ECG_SegNet: An ECG delineation model based on the encoder-decoder structure

Generalising electrocardiogram detection and delineation: training convolutional neural networks with synthetic data augmentation

Introduction

Extracting beat-by-beat information from electrocardiograms (ECGs) is crucial for various downstream diagnostic tasks that rely on ECG-based measurements. However, these measurements can be expensive and time-consuming to produce, especially for long-term recordings. Traditional ECG detection and delineation methods, relying on classical signal processing algorithms such as those based on wavelet transforms, produce high-quality delineations but struggle to generalise to diverse ECG patterns. Machine learning (ML) techniques based on deep learning algorithms have emerged as promising alternatives, capable of achieving similar performance without handcrafted features or thresholds. However, supervised ML techniques require large annotated datasets for training, and existing datasets for ECG detection/delineation are limited in size and the range of pathological conditions they represent.

Methods

This article addresses this challenge by introducing two key innovations. First, we develop a synthetic data generation scheme that probabilistically constructs unseen ECG traces from "pools" of fundamental segments extracted from existing databases. A set of rules guides the arrangement of these segments into coherent synthetic traces, while expert domain knowledge ensures the realism of the generated traces, increasing the input variability for training the model. Second, we propose two novel segmentation-based loss functions that encourage the accurate prediction of the number of independent ECG structures and promote tighter segmentation boundaries by focusing on a reduced number of samples.

Results

The proposed approach achieves remarkable performance, with a F 1 -score of 99.38% and delineation errors of 2.19 ± 17.73  ms and 4.45 ± 18.32  ms for ECG segment onsets and offsets across the P, QRS, and T waves. These results, aggregated from three diverse freely available databases (QT, LU, and Zhejiang), surpass current state-of-the-art detection and delineation approaches.

Discussion

Notably, the model demonstrated exceptional performance despite variations in lead configurations, sampling frequencies, and represented pathophysiology mechanisms, underscoring its robust generalisation capabilities. Real-world examples, featuring clinical data with various pathologies, illustrate the potential of our approach to streamline ECG analysis across different medical settings, fostered by releasing the codes as open source.

Delineation of 12-Lead ECG Representative Beats Using Convolutional Encoder-Decoders with Residual and Recurrent Connections

The aim of this study is to address the challenge of 12-lead ECG delineation by different encoder-decoder architectures of deep neural networks (DNNs). This study compares four concepts for encoder-decoders based on a fully convolutional architecture (CED-Net) and its modifications with a recurrent layer (CED-LSTM-Net), residual connections between symmetrical encoder and decoder feature maps (CED-U-Net), and sequential residual blocks (CED-Res-Net). All DNNs transform 12-lead representative beats to three diagnostic ECG intervals (P-wave, QRS-complex, QT-interval) used for the global delineation of the representative beat (P-onset, P-offset, QRS-onset, QRS-offset, T-offset). All DNNs were trained and optimized using the large PhysioNet ECG database (PTB-XL) under identical conditions, applying an advanced approach for machine-based supervised learning with a reference algorithm for ECG delineation (ETM, Schiller AG, Baar, Switzerland). The test results indicate that all DNN architectures are equally capable of reproducing the reference delineation algorithm's measurements in the diagnostic PTB database with an average P-wave detection accuracy (96.6%) and time and duration errors: mean values (-2.6 to 2.4 ms) and standard deviations (2.9 to 11.4 ms). The validation according to the standard-based evaluation practices of diagnostic electrocardiographs with the CSE database outlines a CED-Net model, which measures P-duration (2.6 ± 11.0 ms), PQ-interval (0.9 ± 5.8 ms), QRS-duration (-2.4 ± 5.4 ms), and QT-interval (-0.7 ± 10.3 ms), which meet all standard tolerances. Noise tests with high-frequency, low-frequency, and power-line frequency noise (50/60 Hz) confirm that CED-Net, CED-Res-Net, and CED-LSTM-Net are robust to all types of noise, mostly presenting a mean duration error < 2.5 ms when compared to measurements without noise. Reduced noise immunity is observed for the U-net architecture. Comparative analysis with other published studies scores this research within the lower range of time errors, highlighting its competitive performance.

ECGVEDNET: A Variational Encoder-Decoder Network for ECG Delineation in Morphology Variant ECGs

Electrocardiogram (ECG) delineation to identify the fiducial points of ECG segments, plays an important role in cardiovascular diagnosis and care. Whilst deep delineation frameworks have been deployed within the literature, several factors still hinder their development: (a) data availability: the capacity of deep learning models to generalise is limited by the amount of available data; (b) morphology variations: ECG complexes vary, even within the same person, which degrades the performance of conventional deep learning models. To address these concerns, we present a large-scale 12-leads ECG dataset, ICDIRS, to train and evaluate a novel deep delineation model-ECGVEDNET. ICDIRS is a large-scale ECG dataset with 156,145 QRS onset annotations and 156,145 T peak annotations. ECGVEDNET is a novel variational encoder-decoder network designed to address morphology variations. In ECGVEDNET, we construct a well-regularized latent space, in which the latent features of ECG follow a regular distribution and present smaller morphology variations than in the raw data space. Finally, a transfer learning framework is proposed to transfer the knowledge learned on ICDIRS to smaller datasets. On ICDIRS, ECGVEDNET achieves accuracy of 86.28%/88.31% within 5/10 ms tolerance for QRS onset and accuracy of 89.94%/91.16% within 5/10 ms tolerance for T peak. On QTDB, the average time errors computed for QRS onset and T peak are -1.86 ± 8.02 ms and -0.50 ± 12.96 ms, respectively, achieving state-of-the-art performances on both large and small-scale datasets. We will release the source code and the pre-trained model on ICDIRS once accepted.

An improved Bi-LSTM method based on heterogeneous features fusion and attention mechanism for ECG recognition

Electrocardiogram (ECG) plays a critical role in early prevention and diagnosis of cardiovascular diseases. However, extracting powerful deep features from ECG signal for recognition is still a challenging problem today due to the variable abnormal rhythms and noise distribution. This work proposes a Bi-LSTM algorithm based on heterogeneous features fusion and attention mechanism (HFFAM + Bi-LSTM). Combining the empirical features and the features learned by the deep learning network, HFFAM + Bi-LSTM can comprehensively extract the temporal frequency information and spatial structure information of the ECG signal. Meanwhile, a novel attention mechanism based on improved DTW (AM-DTW) is designed to analyze and control the fusion process of features. The role of AM-DTW in HFFAM + Bi-LSTM is twofold, one is to measure the feature similarity between ECG signal sets with different labels using the improved DTW, and the other is to distinguish the features into isomorphic and heterogeneous features as well as adaptive weighting of the features. It is worth mentioning that overly similar isomorphic features are filtered out to further optimize the algorithm. Thus, HFFAM + Bi-LSTM has the advantage of strengthening the heterogeneous information in the feature subspace while accounting for the isomorphic features. The accuracy of HFFAM + Bi-LSTM reaches up to 98.1 % and 97.1 % on the simulated and real datasets, respectively. Compared to the all benchmark models, the classification accuracy of HFFAM + Bi-LSTM is 1.3 % higher than the best. The experiments also demonstrate that HFFAM + Bi-LSTM has better performance compared with existing methods, which provides a new scheme for automatic detection of ECG signal.

SEResUTer: a deep learning approach for accurate ECG signal delineation and atrial fibrillation detection

Objective.Accurate detection of electrocardiogram (ECG) waveforms is crucial for computer-aided diagnosis of cardiac abnormalities. This study introduces SEResUTer, an enhanced deep learning model designed for ECG delineation and atrial fibrillation (AF) detection.Approach. Built upon a U-Net architecture, SEResUTer incorporates ResNet modules and Transformer encoders to replace convolution blocks, resulting in improved optimization and encoding capabilities. A novel masking strategy is proposed to handle incomplete expert annotations. The model is trained on the QT database (QTDB) and evaluated on the Lobachevsky University Electrocardiography Database (LUDB) to assess its generalization performance. Additionally, the model's scope is extended to AF detection using the the China Physiological Signal Challenge 2021 (CPSC2021) and the China Physiological Signal Challenge 2018 (CPSC2018) datasets.Main results.The proposed model surpasses existing traditional and deep learning approaches in ECG waveform delineation on the QTDB. It achieves remarkable average F1 scores of 99.14%, 98.48%, and 98.46% for P wave, QRS wave, and T wave delineation, respectively. Moreover, the model demonstrates exceptional generalization ability on the LUDB, achieving average SE, positive prediction rate, and F1 scores of 99.05%, 94.59%, and 94.62%, respectively. By analyzing RR interval differences and the existence of P waves, our method achieves AF identification with 99.20% accuracy on the CPSC2021 test set and demonstrates strong generalization on CPSC2018 dataset.Significance.The proposed approach enables highly accurate ECG waveform delineation and AF detection, facilitating automated analysis of large-scale ECG recordings and improving the diagnosis of cardiac abnormalities.

Robust electrocardiogram delineation model for automatic morphological abnormality interpretation

Knowledge of electrocardiogram (ECG) wave signals is one of the essential steps in diagnosing heart abnormalities. Considerable performance with respect to obtaining the critical point of a signal waveform (P-QRS-T) through ECG delineation has been achieved in many studies. However, several deficiencies remain regarding previous methods, including the effects of noise interference on the performance degradation of delineation and the role of medical knowledge in reaching a delineation decision. To address these challenges, this paper proposes a robust delineation model based on a convolutional recurrent network with grid search optimization, aiming to classify the precise P-QRS-T waves. In order to make a delineation decision, the results from the ECG waveform classification model are utilized to interpret morphological abnormalities, based on medical knowledge. We generated 36 models, and the model with the best results achieved 99.97% accuracy, 99.92% sensitivity, and 99.93% precision for ECG waveform classification (P-wave, QRS-complex, T-wave, and isoelectric line class). To ensure the model robustness, we evaluated delineation model performance on seven different types of ECG datasets, namely the Lobachevsky University Electrocardiography Database (LUDB), QT Database (QTDB), the PhysioNet/Computing in Cardiology Challenge 2017, China Physiological Signal Challenge 2018, ECG Arrhythmia of Chapman University, MIT-BIH Arrhythmia Database and General Mohammad Hossein Hospital (Indonesia) databases. To detect the patterns of ECG morphological abnormalities through proposed delineation model, we focus on investigating arrhythmias. This process is based on two inputs examination: the P-wave and the regular/irregular rhythm of the RR interval. As the results, the proposed method has considerable capability to interpret the delineation result in cases with artifact noise, baseline drift and abnormal morphologies for delivering robust ECG delineation.

Heartbeat detector from ECG and PPG signals based on wavelet transform and upper envelopes

The analysis of cardiac activity is one of the most common elements for evaluating the state of a subject, either to control possible health risks, sports performance, stress levels, etc. This activity can be recorded using different techniques, with electrocardiogram and photoplethysmogram being the most common. Both techniques make significantly different waveforms, however the first derivative of the photoplethysmographic data produces a signal structurally similar to the electrocardiogram, so any technique focusing on detecting QRS complexes, and thus heartbeats in electrocardiogram, is potentially applicable to photoplethysmogram. In this paper, we develop a technique based on the wavelet transform and envelopes to detect heartbeats in both electrocardiogram and photoplethysmogram. The wavelet transform is used to enhance QRS complexes with respect to other signal elements, while the envelopes are used as an adaptive threshold to determine their temporal location. We compared our approach with three other techniques using electrocardiogram signals from the Physionet database and photoplethysmographic signals from the DEAP database. Our proposal showed better performances when compared to others. When the electrocardiographic signal was considered, the method had an accuracy greater than 99.94%, a true positive rate of 99.96%, and positive prediction value of 99.76%. When photoplethysmographic signals were investigated, an accuracy greater than 99.27%, a true positive rate of 99.98% and positive prediction value of 99.50% were obtained. These results indicate that our proposal can be adapted better to the recording technology.

A Deep Learning-Based Semantic Segmentation Model Using MCNN and Attention Layer for Human Activity Recognition

With the development of wearable devices such as smartwatches, several studies have been conducted on the recognition of various human activities. Various types of data are used, e.g., acceleration data collected using an inertial measurement unit sensor. Most scholars segmented the entire timeseries data with a fixed window size before performing recognition. However, this approach has limitations in performance because the execution time of the human activity is usually unknown. Therefore, there have been many attempts to solve this problem through the method of activity recognition by sliding the classification window along the time axis. In this study, we propose a method for classifying all frames rather than a window-based recognition method. For implementation, features extracted using multiple convolutional neural networks with different kernel sizes were fused and used. In addition, similar to the convolutional block attention module, an attention layer to each channel and spatial level is applied to improve the model recognition performance. To verify the performance of the proposed model and prove the effectiveness of the proposed method on human activity recognition, evaluation experiments were performed. For comparison, models using various basic deep learning modules and models, in which all frames were classified for recognizing a specific wave in electrocardiography data were applied. As a result, the proposed model reported the best F1-score (over 0.9) for all kinds of target activities compared to other deep learning-based recognition models. Further, for the improvement verification of the proposed CEF method, the proposed method was compared with three types of SW method. As a result, the proposed method reported the 0.154 higher F1-score than SW. In the case of the designed model, the F1-score was higher as much as 0.184.

A novel P-QRS-T wave localization method in ECG signals based on hybrid neural networks

As the number of people suffering from cardiovascular diseases increases every year, it becomes essential to have an accurate automatic electrocardiogram (ECG) diagnosis system. Researchers have adopted different methods, such as deep learning, to investigate arrhythmias classification. However, the importance of ECG waveform features is generally ignored when deep learning approaches are applied to classification tasks. P-wave, QRS-wave, and T-wave, containing plenty of physiological information, are three critical waves in the ECG heartbeat. The accurate localization of these critical ECG wave components is a prerequisite for ECG classification and diagnosis. In this study, a novel P-QRS-T wave localization method based on hybrid neural networks is proposed. The raw ECG signal is preprocessed sequentially by filtering, heartbeat extraction, and data standardization. The hybrid neural network is constructed by combining the residual neural network (ResNet) and the Long Short-Term Memory (LSTM). It predicts the relative positions of the P-peak, QRS-peak, and T-peak for each heartbeat. The proposed algorithm was validated on four ECG databases with input noise of different signal-to-noise ratio (SNR) levels. The results show that the proposed method can accurately predict the positions of the three key waves. The proposed P-QRS-T localization approach can improve the efficiency of ECG delineation. Integrated with cardiac disease classification methods, it can contribute to the development of advanced automatic ECG diagnosis systems.