracted to obtain spatiotemporal features, and the spatiotemporal features are processed by lead mapping to obtain a standard lead ECG.

[0052] After acquiring the target ECG signal, it can be input into a pre-trained deep learning model. The pre-trained deep learning model extracts spatiotemporal features from the target ECG signal to obtain spatiotemporal features. Then, the spatiotemporal features can be processed by lead mapping to obtain a standard lead ECG, such as a 12-lead ECG.

[0053] In one example, the pre-trained deep learning model can be a CNN-BiLSTM hybrid model based on an attention mechanism. CNN (Convolutional Neural Network) is a convolutional neural network, and BiLSTM (Bidirectional Long Short-Term Memory) is a bidirectional long short-term memory network, thus designing a deep learning model that can capture spatiotemporal features simultaneously.

[0054] The hybrid architecture of the pre-trained deep learning model can be as follows: CNN layer: used to extract local waveform features (such as QRS wave morphology) of the target ECG signal, and gradually abstract signal features through multi-layer convolution and pooling operations.

[0055] BiLSTM layer: Models the long-term temporal dependence of the target ECG signal, such as RR interval and heart rate variability. The bidirectional structure can utilize both past and future information simultaneously.

[0056] Attention mechanism: Dynamically weights spatiotemporal features to highlight key time points and lead correlations.

[0057] Lead mapping layer: 12 independent fully connected branches are used to generate standard 12-lead electrocardiogram waveforms.

[0058] As an example, the training process of a pre-trained deep learning model can be divided into the following steps: 1. Data preparation and preprocessing Collect synchronously acquired single-channel target ECG signals, such as those from wrist or chest leads, along with corresponding standard 12-lead ECGs, construct paired datasets, and perform signal normalization, segmentation, and data augmentation such as noise addition and time shifting.

[0059] 2. Design a multi-task loss function Waveform reconstruction loss: The difference in amplitude between the generated signal and the real signal is measured using mean square error or dynamic time warping.

[0060] Lead correlation loss: Calculate the Pearson correlation coefficient between generated leads to ensure that the spatial distribution conforms to physiological laws.

[0061] Key band constraints: Apply higher weights to clinically important bands, such as the ST segment, to avoid distortion that could affect diagnosis.

[0062] 3. Using self-supervised learning, such as contrastive learning, initialize the CNN-BiLSTM module to learn general ECG feature representations. Then, fix the encoder parameters, train the lead mapping layer separately to optimize waveform generation quality, and finally, unfreeze all parameters and fine-tune the entire model with a low learning rate to balance waveform accuracy and inter-lead consistency.

[0063] 4. Evaluate the model's performance and adjust the model based on the evaluation results to obtain the optimal model. In one example, the evaluation metrics may include waveform similarity, such as signal-to-noise ratio; clinical parameter errors, such as deviations in QT interval and ST segment shift; and computational efficiency.

[0064] In some embodiments of this application, the spatiotemporal features include local waveform features and long-range rhythm features, and the step of performing lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram includes: Sub-step 51 involves spatiotemporal fusion of the local waveform features and the long-range rhythm features to obtain a three-dimensional feature tensor.

[0065] By using a pre-trained deep learning model, spatiotemporal features are extracted from the target electrocardiogram signal. These spatiotemporal features can include local waveform features and long-range rhythm features. The local waveform features can be the time-frequency feature map output by the CNN, and the long-range rhythm features can be the time-series code output by the BiLSTM.

[0066] Local waveform features and long-range rhythm features are aligned along the time axis to ensure that the segment features at each moment match the global rhythm information. Then, the one-dimensional temporal features are expanded into a three-dimensional structure of width × height × number of channels through a fully connected layer. Here, the width can be the time dimension, such as 1000 sampling points, the height can be the feature dimension, such as a 64-dimensional feature vector, and the number of channels can be the lead implicit representation. Finally, the weights of different time points and feature dimensions are dynamically adjusted to highlight the interaction between key waveforms such as the QRS wave and rhythms such as heart rate variability, and weighted fusion is performed to obtain a three-dimensional feature tensor. Sub-step 52 maps the three-dimensional feature tensor to the standard lead space to obtain a standard lead electrocardiogram.

[0067] After obtaining the three-dimensional feature tensor, the pre-trained deep learning model can map the three-dimensional feature tensor to the standard lead space, that is, map the three-dimensional feature tensor to the standard lead electrocardiogram, thereby obtaining the standard lead electrocardiogram, which can be a 12-lead electrocardiogram.

[0068] In some embodiments of the present invention, it further includes: The standard lead electrocardiogram is sent to a cloud server and a terminal display device; the cloud server is used to analyze the standard lead electrocardiogram, obtain the corresponding electrocardiogram analysis results, and send the electrocardiogram analysis results to the terminal display device.

[0069] Specifically, after obtaining a 12-lead electrocardiogram (ECG), it can be transmitted to a terminal display device (such as a mobile phone or computer) via wired or wireless means for viewing. Alternatively, the 12-lead ECG can be sent to a cloud server for analysis, yielding corresponding ECG analysis results, which are then sent to the terminal display device for viewing.

[0070] In one example, a cloud server can perform AI system diagnosis on a 12-lead electrocardiogram, where AI system diagnosis means diagnosis based on artificial intelligence algorithms.

[0071] Specifically, the implementation steps of artificial intelligence algorithms can begin with the collection and preprocessing of electrocardiogram (ECG) data. This involves constructing a large-scale ECG dataset with significant characteristics and rigorously labeling it to ensure sufficient diversity and representativeness, which is crucial for the subsequent training and testing of the algorithm. Secondly, model training and validation can be performed. The preprocessed dataset is used for algorithm training, and methods such as cross-validation are employed to objectively evaluate model performance. Based on the evaluation results, algorithm parameters and structure are continuously optimized to improve the model's accuracy and reliability. Finally, the optimized algorithm can be deployed to an ECG cloud server for clinical use to continuously test diagnostic efficiency and accuracy. Simultaneously, user feedback and operational data are collected to further iterate and optimize algorithm performance, achieving continuous improvement in diagnostic capabilities.

[0072] In embodiments of the present invention, a single-channel raw electrocardiogram (ECG) signal is acquired through the main sensor, and an auxiliary reference signal, simultaneously acquired through an auxiliary sensor, interferes with the raw ECG signal. Based on the auxiliary reference signal, interference correction is performed on the raw ECG signal to obtain a target ECG signal. Spatiotemporal features are extracted from the target ECG signal using a pre-trained deep learning model to obtain spatiotemporal features, and lead mapping processing is performed on these features to obtain a standard lead ECG. This achieves the generation of a high-quality ECG based on a single-channel ECG signal, thereby improving the accuracy of reconstructing the standard lead signal and enhancing the convenience and accessibility of ECG monitoring.

[0073] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0074] Reference Figure 2 The diagram illustrates a structural schematic of an electrocardiogram processing device based on a wearable thin-film sensor according to some embodiments of the present invention. The wearable thin-film sensor includes a main sensor and an auxiliary sensor, and may specifically include the following modules: The signal acquisition module 201 is used to acquire the raw ECG signal of a single channel collected by the main sensor, and to acquire the auxiliary reference signal that interferes with the raw ECG signal by the auxiliary sensor. The signal correction module 202 is used to perform interference correction on the original electrocardiogram signal based on the auxiliary reference signal to obtain the target electrocardiogram signal; The signal reconstruction module 203 is used to extract spatiotemporal features from the target electrocardiogram signal based on a pre-trained deep learning model, obtain spatiotemporal features, and perform lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram.

[0075] In one embodiment of the present invention, the spatiotemporal features include local waveform features and long-range rhythm features, and the step of performing lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram includes: The feature fusion submodule is used to perform spatiotemporal fusion of the local waveform features and the long-range rhythm features to obtain a three-dimensional feature tensor; the mapping submodule is used to map the three-dimensional feature tensor to the standard lead space to obtain a standard lead electrocardiogram.

[0076] In one embodiment of the present invention, the wearable thin-film sensor is attached to the user's wrist, the auxiliary sensor is an accelerometer, and the step of performing interference correction on the original electrocardiogram signal based on the data from the auxiliary sensor to obtain the target electrocardiogram signal includes: An amplitude acquisition submodule is used to obtain the corresponding acceleration amplitude based on the auxiliary reference signal; The noise signal acquisition submodule is used to filter the acceleration amplitude using a first preset frequency to obtain a motion noise signal; The first signal acquisition submodule is used to eliminate motion artifacts in the original electrocardiogram signal based on motion noise signals to obtain the target electrocardiogram signal.

[0077] In one embodiment of the present invention, the wearable thin-film sensor is attached to the user's chest, the auxiliary sensor is a respiratory sensor, and the step of performing interference correction on the original electrocardiogram signal based on the data from the auxiliary sensor to obtain the target electrocardiogram signal includes: The wave signal acquisition submodule is used to filter the auxiliary reference signal according to a second preset frequency to obtain a respiratory wave signal; The interference signal acquisition submodule is used to scale the respiratory fluctuation signal according to the baseline fluctuation amplitude of the original electrocardiogram signal to obtain a respiratory interference signal; The second signal acquisition module is used to perform baseline drift correction on the original electrocardiogram signal based on the respiratory interference signal to obtain the target electrocardiogram signal.

[0078] In one embodiment of the present invention, before performing signal correction on the original electrocardiogram signal based on the auxiliary reference signal to obtain the target electrocardiogram signal, the method further includes: The interference rate determination module is used to perform power frequency interference detection on the original electrocardiogram signal to obtain the corresponding power frequency interference frequency; The interference frequency removal module is used to filter the original electrocardiogram (ECG) signal based on the notch filter corresponding to the power frequency interference frequency to obtain the original ECG signal after suppressing the power frequency interference.

[0079] In one embodiment of the present invention, after denoising the original electrocardiogram signal based on the auxiliary sensor data to obtain the target electrocardiogram signal, the method further includes: The signal-to-noise ratio (SNR) determination module is used to perform quality assessment on the target ECG signal by employing a sliding window mechanism, and to determine the SNR of the target ECG signal. The signal resampling module is used to determine the signal-to-noise ratio using a preset threshold, and to re-acquire the original ECG signal when the determination result is that the target ECG signal is invalid.

[0080] In one embodiment of the present invention, it further includes: An electrocardiogram (ECG) sending module is used to send the standard lead ECG to a cloud server and a terminal display device; the cloud server is used to analyze the standard lead ECG, obtain the corresponding ECG analysis results, and send the ECG analysis results to the terminal display device.

[0081] In embodiments of the present invention, a single-channel raw electrocardiogram (ECG) signal is acquired through the main sensor, and an auxiliary reference signal, simultaneously acquired through an auxiliary sensor, interferes with the raw ECG signal. Based on the auxiliary reference signal, interference correction is performed on the raw ECG signal to obtain a target ECG signal. Spatiotemporal features are extracted from the target ECG signal using a pre-trained deep learning model to obtain spatiotemporal features, and lead mapping processing is performed on these features to obtain a standard lead ECG. This achieves the generation of a high-quality ECG based on a single-channel ECG signal, thereby improving the accuracy of reconstructing the standard lead signal and enhancing the convenience and accessibility of ECG monitoring.

[0082] Some embodiments of the present invention also provide an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the method described above.

[0083] Some embodiments of the present invention also provide a computer-readable storage medium on which a computer program is stored, and which, when executed by a processor, implements the method described above.

[0084] Some embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0085] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0086] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0087] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0088] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0090] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0092] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0093] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the aforementioned element.

[0094] The above provides a detailed description of an electrocardiogram processing method and apparatus based on a wearable thin-film sensor. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for processing electrocardiograms based on wearable thin-film sensors, characterized in that, The wearable thin-film sensor includes a main sensor and an auxiliary sensor, and the method includes: Acquire the raw ECG signal from a single channel obtained by the main sensor, and acquire the auxiliary reference signal that interferes with the raw ECG signal by an auxiliary sensor. Based on the auxiliary reference signal, the original electrocardiogram signal is subjected to interference correction to obtain the target electrocardiogram signal; Based on a pre-trained deep learning model, spatiotemporal features are extracted from the target electrocardiogram signal to obtain spatiotemporal features, and lead mapping is performed on the spatiotemporal features to obtain a standard lead electrocardiogram.

2. The method according to claim 1, characterized in that, The spatiotemporal features include local waveform features and long-range rhythm features. The step of performing lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram includes: The local waveform features and the long-range rhythm features are spatiotemporally fused to obtain a three-dimensional feature tensor; the three-dimensional feature tensor is then mapped to the standard lead space to obtain a standard lead electrocardiogram.

3. The method according to claim 1, characterized in that, The wearable thin-film sensor is attached to the user's wrist. The auxiliary sensor is an accelerometer. Based on the data from the auxiliary sensor, interference correction is performed on the original electrocardiogram (ECG) signal to obtain the target ECG signal, including: Based on the auxiliary reference signal, the corresponding acceleration amplitude is obtained; The acceleration amplitude is filtered using a first preset frequency to obtain a motion noise signal; Based on the motion noise signal, motion artifacts in the original ECG signal are eliminated to obtain the target ECG signal.

4. The method according to claim 1, characterized in that, The wearable thin-film sensor is attached to the user's chest, and the auxiliary sensor is a respiratory sensor. Based on the data from the auxiliary sensor, interference correction is performed on the raw electrocardiogram (ECG) signal to obtain the target ECG signal, including: The auxiliary reference signal is filtered according to the second preset frequency to obtain the respiratory fluctuation signal; Based on the baseline fluctuation amplitude of the original electrocardiogram signal, the respiratory fluctuation signal is scaled to obtain a respiratory interference signal; Based on the respiratory interference signal, baseline drift correction is performed on the original electrocardiogram signal to obtain the target electrocardiogram signal.

5. The method according to any one of claims 1-4, characterized in that, Before performing signal correction on the original ECG signal based on the auxiliary reference signal to obtain the target ECG signal, the method further includes: Power frequency interference is detected on the original electrocardiogram signal to obtain the corresponding power frequency interference frequency; The original electrocardiogram (ECG) signal is filtered based on the notch filter corresponding to the power frequency interference frequency to obtain the original ECG signal after power frequency interference suppression.

6. The method according to any one of claims 1-4, characterized in that, After denoising the original ECG signal based on the auxiliary sensor data to obtain the target ECG signal, the process further includes: By employing a sliding window mechanism, the quality of the target ECG signal is assessed, and the signal-to-noise ratio of the target ECG signal is determined. The signal-to-noise ratio is determined using a preset threshold, and if the determination result indicates that the target ECG signal is invalid, the original ECG signal is re-acquired.

7. The method according to any one of claims 1-4, characterized in that, Also includes: The standard lead electrocardiogram is sent to a cloud server and a terminal display device; the cloud server is used to analyze the standard lead electrocardiogram, obtain the corresponding electrocardiogram analysis results, and send the electrocardiogram analysis results to the terminal display device.

8. An electrocardiogram processing device based on a wearable thin-film sensor, characterized in that, The wearable thin-film sensor includes a main sensor and an auxiliary sensor, and the device includes: The signal acquisition module is used to acquire the raw ECG signal of a single channel collected by the main sensor, and to acquire the auxiliary reference signal that interferes with the raw ECG signal by an auxiliary sensor. The signal correction module is used to perform interference correction on the original electrocardiogram signal based on the auxiliary reference signal to obtain the target electrocardiogram signal; The signal reconstruction module is used to extract spatiotemporal features from the target electrocardiogram signal based on a pre-trained deep learning model, obtain spatiotemporal features, and perform lead mapping processing on the spatiotemporal features to obtain a standard lead electrocardiogram.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the method as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, A computer program is stored on the readable storage medium, which, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images