Search

• Compared with previous research on ECG heartbeat classification, the proposed method offers the following four advantages. • We utilize spatio-temporal domain features as expert features to represent contextual information between heartbeats. There are no high-dimensional features with high complexity, and the calculation is relatively simple. • We construct a multi-input model that combines the raw ECG signal with nine expert features, which perfectly matches the actual scenario of ECG diagnosis by physicians. The improved SE-ResNet network can extract long-time series features from ECG signals more effectively. • The proposed method achieves an overall F1-score of 83.39 % for inter-patient heartbeat classification. Satisfactory classification results are obtained for both class S and class F with imbalanced data. • Three open-source databases are used to evaluate the proposed approach and cross-database validation is performed, which demonstrates that our method has good generalization ability. As the number of patients with cardiovascular diseases (CVDs) increases annually, a reliable and automated system for detecting electrocardiogram (ECG) abnormalities is becoming increasingly essential. Scholars have developed numerous methods of arrhythmia classification using machine learning or deep learning. However, the issue of low classification rates of individual classes in inter-patient heartbeat classification remains a challenge. This study proposes a method for inter-patient heartbeat classification by fusing dual-channel squeeze-and-excitation residual neural networks (SE-ResNet) and expert features. In the preprocessing stage, ECG heartbeats extracted from both leads of ECG signals are filtered and normalized. Additionally, nine features representing waveform morphology and heartbeat contextual information are selected to be fused with the deep neural networks. Using different filter and kernel sizes for each block, the SE-residual block-based model can effectively learn long-term features between heartbeats. The divided ECG heartbeats and extracted features are then input to the improved SE-ResNet for training and testing according to the inter-patient scheme. The focal loss is utilized to handle the heartbeat of the imbalance category. The proposed arrhythmia classification method is evaluated on three open-source databases, and it achieved an overall F1-score of 83.39 % in the MIT-BIH database. This system can be applied in the scenario of daily monitoring of ECG and plays a significant role in diagnosing arrhythmias. [ABSTRACT FROM AUTHOR]

• This paper presents a novel deep wavelet convolutional neural network for end-to-end ECG signal denoising. • The proposed DW-CNN avoids QRS waveform detection, which can avoid iterative errors caused by QRS complex detection. Additionally, there is no need to perform any other normalization operation to simplify the preprocess process. • CNN is used as an automatic feature extractor to build an encoder and decoder, which can effectively automatically extract numerous deep features and achieve end-to-end automatic noise reduction. This end-to-end method reaches better denoising performance and generalization ability. • For overcoming the problem of simply discarding the input information caused by the pooling layer, this paper utilizes DWT and IDWT to build DWT-based pooling layer and IDWT-based up-sampling layer, which can retain all the effective information and improve the denoising performance. • Based on the dataset under the real noise conditions, compared with the existing methods, the proposed denoising model has better denoising performance, which reduces 0.194 RMSE and improves 5.99 SNR on average. Recently, more than 80% of sudden cardiac death is caused by arrhythmia, whose incidence has increased rapidly. In the actual wearable device acquisition process, ECG recordings are polluted by many different types of noise, mainly including baseline drift (BW), muscle artifact (MA) and electrode motion (EM). Therefore, the effective de-noising of ECG signals plays an important role in the subsequent accurate arrhythmia diagnosis. Nowadays, traditional filtering methods often only focus on one or two kinds of noise. When encountering different noises, different filtering algorithms need to be used, resulting in insufficient generalization ability of the algorithms. Aiming to effectively denoising ECG signals, this paper proposes a novel signal denoising method based on deep wavelet convolutional neural network. Inspired by the structure of the de-noising self-encoder, the convolution layers are used to replace the simple full-connected layers to build the encoder and decoder. Based on automatic feature extraction in convolution layer, the discrete wavelet transform is used to convert the signal into high-frequency and low-frequency components for replacing pooling layers to compress the input data and fully preserve the effective information. Finally, based on the dataset under the real noise conditions, compared with the existing methods, the proposed denoising model has better denoising performance, which reduces 0.194 RMSE and improves 5.99 SNR on average. Thus, the proposed denoising model has good potential in wearable devices. [ABSTRACT FROM AUTHOR]

• This paper presents a novel dual attentional convolutional long short-term memory neural network to reject noise for ECG quality assessment. • Convolution layers and bidirectional long-short memory layer are integrated to extract features from dual-scale inputs and acquire good classification performance. • Channel-based and time-based attention mechanisms are presented to visually show the attention of the model to different leads and different periods on the multi-leads ECG signals. To our knowledge, it is the first effort for introducing attention mechanism in ECG quality assessment. • The experimental results show that the proposed method achieves 76.47% of specificity, 97.59% of sensitivity, and 94.0% of accuracy, especially improving 3.35% accuracy on average and 4.27% sensitivity on average. ECG quality assessment is of great significance to reduce false alarms in automatic arrhythmia and other cardiovascular diseases diagnoses and reduce the workload of clinicians. Recently, developing an automatic noise rejection algorithm attracts much attention. Nowadays, many researchers have applied deep learning (DL) algorithms into evaluating ECG quality. However, traditional DL approaches improve model accuracy but cannot show the concerned area of ECG signals during detection process. Hence, this paper presents dual attentional convolutional long short-term memory neural network (DAC-LSTM) to evaluate ECG quality. Firstly, convolutional and bidirectional long short-term memory layers are utilized for acquiring numerous deep features from ECG recordings. And then, for enhancing model interpretability, dual-layer attention mechanisms, including channel-based attention mechanism and time-based attention mechanism, are built to visually show the attention of the model to different leads and different periods on the multi-leads ECG signals. Finally, compared with baseline models and the existing methods, DAC-LSTM achieves 76.47% of specificity, 97.59% of sensitivity, and 94.0% of accuracy, especially improving 3.35% accuracy on average and 4.27% sensitivity on average on the commonly used ECG dataset. Generally, DAC-LSTM achieves competitive and interpretable performance and has the potential for practical ECG quality assessment. [ABSTRACT FROM AUTHOR]

Tubing corrosion is becoming more and more serious due to the harsh operating environment and corrosive medium transportation. It results in frequent seal failure of plugs used for tubing lift-off in snubbing operations. This paper develops a finite element model of a plug rubber cylinder sealing a defected tubing. The sealing performance of the rubber cylinder under the defected tubing is studied. The accuracy of model is tested experimentally. The related structural parameters affecting the sealing performance of rubber cylinder are pointed out by numerical derivation. The results show that when selecting the equivalent defected tubing size, smaller depth and radius and larger axial height can be selected. The expression of sealing performance of rubber cylinder in defected tubing environment is obtained. The influence of the key parameters on the sealing performance of the rubber cylinder is the axial height, depth and radius from high to low. The research work in this paper can provide theoretical guidance for the behavior of rubber cylinder seal corroding tubing in snubbing operation. • Performed a finite element analysis of a plug rubber cylinder sealing a defected tubing. • Got the prediction formula of sealing performance of rubber cylinder in the defected tubing. • The influence weights of the axial height, depth and radius of the defected area on the sealing performance of the rubber cylinder were obtained. [ABSTRACT FROM AUTHOR]

• We propose a novel lightweight-modified two-dimensional neural network for heartbeat classification. • It takes the modified VGGNet network as the backbone and contains inverted residual with linear bottleneck module. • Compared with the existing methods, it can reduce the computational complexity and improve classification efficiency. • It could effectively ensure the accuracy of heartbeat classification. • It would be helpful for the application of original two-dimensional ECG in computer-aided diagnosis. Development of lightweight model with high accuracy is an important study in order to better transfer arrhythmia diagnostic models to mobile terminals or embedded devices. A novel lightweight-modified two-dimensional neural network is presented for accurately identifying whether certain types of heartbeats are occurred based on information of heartbeat images. The MIT-BIH arrhythmia database is taken as the dataset. Firstly, data preprocessing is performed that one-dimensional ECG signals are converted into two-dimensional heartbeat images. Then, the obtained heartbeat images are fed into the proposed two-dimensional network for training. Moreover, in order to reduce the network model parameters, we integrate the small-scale convolution kernels and inverted residual with linear bottleneck module into the proposed network. Finally, comparisons of several related classification networks are conducted to verify the effectiveness and superiority of the proposed method. Result shows the classification accuracy is 99.41 % in the test dataset originated in the aforementioned MIT-BIH database. The TPR of APC, LBBB, Normal, RBBB and VPC are 0.995, 1.0, 0.993, 1.0, 0.994, respectively. Compared with VGG16, the number of parameters is reduced by 8, 525, 312, and it achieves the optimal in network complexity and accuracy among the related networks of heartbeat classification. In terms of time consumption, the network proposed is better than some networks suitable for image classification. It takes around 0.0039478 s to complete the heartbeat classification of each ECG image using the proposed method. Result shows the effectiveness of the new method and the simplicity of the model parameters. [ABSTRACT FROM AUTHOR]

Electrocardiogram (ECG) is a critical diagnostic tool used for screening atrial fibrillation (AF). In recent years, several deep learning-based ECG automatic diagnosis methods have been proposed and have yielded satisfactory results. However, the lack of interpretability in these systems limits their clinical usage. To address this issue, we propose a medical explainable AF diagnosis system (M-XAF) that not only predicts AF diagnostic results but also generates an analysis report of the ECG medical features that influenced the classification model's decision. Using a novel ECG signal features occlusion analysis method, M-XAF establishes a connection between the medical knowledge space and the representation space from the convolutional neural network (CNN) by extracting understandable semantic sensitive channels (SSCs) via gradient-weighted channel sensitivity. Through the visualization and statistical modeling of SSCs, M-XAF provides multimodal explanations, including ECG feature descriptions, regions of interest (ROI), and class activation mapping (CAM), for AF diagnostic results. The experimental results demonstrate that the M-XAF has achieved satisfactory results in the AF identification task, with accuracy, specificity, PPV , F1 , and MCC reaching over 0.9715, 0.9896, 0.9778, 0.9567, and 0.9365 on two public ECG datasets. In addition, M-XAF generates quantitative explainable reports and CAM for each predicted sample from the perspective of classification models. Furthermore, the explainable reports provided by M-XAF can address the high-confidence problem in classification models. Ultimately, the explainable framework in M-XAF can be widely applied in AI-ECG classification scenarios. [ABSTRACT FROM AUTHOR]

To achieve optimal performance in practical applications, the electrocardiogram (ECG) diagnosis models have to be personalized using the ECG data of specific patients. Most current research on personalized ECG diagnosis centres around large-sample transfer learning aimed at specific populations or collection conditions. However, the time and economic costs of collecting such a large amount of data are unacceptable for specific individuals. Moreover, the characteristics of ECG signals mean that most data augmentation methods may not produce positive effects. To address these issues, we propose a novel differentiated knowledge distillation (DKD) method, which purposefully transfers specific knowledge in knowledge distillation by inputting differentiated data to the teacher and student models. Specifically, this method constructs differences between the input data of student and teacher models using one single ECG record of a specific patient, so as to transfer individual knowledge of the patient into the student model while maintaining the diagnostic ability of the baseline teacher model. For other ECG data of the target patient, the model personalized by this method shows a significant improvement of up to 10% in accuracy, micro-F1-score, and micro-sensitivity compared to the baseline model. Additionally, this method demonstrates excellent efficacy for various existing networks in the field of ECG diagnosis. [ABSTRACT FROM AUTHOR]