Electrocardiogram Detection Research Articles

Comprehensive Learning Salp Swarm Algorithm with Ensemble Deep Learning-based ECG Signal Classification on Internet of Things Environment

The Internet of Things (IoT) in healthcare relates to implementing interconnected devices and systems for collecting and sharing healthcare information in real time. The integration of IoT in healthcare has the potential to enhance patient outcomes, reduce healthcare costs, and improve the efficacy of medical services. Electrocardiogram (ECG) is a non-invasive heart monitoring method that has become widely accessible due to user-friendly, low-cost, and lead-free wearable heart monitors. However, relying on overworked caregivers for manual monitoring is inefficient. This study develops a Comprehensive Learning Salp Swarm Algorithm with Ensemble Deep Learning (CLSSA-EDL) technique for ECG signal classification in IoT healthcare. The objective of CLSSA-EDL is to detect and classify ECG signals to support decision-making in the IoT healthcare environment. The CLSSA-EDL approach employs the DenseNet201 feature extraction method, with hyperparameters optimally selected by the CLSSA system. For ECG signal detection and classification, an ensemble model using a Stacked Autoencoder (SAE), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM) is utilized. The CLSSA-EDL technique was evaluated on a benchmark ECG dataset, achieving an accuracy of 98.7%, sensitivity of 97.5%, and specificity of 99.1%, demonstrating superior performance compared to recent algorithms.

Accuracy of the Apple Watch in Detecting Atrial Fibrillation Among Patients Undergoing 24-Hour Holter Monitoring: A Prospective, Pragmatic Study

BACKGROUND: As smartwatches with atrial fibrillation detection features gain popularity, it is important to assess the accuracy of these devices to guide decision-making. OBJECTIVES: Our study aimed to assess the sensitivity and specificity of the irregular rhythm notification and the electrocardiogram (ECG)–based detection features of a commonly used smart wearable device (Apple Watch) in detecting atrial fibrillation. METHODS: This was a prospective, pragmatic study conducted in Perpetual Succour Hospital–Cebu Heart Institute from August 2023 to January 2024. To assess the irregular rhythm notification feature, participants were asked to wear an Apple Watch alongside a 24-hour Holter monitor to verify notifications. For the ECG-based detection feature, participants had to tap the crown of the Apple Watch for 30 seconds to get a single-lead ECG similar to a lead I ECG tracing. They were instructed to get manual ECGs hourly, or more often while awake. Irregular rhythm notifications and ECG readings were then compared with that of the 24-hour Holter monitor. Sensitivity and specificity were then computed. RESULTS: A total of 140 participants consented to join after full study disclosure. The irregular rhythm notification feature of the Apple Watch exhibited a low sensitivity of 21.4% but achieved a high specificity of 100% in detecting atrial fibrillation. Meanwhile, the ECG-based detection feature, analyzed from 1295 manually taken ECGs with interpretable sinus rhythm or atrial fibrillation, demonstrated a high level of agreement with the Holter monitor, with a sensitivity of 100% and a specificity of 99.1%. CONCLUSION: The low sensitivity of the irregular rhythm notification feature of the Apple Watch in detecting atrial fibrillation cautions against relying on it as a primary screening tool. However, the high concordance of manually taken Apple Watch ECGs positions the device as a robust tool for detecting atrial fibrillation through manual ECG detection. KEYWORDS: Apple Watch, atrial fibrillation, smartwatches

ECG signal analysis and detection system based on deep learning

In view of the problems of large workload and easy error in traditional manual recognition of ECG signals, and the existing ECG monitoring equipment still has few ECG signal recognition types, low diagnostic accuracy, and excessive reliance on network services, in order to improve the performance of ECG monitoring equipment, an ECG (Electrocardiogram Signals) analysis and detection system is designed based on deep learning technology. The SENet-LSTM (Squeeze-and-Excitation Networks Long Short Term Memory) network model is built to realize automatic diagnosis of 7 categories of ECG signals. The model is built on an intelligent hardware platform that uses ADS1292R as ECG acquisition module, STM32F103 as data processing module, and Raspberry Pi as central processing module. The system uses the integrated high- performance microcomputer Raspberry Pi for calculation and analysis, providing users with offline artificial intelligence (AI: Artificial Intelligence) services. At the same time, the accuracy and precision of the model are and respectively 98.44%, 90.00%thereby realizing real-time detection and accurate classification of ECG, providing patients with accurate disease diagnosis.

An electrocardiogram signal classification using a hybrid machine learning and deep learning approach

An electrocardiogram (ECG) is a diagnostic tool that captures the electrical activity of the heart. Any irregularity in the heart's electrical system is referred to as an arrhythmia, which can be identified through the analysis of ECG signals. Timely diagnosis of cardiac arrhythmias is crucial in order to mitigate their potentially harmful consequences. However, manual analysis of ECG signals is time-consuming and prone to inaccuracies. Therefore, researchers have developed medical decision support systems that utilize machine learning techniques to automate the analysis of ECG signals. In this study, we propose a novel method for classifying ECG signals into four distinct types of heartbeats: normal, supraventricular, ventricular, and fusion. Our method consists of two subsystems that integrate both machine learning and deep learning approaches. The first subsystem uses a residual network block to extract features from the input ECG signal, followed by an LSTM network for learning and classification of these features. The second subsystem uses several feature extraction methods and a random forest to classify the ECG signals. Furthermore, it employs a Synthetic Minority Over-Sampling Technique to improve dataset balance and overall performance. The ultimate result is achieved by merging the results of both subsystems together. An assessment of our approach was carried out on the MIT-BIH dataset, which acts as a recognized ECG signal classification benchmark. Our technique attained an impressive accuracy rate of 99.26%, ranking it as one of the most superior methods in the current literature. Our findings demonstrate the effectiveness and efficiency of our approach in accurately classifying ECG signals for arrhythmia detection.

Catch-AF—Early Diagnosis of Symptomatic Arrythmias in the Waiting Period Prior to Seeing a Cardiologist in Victoria, British Columbia

BackgroundAtrial fibrillation (AF) is the most common cardiac arrhythmia. Given its often-paroxysmal nature, screening at a single time point, using a 12-lead electrocardiogram (ECG) or a Holter monitor, has limited benefit. The AliveCor KardiaMobile device is a validated ECG recorder that can be used for patient-directed arrhythmia diagnosis and symptom–rhythm correlation. The aim of this study was to evaluate whether using the KardiaMobile device could reduce the time-to-diagnosis, for AF as well as other arrhythmias. We hypothesized that providing patients with a KardiaMobile device during their waiting period for specialist care could reduce the length of time that passes before ECG detection of arrhythmia. MethodsPatients were randomized 1:1 to receive either standard monitoring (ECG and a Holter monitor) or enhanced monitoring (ECG, a Holter monitor, and a KardiaMobile device). Patients were instructed to upload ECG recordings if they had cardiac symptoms, so that symptom–rhythm correlation could be achieved. The primary outcome was the time-to-diagnosis for AF. The secondary endpoint was the time-to-diagnosis for any arrhythmias. ResultsFrom October 2018 to October 2022, a total of 69 patients were enrolled, and they were followed up to 12 months. Overall, 6 of the 7 patients diagnosed with AF were in the enhanced-monitoring group (P = 0.106). The time-to-diagnosis was not significantly different in the 2 groups (P = 0.053). Overall arrhythmias were diagnosed in 10 patients (29%) in the standard-monitoring arm, compared to 22 patients (63%) in the enhanced-monitoring arm (P = 0.008). The time-to-diagnosis was reduced in the enhanced-monitoring arm (P = 0.010). ConclusionsThe time-to-diagnosis of any arrhythmia was reduced significantly in patients randomized to receive KardiaMobile device monitoring. Providing patients with a KardiaMobile device may expedite the diagnosis of arrhythmias during the waiting period for specialist care. Clinical Trial RegistrationNCT04302311.

Secure hardware IP of GLRT cascade using color interval graph based embedded fingerprint for ECG detector

This paper presents a security aware design methodology to design secure generalized likelihood ratio test (GLRT) hardware intellectual property (IP) core for electrocardiogram (ECG) detector against IP piracy and fraudulent claim of IP ownership threats. Integrating authentic (secure version) GLRT hardware IP core in the system-on-chip (SoC) of ECG detectors is paramount for reliable operation and estimation of ECG parametric data, such as Q wave, R wave and S wave (QRS) complex detection. A pirated GLRT hardware IP integrated into an ECG detector may result in an unreliable/erratic estimation of ECG parametric data that can be hazardous and fatal for the end patient. The proposed methodology presents an integrated design flow to secure micro GLRT and GLRT cascade hardware IP cores for the ECG detector, using the colored interval graph (CIG) framework based fingerprint biometric, during high level synthesis (HLS). The proposed approach integrates a fingerprint biometric based security constraint generation process for securing the GLRT hardware IP core. This paper also presents a secure register transfer level (RTL) datapath design corresponding to micro GLRT and GLRT cascade hardware IP cores with embedded IP vendor's fingerprint. The proposed secure GLRT hardware IP core embedded with fingerprint biometric achieves superior results in terms of probability of coincidence and tamper tolerance than other security approaches. More explicitly, the proposed approach reports a significantly lower value of probability of coincidence and stronger value for tamper tolerance. Further, the proposed approach incurs zero design cost overhead.

Design and Analysis of a High-Gain, Low-Noise, and Low-Power Analog Front End for Electrocardiogram Acquisition in 45 nm Technology Using gm/ID Method

In this work, an analog front-end (AFE) circuit for an electrocardiogram (ECG) detection system has been designed, implemented, and investigated in an industry-standard Cadence simulation framework using an advanced technology node of 45 nm. The AFE consists of an instrumentation amplifier, a Butterworth band-pass filter (with fifth-order low-pass and second-order high-pass sections), and a second-order notch filter—all are based on two-stage, Miller-compensated operational transconductance amplifiers (OTA). The OTAs have been designed employing the gm/ID methodology. Both the pre-layout and post-layout simulation are carried out. The layout consumes an area of 0.00628 mm2 without the resistors and capacitors. Analysis of various simulation results are carried out for the proposed AFE. The circuit demonstrates a post-layout bandwidth of 239 Hz, with a variable gain between 44 and 58 dB, a notch depth of −56.4 dB at 50.1 Hz, a total harmonic distortion (THD) of −59.65 dB (less than 1%), an input-referred noise spectral density of <34 μVrms/Hz at the pass-band, a dynamic range of 52.71 dB, and a total power consumption of 10.88 μW with a supply of ±0.6 V. Hence, the AFE exhibits the promise of high-quality signal acquisition capability required for portable ECG detection systems in modern healthcare.

Skin Electrodes Based on TPU Fiber Scaffolds with Conductive Nanocomposites with Stretchability, Breathability, and Washability.

In the context of an aging population and escalating work pressures, cardiovascular diseases pose increasing health risks. Electrocardiogram (ECG) monitoring presents a preventive tool, but conventional devices often compromise comfort. This study proposes an approach using Ag NW/TPU composites for flexible and breathable epidermal electronics. In this new structure, TPU fibers are used to support Ag NWs/TPU nanocomposites. The TPU fiber-reinforced Ag NW/TPU (TFRAT) nanocomposites exhibit excellent conductivity, stretchability, and electromechanical durability. The composite ensures high steam permeability, maintaining stable electrical performance after washing cycles. Employing this technology, a flexible ECG detection system is developed, augmented with a convolutional neural network (CNN) for automated signal analysis. The experimental results demonstrate the system's reliability in capturing physiological signals. Additionally, a CNN model trained on ECG data achieves over 99% accuracy in diagnosing arrhythmias. This study presents TFRAT as a promising solution for wearable electronics, offering both comfort and functionality in long-term epidermal applications, with implications for healthcare and beyond.

Universal 12-lead ECG representation for signal denoising and cardiovascular disease detection by fusing generative and contrastive learning

With the wide use of electrocardiogram (ECG) technology, more and more ECGs have been collected and stored. However, ECG labeling is costly and laborious, the utilization of unlabeled ECG is still a critical challenge. Self-supervised learning (SSL) is a way to deal with this problem, which can learn representations from unlabeled ECG and use them for downstream tasks. In this study, a novel SSL model that fuses generative learning (denoising autoencoder, DAE) and contrastive learning (CL) is proposed to learn robust representations from unlabeled ECG for downstream denoising and classification. The 12-lead ECGs are separated into one-dimensional single-lead ECGs and ECG-specific noises are added to the original ECGs as the data augmentation method. To improve the model's denoising and feature extraction abilities, the reconstruction loss and contrastive loss are combined during the pretraining phase. In the downstream classification task, a multi-branch network is used to enhance the correlation between different ECG leads. As a result, it improves the denoising performance over the standard DAE by an average of 5.20%. In the classification tasks for the Chapman, PTB-XL, and CPSC2018 databases, compared to existing work, the accuracies in linear probe are increased by at least 2.49%, 1.09%, and 2.61%, respectively. In addition, an SoC (system-on-a-chip) based heterogeneous deployment scheme is designed. It is 13.69–14.26 times faster in inference than pure software deployment and enables real-time detection of ECG signals. The proposed method provides a good way to utilize unlabeled ECG, and the designed deployment scheme has great potential for medical applications.

Light weight neural network for ECG and EEG anomaly detection in IOT edge sensors

Lightweight neural network designed for detecting anomalies in Electrocardiogram (ECG) and Electroencephalogram (EEG) signals at IoT edge sensors. By optimizing neural network architectures, we achieve high accuracy in anomaly detection while minimizing computational demands and memory usage. Experimental results validate the effectiveness of our approach in real-world scenarios, promising improved healthcare monitoring with early detection of abnormal ECG and EEG patterns at the edge.

MI-OPTNET: AN OPTIMIZED DEEP LEARNING FRAMEWORK FOR MYOCARDIAL INFARCTION DETECTION

The conventional means of myocardial infarction (MI) detection using a 12-lead electrocardiogram (ECG) system include a pretrained network and machine learning interpretation of the complex ECG signals. They are computationally inefficient and demand high-performance hardware. Here, for the first time, we introduce an effective framework (MI-OptNet) using the particle swarm optimization model (PSO) in the design of a lightweight hybrid network combining convolutional neural network (CNN)-long short terms memory (LSTM) for MI and normal ECG detection. We optimized important design and training parameters based on limb leads’ signals and identified leads III and VI as the best ECG leads for the task based on their high classification performance ranging between 80 – 90 %, suggesting that they may provide more information about MI than the others. The other strategy of fusing the scores from all models at the decision level achieved the best result with a 10 % increase in the evaluated metrics. Our findings support the flexibility and adaptability of our framework for the design process using minimal computer efforts. We concluded that this approach may be used for other classification problems to assist engineers and designers in efficient decision-making and to solve complex signal classification and recognition problems.

Advancements in Artificial Intelligence for ECG Signal Analysis and Arrhythmia Detection: A Review

Context: With the widespread availability of portable electrocardiogram (ECG) devices, there is an increasing interest in utilizing artificial intelligence (AI) methods for ECG signal analysis and arrhythmia detection. The potential benefits of AI-assisted arrhythmia prognosis, early screening, and improved accuracy in arrhythmia classification are discussed. Evidence Acquisition: Artificial intelligence methods are a new way to classify different types of arrhythmias. For example, deep learning (DL) algorithms, including long short-term memory (LSTM) networks, convolutional neural networks (CNN), CNN-based autoencoders (AE), and convolutional recurrent neural networks (CRNN), have been extensively utilized for ECG signal analysis and arrhythmia detection. Results: This study explores different DL techniques for classifying arrhythmias. The two-dimensional (2D) CNN model achieved an accuracy of 97.42% in classifying five different arrhythmias. After classifying five types of ECG signals, an accuracy of 99.53% was achieved by the CNN-based AE and transfer learning (TL) models. The CNN-Bi-LSTM model achieved an accuracy of 98.0% in categorizing five categories of ECG signals. The CNN+LSTM model achieved an accuracy of 98.24% in classifying five classes of arrhythmias. The CNN-support vector machine (SVM) classifier model demonstrated an accuracy of 98.64% in detecting seventeen classes of heartbeats. The results indicated that the CNN-based AE and TL models perform exceptionally well with high accuracy in detecting ECG signals. Conclusions: The present study demonstrates the growing interest in utilizing DL for ECG signal detection in medical and healthcare applications over the past decade. Deep learning models have been shown to outperform experienced cardiologists, delivering state-of-the-art and more accurate results.

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.

Deep Learning Autoencoder Study on ECG Signals

Arrhythmia refers to an irregular heart rhythm resulting from disruptions in the heart's electrical activity. To identify arrhythmias, an electrocardiogram (ECG) is commonly employed, as it can record the heart's electrical signals. However, ECGs may encounter interference from sources like electromagnetic waves and electrode motion. Several researchers have investigated the denoising of electrocardiogram signals for arrhythmia detection using deep autoencoder models. Unfortunately, these studies have yielded suboptimal results, indicated by low Signal-to-Noise Ratio (SNR) values and relatively large Root Mean Square Error (RMSE). This study addresses these limitations by proposing the utilization of a Deep LSTM Autoencoder to effectively denoise ECG signals for arrhythmia detection. The model's denoising performance is evaluated based on achieved SNR and RMSE values. The results of the denoising evaluations using the Deep LSTM Autoencoder on the AFDB dataset show SNR and RMSE values of 56.16 and 0.00037, respectively. Meanwhile, for the MITDB dataset, the corresponding values are 65.22 and 0.00018. These findings demonstrate significant improvement compared to previous research. However, it's important to note a limitation in this study—the restricted availability of arrhythmia datasets from MITDB and AFDB. Future researchers are encouraged to explore and acquire a more extensive collection of arrhythmia data to further enhance denoising performance.

Artificial intelligence-enabled 8-lead ECG detection of atrial septal defect among adults: a novel diagnostic tool.

Patients with atrial septal defect (ASD) exhibit distinctive electrocardiogram (ECG) patterns. However, ASD cannot be diagnosed solely based on these differences. Artificial intelligence (AI) has been widely used for specifically diagnosing cardiovascular diseases other than arrhythmia. Our study aimed to develop an artificial intelligence-enabled 8-lead ECG to detect ASD among adults. In this study, our AI model was trained and validated using 526 ECGs from patients with ASD and 2,124 ECGs from a control group with a normal cardiac structure in our hospital. External testing was conducted at Wuhan Central Hospital, involving 50 ECGs from the ASD group and 46 ECGs from the normal group. The model was based on a convolutional neural network (CNN) with a residual network to classify 8-lead ECG data into either the ASD or normal group. We employed a 10-fold cross-validation approach. Statistically significant differences (p < 0.05) were observed in the cited ECG features between the ASD and normal groups. Our AI model performed well in identifying ECGs in both the ASD group [accuracy of 0.97, precision of 0.90, recall of 0.97, specificity of 0.97, F1 score of 0.93, and area under the curve (AUC) of 0.99] and the normal group within the training and validation datasets from our hospital. Furthermore, these corresponding indices performed impressively in the external test data set with the accuracy of 0.82, precision of 0.90, recall of 0.74, specificity of 0.91, F1 score of 0.81 and the AUC of 0.87. And the series of experiments of subgroups to discuss specific clinic situations associated to this issue was remarkable as well. An ECG-based detection of ASD using an artificial intelligence algorithm can be achieved with high diagnostic performance, and it shows great clinical promise. Our research on AI-enabled 8-lead ECG detection of ASD in adults is expected to provide robust references for early detection of ASD, healthy pregnancies, and related decision-making. A lower number of leads is also more favorable for the application of portable devices, which it is expected that this technology will bring significant economic and societal benefits.

HA-ResNet: Residual Neural Network With Hidden Attention for ECG Arrhythmia Detection Using Two-Dimensional Signal.

Arrhythmia is an abnormal heart rhythm, a common clinical problem in cardiology. Long-term or severe arrhythmia may lead to stroke and sudden cardiac death. The electrocardiogram (ECG) is the most commonly used tool to diagnose arrhythmia. However, the traditional diagnosis relies on experts for manual interpretation, which is time-consuming and laborious. In recent years, many automatic arrhythmia detection methods have emerged due to advancements in deep learning. These methods can reduce manual intervention and improve diagnostic efficiency. However, extracting useful features from raw ECG signals for arrhythmia detection is still challenging due to the low frequency of ECG signals and noise distribution. In this paper, we propose a novel hidden attention residual network (HA-ResNet) for automated arrhythmia classification. In this model, the one-dimensional ECG signals are first converted into two-dimensional images and fed into an embedding layer to obtain the relevant shallow features in ECG. Then, a hidden attention layer combining Squeeze-and-Excitation (SE) block and Bidirectional Convolutional LSTM (BConvLSTM) is used to further capture the deep Spatio-temporal features. We evaluate our HA-ResNet on two public datasets and achieve F1 scores of 96.0%, 96.7%, and 87.6% on 2s segments, 5s segments, and 10s segments, respectively, which significantly outperform the existing state-of-the-art approaches. The experimental results demonstrate the effectiveness and generalization of our method.

IM-ECG: An interpretable framework for arrhythmia detection using multi-lead ECG

Multi-lead electrocardiogram (ECG) is a fundamental and reliable diagnostic tool for the detection of heart arrhythmias. An increasing number of deep neural network models have been proposed for automatic arrhythmia detection using multi-lead ECGs. However, most of the existing models focus on intra-lead features while neglecting the inter-lead ones and fail to explain the reasons behind decisions due to their black-box nature. To address these challenges, we propose an interpretable multi-lead ECG detection framework that tackles these limitations in two key areas. First, a flexible double-kernel residual block (DKR-block) is proposed to effectively extract inter-lead and intra-lead features. In addition, a new two-dimensional ECG classification model is built based on the DKR-block for accurate arrhythmia detection. Second, a visualization method based on gradient-weighted class activation mapping is used to visualize the indicative features of the model’s decisions. The proposed model achieves an AUCmacro of 0.929 on the PTB-XL data, a F1micro of 0.932 on the HFECGIC dataset, and an F1macro of 0.989 on the LUDB dataset. The results show that the proposed model outperforms several state-of-the-art methods, and the visualization features of its classification decision are consistent with the medical background knowledge.

ECG based Categorical emotion classification using time-domain features and Machine Learning

Abstract Emotions are mental states that result from neurophysiological changes associated with thoughts, feelings, and behavioral responses. Emotions lead to modifications in heart rate variability, which can be identified through electrocardiogram (ECG) signals. In this study, we attempted to analyze the ECG signals to detect categorical emotions using time-domain features and machine-learning algorithms. Initially, the ECG signals of 30 subjects were obtained from the publicly available continuously annotated signals of emotion dataset. Further, the signals were preprocessed and extracted 32-time domain features from ECG signals which were recorded during different emotional states such as amusing, boring, relaxing, and scary. The extracted features were fed to a random forest (RF) classifier to rank the features and to build the three machine learning models such as logistic regression (LR), support vector machine, and RF. We achieved the highest average classification accuracy, sensitivity, specificity, precision, and f1-score of 71.04%, 42.08%, 80.69%, 43.03%, and 42.32%, respectively, with the top 4 features using the LR classifier. We found that the mean of peaks, slope sign change, dynamic range, and mean of first derivative were ranked top and played a significant role in the classification model. Our study shows the effectiveness of utilizing ECG signals for emotion detection in wearable devices.

An 82-nW 0.53-pJ/SOP Clock-Free Spiking Neural Network With 40-μs Latency for AIoT Wake-Up Functions Using a Multilevel-Event-Driven Bionic Architecture and Computing-in-Memory Technique

This article presents a clock-free spiking neural network (SNN) intelligent inference engine (IIE) for artificial intelligence of things (AIoT) sensor nodes, which often operate in random-sparse-event (RSE) scenarios. The IIE drastically reduces the system’s long-term average (LTA) power consumption, improves energy efficiency, and achieves microsecond level inference latency. Three techniques are proposed: 1) A clock-free SNN architecture without clock tree, frame generator, and arbiter, is driven by the output spikes, which are encoded with level-crossing (LC) sampling method; the circuit activity is completely related to event activity and spike rates, dramatically reducing the overall power consumption and latency. 2) The bioinspired leaky-integrate-fire (LIF) neurons directly extract the time-domain information from asynchronous spikes, reducing the network size and number of operations. 3) The computing-in-memory (CIM) and mixed-signal synapse-neuron circuits are employed to increase the SNN parallelism and avoid weight movements, thus improving the energy efficiency and response speed. The measured LTA power is bounded at 82 nW while the event-driven chip is on call and waiting for events; the energy efficiency is 0.53 pJ per synapse operation (SOP), only 1/3 that of state-of-the-art methods at 4bit weights even with 180 nm technology. We demonstrate electrocardiogram (ECG) recognition as a typical AIoT application, and the power consumption is less than 350 nW. The measured accuracy of abnormal ECG detection is 90.5%. Moreover, the latency is only 40 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu $</tex-math> </inline-formula> s to realize real-time NN inference. This work provides an effective solution for AIoT nodes that require both ultralow power and fast response.

Electrocardiogram signal classification based on fusion method of residual network and self-attention mechanism

In the diagnosis of cardiovascular diseases, the analysis of electrocardiogram (ECG) signals has always played a crucial role. At present, how to effectively identify abnormal heart beats by algorithms is still a difficult task in the field of ECG signal analysis. Based on this, a classification model that automatically identifies abnormal heartbeats based on deep residual network (ResNet) and self-attention mechanism was proposed. Firstly, this paper designed an 18-layer convolutional neural network (CNN) based on the residual structure, which helped model fully extract the local features. Then, the bi-directional gated recurrent unit (BiGRU) was used to explore the temporal correlation for further obtaining the temporal features. Finally, the self-attention mechanism was built to weight important information and enhance model's ability to extract important features, which helped model achieve higher classification accuracy. In addition, in order to mitigate the interference on classification performance due to data imbalance, the study utilized multiple approaches for data augmentation. The experimental data in this study came from the arrhythmia database constructed by MIT and Beth Israel Hospital (MIT-BIH), and the final results showed that the proposed model achieved an overall accuracy of 98.33% on the original dataset and 99.12% on the optimized dataset, which demonstrated that the proposed model can achieve good performance in ECG signal classification, and possessed potential value for application to portable ECG detection devices.