ECG Database Research Articles

A lightweight U-Net model for denoising and noise localization of ECG signals

ObjectiveElectrocardiogram (ECG) is a diagnostic tool for cardiovascular disease in clinical medicine, but noisy ECG signals can easily lead to misdiagnosis. This paper presents a lightweight U-Net model (LUNet) for denoising and noise localization of ECG signals of varying lengths. MethodsThe proposed model acts as a denoising autoencoder, which accepts an ECG signal containing noises and outputs the denoised signal. The locations of unacceptable noise are determined by comparing the difference between the input and output signals of the model. ResultsThe proposed method for ECG signal denoising achieved Root Mean Square Error (RMSE) scores of 0.0116 ± 0.0016, 0.0149 ± 0.0033, 0.0122 ± 0.0015, and 0.0098 ± 0.0012 for baseline wander (BW), muscle artifacts (MA), electrode motion artifacts (EM), and Additive white Gaussian noise (AWGN), respectively. For ECG signal noise localization, the proposed method obtained high Intersection over Union (IoU) scores and Dice scores. Meanwhile, our model showed excellent generalization performance on two public ECG databases. On the 12-lead ECG arrhythmia classification task, the results using the processed dataset with the proposed method were 2.4% higher in F1 score than using the original dataset. ConclusionThe proposed method is highly effective in removing acceptable noise from ECG signals of varying lengths and locating the positions of unacceptable noise. SignificanceThis paper presents a promising solution for denoising and noise localization of ECG signals, which can improve ECG interpretation and potentially lead to better diagnosis and treatment of cardiovascular diseases.

Heart condition determination based on MET value to NYHA Classification and abnormal ST segment identification

Cardiovascular diseases have always been among the top causes of death. Thus, there were variety of research focuses on cardiac stress test and on self-stress test as well. However, the portable device existed for heart monitoring were still insufficient where most device are for fitness monitoring rather than functional capacity. Thus, this paper determines heart condition based on New York Heart Classification (NYHA) for The Rockport Walking Fitness Test (TRWFT). TRWFT gave out parameters to calculate maximum oxygen consumption (VO2max) to find metabolic equivalent (MET) value. The MET values were compared with NYHA functional capacity classes. Hence, the comparison simplified the heart condition to the urgency to visit hospitals. Plus, ST segments of MIT-BIH ECG database were extracted and assessed for abnormality. Accordingly, a mobile application was developed. The prototype is built using Genuino 101 microcontroller, an AD8232 ECG module with leads to sense heart pulse, and a SD card module. Meanwhile, simulation test on the prototype shows that the prototype succeeds in produce the same values with manual calculation and managed to assess ST segments as programmed. In conclusion, a prototype to demonstrate the implementation of TRFWT and prediction of cardiac condition based on MET value is successful.

Detecting anomalies in fetal electrocardiogram records using deep learning models

In order to assess the fetus health and make timely decisions throughout pregnancy, Fetal Electrocardiography (FECG) monitoring is essential. Huge datasets for electrocardiograms are freely accessible from Physionet ATM Dataset1- Abdominal and Direct Fetal ECG Database (adfecgdb), Dataset2- Fetal ECG Synthetic Database (fecgsyndb), Dataset3- Non-Invasive Fetal ECG Database(nifecgdb). In this study, we strive to categorize normal and abnormal (Atrial fibrillation) FECG from three online dataset which contains FECG recordings as major details. Deep learning models like Transfer Learning (TL) and Convolutional Neural Networks (CNN) are being investigated. The composite abdominal signal and the FECG are separated using a wavelet transform approach. The best model for categorizing the parameters of the FECG is determined through a comparative analysis and performance is improved using Continuous Wavelet Transform (CWT). The accuracy of the CNN-based technique is found to be 98.59%, whereas the accuracy of the transfer learning model is 99.01% for FECG classification. The computation of metric parameters for all the datasets is done. The classification of normal and abnormal (Atrial fibrillation) is best performed in TL model compared to CNN. Real-time data analysis is done for PQRST plotting and comparative study is done using Net Reclassification Improvement (NRI) and obtained NRI = 13%, z static 0f 3.7641, p-Value of 0.00016721. Acute Myocardial Infraction (AMI) identification is done based on ST segment of Maternal ECG (MECG) images to analyze the heart attack risk. The proposed work can be utilized to track FECG waveforms in real-time for wearable technology because of its end-to-end properties and expandable intrinsic for diagnosing multi-lead heart disorders.

Evaluating Goldberger's triad as a marker of congestive heart failure: A systematic review and retrospective study

BackgroundIn 1982, Drs. Barold and Goldberger described an ECG triad associated with left ventricular dysfunction (LVD) consisting of high precordial QRS voltage, low limb lead voltage, and poor precordial R wave progression. Studies have since attempted to replicate the originally reported sensitivity (70%), specificity (>99%), and positive predictive value (PPV, 100%) of Goldberger's triad (GT) with variable results. PurposeTo assess sensitivity, specificity and PPV of GT as a screening tool for LVD in the current era. MethodsWe performed: (1) A systematic review of the published studies; (2) Searched our hospital ECG database (GE MUSE) for diagnoses of “low limb-voltage” and “left ventricular hypertrophy” from 2017 to 2022; identified ECGs were analyzed for GT criteria and their medical records were screened for LVD. (3) ECG analysis of patients with known idiopathic LVD for the GT. ResultsA total of 11,115 patients from 8 studies were included in the systematic review of published studies and showed widely varying sensitivity, specificity and PPV. A total of 4576 ECGs (in GE MUSE) from 372 patients met initial screening criteria of low limb lead voltage and LVH; only 12 patients had ECGs that satisfied GT. Of these 12, only 1 patient had evidence of LVD, yielding a PPV of 8%. Finally, of the 40 patients with known LVD, only 1 met the ECG criteria for GT, resulting in a sensitivity of 2.5%. ConclusionOur literature review does not support the original results of GT. ECGs from our database that met GT (searched by low limb-voltage and left ventricular hypertrophy) over a span of 5 years were rare. When present, the PPV of GT was 8%. In patients with established LVD, the sensitivity was 2.5%. These data do not validate GT as tool to identify LVD in the current era.

Post hoc labeling an acute ECG as ischemic or non-ischemic based on clinical data: A necessary challenge

The ECG is crucial in the prehospital (and early inhospital) phase of patients with symptoms suggestive of myocardial ischemia. Therefore, new algorithms for ECG-based myocardial ischemia detection are continuously being researched. Development and validation of these algorithms require a database of acute ECGs (from the prehospital or emergency department setting) including a representative mix of cases (ischemia present) and controls (no ischemia present). Therefore, for every patient in this mix, the “truth” regarding the actual presence or absence of myocardial ischemia during the recording of the acute ECG has to be determined to compare the newly developed algorithm against. This post hoc adjudication process of determining whether an acute (either prehospitally acquired or acquired in the emergency department) ECG was made under ischemic conditions should use all available clinical data (the clinical diagnosis, cardiac imaging data, and laboratory values) of the subsequent patient's admission.Even with all data at hand, post hoc labeling a patient and their acute ECG as a myocardial ischemia case or control cannot be forced into a binary division between definite cases and definite controls. More specifically, to be used for the development of a new algorithm, the patients' ECG has to be scored for the presence or absence of myocardial ischemia at the exact moment of its recording, which renders the classification even more difficult. For instance, even though it may be plausible that myocardial ischemia was present at a given moment during the patient's admission, this is not necessarily proof that the prehospital (or early inhospital) ECG was also made in ischemic conditions: ischemia can be a fluctuating process (as is, e.g., the case in unstable angina pectoris). Therefore, post hoc classification of an acute ECG in terms of the absence or presence of ischemia requires a multipoint scale ranging between definite ischemic to definite non-ischemic, for instance using a 5-point scale (presumed non-ischemic, probably non-ischemic, uncertain, probably ischemic, presumed ischemic).To summarize, the post hoc adjudication process of ECGs of ambulance (and emergency department) patients cannot result in a binary division into definite cases and controls (i.e., patients with or without myocardial ischemia during the recording of the acute ECG), as myocardial ischemia is often dynamic rather than constant. ECGs could be labeled on a multi-point scale, in which the label represents the probability of the actual presence (or absence) of myocardial ischemia at the exact moment of the recording of that ECG. Further development of algorithms for myocardial ischemia detection should consider this concept.

A novel deep learning technique for morphology preserved fetal ECG extraction from mother ECG using 1D-CycleGAN

The non-invasive fetal electrocardiogram (fECG) enables easy detection of developing heart abnormalities, leading to a significant reduction in infant mortality rate and post-natal complications. Due to the overlapping of maternal and fetal R-peaks, the low amplitude of the fECG, systematic and ambient noises, typical signal extraction methods, such as adaptive filters, independent component analysis, empirical mode decomposition, etc., are unable to produce satisfactory fECG. While some techniques can produce accurate QRS waves, they often ignore other important aspects of the ECG. Utilizing extensive preprocessing and an appropriate framework, our approach, built upon 1D CycleGAN, achieves fECG signal reconstruction from the mECG signal while preserving its morphology. The performance of our solution was evaluated by combining two available datasets from Physionet, “Abdominal and Direct Fetal ECG Database” and “Fetal electrocardiograms, direct and abdominal with reference heartbeat annotations”, where it achieved an average PCC and Spectral-Correlation score of 88.4% and 89.4%, respectively. It detects the fQRS of the signal with accuracy, precision, recall and F1 score of 92.6%, 97.6%, 94.8% and 96.4%, respectively. It can also accurately produce the estimation of fetal heart rate and R-R interval with an error of 0.25% and 0.27%, respectively. The main contribution of our work is that, unlike similar studies, it can retain the morphology of the ECG signal with high fidelity. The accuracy of our solution for fetal heart rate and R-R interval length is comparable to existing state-of-the-art techniques. This makes it a highly effective tool for early diagnosis of fetal heart diseases and regular health checkups of the fetus.

Left-bundle branch block, an alert for cardiac resynchronization therapy Characteristics of an unselected population with LBBB in Stockholm County.

The left bundle branch block (LBBB) might be the first finding of cardiovascular diseases but also the prerequisite for cardiac resynchronization therapy (CRT) in heart failure (HF) with reduced ejection fraction (HFrEF). The prognosis for patients with LBBB and the implications of CRT in an unselected real-world setting are of great interest. A central ECG database and national registers have been screened to identify patients with LBBB. Predictors of HF and use of CRT were identified with Cox models. The hazard ratios (HR) of death, cardiovascular death (CVD) and HF hospitalization (HFH) were estimated according to CRT use. Of 5,359 patients with LBBB and QRS>150ms, median age 76 years, 36% were female. At the time of index ECG, 41% had a previous history of HF and 27% developed HF. Among 1053 patients with a class I indication for CRT, only 60% received the CRT with a median delay of 137 days and it was associated with a lower risk of death (HR: 0.45 95% CI: 0.36-0.57), CVD (HR: 0.47 95% CI: 0.35-0.63) and HFH (HR: 0.56 CI: 0.48-0.66). The age over 75 years and the diagnosis of dementia and chronic obstructive pulmonary disease were predictors of CRT non-use while having a pacing/defibrillator device independently predicted CRT use. In an unselected LBBB population, CRT is underused but of great value for HF patients. Therefore, it is crucial to find ways of better implementing and understanding CRT utilization and characteristics that influence the management of our patients.

Left-bundle branch block, an alert for cardiac resynchronization therapy

Abstract Funding Acknowledgements Type of funding sources: Private company. Main funding source(s): Medtronic. Background Left bundle branch block (LBBB) might be the first finding of cardiovascular diseases but it is also the target of cardiac resynchronization therapy (CRT) in heart failure (HF) with reduced ejection fraction (HFrEF). LBBB prognosis and the implication of CRT in an unselected real-world setting are of great potential clinical impact. Methods A central ECG database together with national registries has been screened to identify LBBB patients. Cox models were fitted to assess the hazard ratio (HR) of death, cardiovascular death (CVD) and HF hospitalization (HFH) according to gender and CRT use. A subdistribution of the hazard ratio (SHR) was used to account for non-cardiovascular death as a competing risk. Logistic regression was fitted to investigate characteristics associated with CRT use. Results Of 5359 patients with LBBB and QRS duration over 150ms, 36% were female. The median age was 76 years and males had a significantly higher risk of death after 5 years from the LBBB diagnosis. CRT, when indicated, has a large benefit on all-causes of death (HR: 0.53, 95% confidence intervals (CI): 0.42-0.67), CVD (HR: 0.53 CI: 0.39-0.71) and HFH (HR: 0.71 CI: 0.60-0.82), especially in the first 5 years after the indication. These results are consistent with the competing risk analysis. Conclusion In an unselected LBBB population, CRT is underused but extremely beneficial. Therefore it’s crucial to find ways of better implementing and understanding CRT utilization, focusing on characteristics that influence recommendations.

Normal ECG standards and percentile charts for infants, children and adolescents.

Normal ECG standards in newborns, infants, children and adolescents have been collected and published by many authors. Only those by Davignon etal., Rijinbeek etal. and our two studies covered all ages from birth to adolescence. The standards reflecting the growth and development of the heart in infants, children and adolescents remained to be studied and explored. We selected from our ECG database, after discussions and consultation, 15 key ECG parameters and analyzed for their age- and sex-specific mean, standard deviation and 2nd to 98th percentiles and their percentile charts were constructed. The ranges and distributions of the normal ECG standards, means and 2nd to 98th percentiles of 15 key parameters were established. A complete set of normal ECG standards of 15 key parameters from birth to adolescents is available for clinicians and researchers.

AI-Assisted QT Measurements for Highly Automated Drug Safety Studies.

Rate-corrected QT interval (QTc) prolongation has been suggested as a biomarker for the risk of drug-induced torsades de pointes, and is therefore monitored during clinical trials for the assessment of drug safety. Manual QT measurements by expert ECG analysts are expensive, laborious and prone to errors. Wavelet-based delineators and other automatic methods do not generalize well to different T wave morphologies and may require laborious tuning. Our study investigates the robustness of convolutional neural networks (CNNs) for QT measurement. We trained 3 CNN-based deep learning models on a private ECG database with human expert-annotated QT intervals. Among these models, we propose a U-Net model, which is widely used for segmentation tasks, to build a novel clinically useful QT estimator that includes QT delineation for better interpretability. We tested the 3 models on four external databases, amongst which a clinical trial investigating four drugs. Our results show that the deep learning models are in stronger agreement with the experts than the state-of-the-art wavelet-based algorithm. Indeed, the deep learning models yielded up to 71% of accurate QT measurements (absolute difference between manual and automatic QT below 15 ms) whereas the wavelet-based algorithm only allowed 52% of QT accuracy. For the 2 studies of drugs with small to no QT prolonging effect, a mean absolute difference of 6 ms (std = 5 ms) was obtained between the manual and deep learning methods. For the other 2 drugs with more significant effect on the volunteers, a mean difference of up to 17 ms (std = 17 ms) was obtained. The proposed models are therefore promising for automated QT measurements during clinical trials. They can analyze various ECG morphologies from a diversity of individuals although some QT-prolonged ECGs can be challenging. The U-Net model is particularly interesting for our application as it facilitates expert review of automatic QT intervals, which is still required by regulatory bodies, by providing QRS onset and T offset positions that are consistent with the estimated QT intervals.

Analysis on population-based algorithm optimized filter for non-invasive fECG extraction

Metaheuristic algorithms (MAs) have become one of the primary tools for optimization in diverse domains, including non-invasive fetal electrocardiogram (fECG) extraction. Research reveals that hyperparameters affect the performance of MAs differently to problems, and some algorithms are problem-specific designed and may produce bad results for different problems. Hence, three questions arise, (1) how much can we trust MAs when solving boxed-constrained fECG extraction problems, and (2) which type of MAs are suitable and adequate for fECG extraction? (3) do MAs find acceptable solutions to a problem that does not formulate the problem comprehensively with an imperfect objective function? This paper focuses on these three inquiries and proposes a framework providing an MA-assisted adaptive filter for non-invasive fECG extraction. The proposed framework has three key components: data pre-processing, MA-based adaptive filter, and post-processing. The pre-processing starts to process the signals detected by pregnant women and extract the desired signal by independent component analysis. Such output signals are then passed to a filter and optimized by population-based algorithms, producing an optimal solution. After that, this solution, as filter weights, will be used for signal extraction and be processed by post-processing. Importantly, the proposed framework is user-friendly that can import any MA algorithm to run and disassemble as a separate assisting tool. This work investigated eight classic and disparate MA algorithms on the Abdominal and Direct Fetal ECG Database (ADFECGDB) dataset, offering comprehensive experimental analysis. We infer from the results that the recordings r01,r02,r03,r05,r08, and r09, where the signals are less noisy and can be solved well. While for recordings r04,r07,r06, and r10, the performance can be influenced by hyperparameters for any test algorithms, such as the population size, window size, and other parameters in MAs. Meanwhile, we found that MA may not be the primary target influencing the performance vary.

Deep Learning-Based ECG Arrhythmia Classification: A Systematic Review

Deep learning (DL) has been introduced in automatic heart-abnormality classification using ECG signals, while its application in practical medical procedures is limited. A systematic review is performed from perspectives of the ECG database, preprocessing, DL methodology, evaluation paradigm, performance metric, and code availability to identify research trends, challenges, and opportunities for DL-based ECG arrhythmia classification. Specifically, 368 studies meeting the eligibility criteria are included. A total of 223 (61%) studies use MIT-BIH Arrhythmia Database to design DL models. A total of 138 (38%) studies considered removing noise or artifacts in ECG signals, and 102 (28%) studies performed data augmentation to extend the minority arrhythmia categories. Convolutional neural networks are the dominant models (58.7%, 216) used in the reviewed studies while growing studies have integrated multiple DL structures in recent years. A total of 319 (86.7%) and 38 (10.3%) studies explicitly mention their evaluation paradigms, i.e., intra- and inter-patient paradigms, respectively, where notable performance degradation is observed in the inter-patient paradigm. Compared to the overall accuracy, the average F1 score, sensitivity, and precision are significantly lower in the selected studies. To implement the DL-based ECG classification in real clinical scenarios, leveraging diverse ECG databases, designing advanced denoising and data augmentation techniques, integrating novel DL models, and deeper investigation in the inter-patient paradigm could be future research opportunities.

Imbalanced Ectopic Beat Classification Using a Low-Memory-Usage CNN LMUEBCNet and Correlation-Based ECG Signal Oversampling

Objective: This study presents a low-memory-usage ectopic beat classification convolutional neural network (CNN) (LMUEBCNet) and a correlation-based oversampling (Corr-OS) method for ectopic beat data augmentation. Methods: A LMUEBCNet classifier consists of four VGG-based convolution layers and two fully connected layers with the continuous wavelet transform (CWT) spectrogram of a QRS complex (0.712 s) segment as the input of the LMUEBCNet. A Corr-OS method augmented a synthetic beat using the top K correlation heartbeat of all mixed subjects for balancing the training set. This study validates data via a 10-fold cross-validation in the following three scenarios: training/testing with native data (CV1), training/testing with augmented data (CV2), and training with augmented data but testing with native data (CV3). Experiments: The PhysioNet MIT-BIH arrhythmia ECG database was used for verifying the proposed algorithm. This database consists of a total of 109,443 heartbeats categorized into five classes according to AAMI EC57: non-ectopic beats (N), supraventricular ectopic beats (S), ventricular ectopic beats (V), a fusion of ventricular and normal beats (F), and unknown beats (Q), with 90,586/2781/7236/803/8039 heartbeats, respectively. Three pre-trained CNNs: AlexNet/ResNet18/VGG19 were utilized in this study to compare the ectopic beat classification performance of the LMUEBCNet. The effectiveness of using Corr-OS data augmentation was determined by comparing (1) with/without using the Corr-OS method and (2) the Next-OS data augmentation method. Next-OS augmented the synthetic beat using the next heartbeat of one subject. Results: The proposed LMUEBCNet can achieve a 99.4% classification accuracy under the CV2 and CV3 cross-validation scenarios. The accuracy of the proposed LMUEBCNet is 0.4–0.5% less than the performance obtained from AlexNet/ResNet18/VGG19 under the same data augmentation and cross-validation scenario, but the parameter usage is only 10% or less than that of the AlexNet/ResNet18/VGG19 method. The proposed Corr-OS method can improve ectopic beat classification accuracy by 0.3%. Conclusion: This study developed a LMUEBCNet that can achieve a high ectopic beat classification accuracy with efficient parameter usage and utilized the Corr-OS method for balancing datasets to improve the classification performance.

Bimodal CNN for cardiovascular disease classification by co-training ECG grayscale images and scalograms

This study aimed to develop a bimodal convolutional neural network (CNN) by co-training grayscale images and scalograms of ECG for cardiovascular disease classification. The bimodal CNN model was developed using a 12-lead ECG database collected from Chapman University and Shaoxing People's Hospital. The preprocessed database contains 10,588 ECG data and 11 heart rhythms labeled by a specialist physician. The preprocessed one-dimensional ECG signals were converted into two-dimensional grayscale images and scalograms, which are fed simultaneously to the bimodal CNN model as dual input images. The proposed model aims to improve the performance of CVDs classification by making use of ECG grayscale images and scalograms. The bimodal CNN model consists of two identical Inception-v3 backbone models, which were pre-trained on the ImageNet database. The proposed model was fine-tuned with 6780 dual-input images, validated with 1694 dual-input images, and tested on 2114 dual-input images. The bimodal CNN model using two identical Inception-v3 backbones achieved best AUC (0.992), accuracy (95.08%), sensitivity (0.942), precision (0.946) and F1-score (0.944) in lead II. Ensemble model of all leads obtained AUC (0.994), accuracy (95.74%), sensitivity (0.950), precision (0.953), and F1-score (0.952). The bimodal CNN model showed better diagnostic performance than logistic regression, XGBoost, LSTM, single CNN model training with grayscale images alone or with scalograms alone. The proposed bimodal CNN model would be of great help in diagnosing cardiovascular diseases.

Atrial fibrillation classification based on the 2D representation of minimal subset ECG and a non-deep neural network.

Atrial fibrillation (AF) is the most common cardiac arrhythmia, and its early detection is critical for preventing complications and optimizing treatment. In this study, a novel AF prediction method is proposed, which is based on investigating a subset of the 12-lead ECG data using a recurrent plot and ParNet-adv model. The minimal subset of ECG leads (II &V1) is determined via a forward stepwise selection procedure, and the selected 1D ECG data is transformed into 2D recurrence plot (RP) images as an input to train a shallow ParNet-adv Network for AF prediction. In this study, the proposed method achieved F1 score of 0.9763, Precision of 0.9654, Recall of 0.9875, Specificity of 0.9646, and Accuracy of 0.9760, which significantly outperformed solutions based on single leads and complete 12 leads. When studying several ECG datasets, including the CPSC and Georgia ECG databases of the PhysioNet/Computing in Cardiology Challenge 2020, the new method achieved F1 score of 0.9693 and 0.8660, respectively. The results suggested a good generalization of the proposed method. Compared with several state-of-art frameworks, the proposed model with a shallow network of only 12 depths and asymmetric convolutions achieved the highest average F1 score. Extensive experimental studies proved that the proposed method has a high potential for AF prediction in clinical and particularly wearable applications.

Prediction of Arrhythmias and Acute Myocardial Infarctions using Machine Learning

Cardiovascular diseases such as Acute Myocardial Infarction is one of the 3 leading causes of death in the world according to WHO data, in the same way cardiac arrhythmias are very common diseases today, such as atrial fibrillation. The ECG electrocardiogram is the means of cardiac diagnosis that is used in a standardized way throughout the world. Machine learning models are very helpful in classification and prediction problems. Applied to the field of health, ANN, and CNN artificial and neural networks, added to tree-based models such as XGBoost, are of vital help in the prevention and control of heart disease. The present study aims to compare and evaluate learning based on ANN, CNN and XGBoost algorithms by using the Physionet MIT-BIH and PTB ECG databases, which provide ECGs classified with Arrhythmias and Acute Myocardial Infarctions respectively. The learning times and the percentage of Accuracy of the 3 algorithms in the 2 databases are compared separately, and finally the data are crossed to compare the validity and safety of the learning prediction.

Fetal Electrocardiogram Extraction Using Dual-Path Source Separation of Single-Channel Non-Invasive Abdominal Recordings.

The development of a method for non-invasive monitoring of fetal electrocardiogram (FECG) signals from single-channel abdominal recordings. The dual-path source separation (DPSS) architecture is introduced for the simultaneous separation of fetal and maternal ECG signals from abdominal ECG recordings. DPSS initially denoises abdominal ECG (AECG) recordings using a generative dual-path long short-term memory (DP-LSTM) network. An inception module along with a series of DP-LSTM blocks is then employed to extract the masking maps associated with fetal and maternal components. Finally, these masking maps are weighted by the AECG recording to separate maternal and fetal ECG signals. The performance of this network is evaluated on 10 pregnancies from the fetal ECG synthetic database (FECGSYNDB), 22 cases of labor and pregnancy from the abdominal and direct fetal ECG database (ADFECGDB), and 69 pregnancies from set A of non-invasive FECG challenge (NIFECGC) datasets. F1-scores of 99.03%, 97.7%, and 95.3% are reported for the detection of fetal QRS complexes in FECGSYNDB, ADFECGDB, and NIFECGC respectively. DPSS technique is also evaluated in terms of separability of fetal and maternal clusters. According to the clustering-based analyses, the average purity index of 0.9750, Jaccard index of 0.9705, and Davies-Bouldin index of 0.7429 demonstrate the high source separation capability of DPSS. The achieved performance suggests that DPSS enables accurate single-channel FECG extraction, and can replace state-of-the-art source separation techniques for this purpose. This study signifies a fundamental step towards non-invasive fetal ECG monitoring systems, which favors at-home prenatal care.

Variability of expert assessments of ECG time domain parameters

BACKGROUND In anaesthesiology, time domain parameters of the ECG are frequently used for long QT diagnosis, determination of branch blocks and for identifying atrioventricular blocks. However, this analysis depends on several factors and therefore the time domain parameters are prone to variability. OBJECTIVE To determine the variability of expert assessments of ECG time domain parameters in daily clinical practice. DESIGN In an observational study 18 physicians labelled the onset, peak and end of the waves (P-QRS-T) for 26 preselected, distinctive beats. Based on this, clinically important durations and intervals were derived: the duration of the P wave, T wave and QRS complex, as well as the length of the PQ and QT interval. These parameters were subsequently analysed with respect to inter-expert variability (for all experts and discipline-based subgroups) and, additionally, were compared with computer-aided analysis. SETTING ECG recordings from participating patients were recorded in March 2015 during standard examination in the hospital and stored in the local ECG database. PATIENTS We did not define inclusion or exclusion criteria for the patients themselves; the ECG beats were selected by a medical expert with respect to their shapes and abnormalities. MAIN OUTCOME MEASURES Variability of clinically relevant ECG wave durations and intervals quantified were expressed as the interquartile range and the Q2/Q98 span for 18 investigators. RESULTS The resulting wave durations (Pdur, QRSdur and Tdur) and intervals (PQ and QT) showed high variability, for instance, captured by the Q2/Q98 span ranging from 39 to 99 ms. CONCLUSION The observed, inter-investigator variability of assessing the PQ and QT intervals, as well as the wave durations, might result in important variance regarding ECG-associated diagnoses. Whether these variances play a major role in typical clinical situations would have to be demonstrated with further clinical observational studies.

Multi-lead ECG heartbeat classification of heart disease based on HOG local feature descriptor

IntroductionECG data play an important role in the diagnostics of various cardiovascular diseases. Classification of multi-lead ECG signals could be challenging even for well-trained physicians. In this study we propose a new approach for multi-lead ECG classification. MethodFive-types of 15-lead ECG data namely healthy control, bundle branch block, cardiomyopathy, Dysrhythmia, and myocardial infarction patients from two types of datasets, 5319 and 6647 heartbeats from Baqiyatallah and PTB Diagnostic ECG database, were used, respectively. One-dimensional total variation regularization was used to denoising ECG data. Heartbeats were extracted by one cardiologist and saved as images with jpg format. Histogram of oriented gradients method was used to extract feature of images. for classification task support vector machine and fully connected neural network were used. Five-fold cross validation was used for validating the models. ResultFor 15-lead ECG PTB Diagnostic database, the best classification models are SVM model with cubic (accuracy: 99.9%, Range: 99.77% - 100%) and quadratic (accuracy: 99.88%, Range: 99.77%-100%) kernel function, for this dataset fully connected accuracy is 99.4% with range of 99.02%- 99.70%. Regarding to the Baqyatallah dataset SVM with cubic (accuracy: 99.83%, Range:99.72%-100%) and quadratic (accuracy: 99.77%, Range: 99.62%-99.9%) were the best classification model and the accuracy for fully connected neural network was 99.1% with the range of 98.59%-99.62% based on HOG descriptors. Expected sigmodal kernel all classification method have accuracy more than 99%. Discussionsimultaneous use of HOG feature extraction method and appropriate classification algorithm such as SVM or fully connected neural network can classify 15-lead ECG heart-beat for different heart disease with high accuracy and adding other relevant patients’ information can be easily done in order to increase the method performance.

Improving ECG Classification Performance by Using an Optimized One-Dimensional Residual Network Model

Cardiovascular disease and its consequences on human health have never stopped and even show a trend of appearing in increasingly younger generations. The establishment of an excellent deep learning algorithm model to assist physicians in identifying and the early screening of ECG abnormalities can effectively improve the accuracy of diagnosis. Therefore, in this study, the deep residual network model is adapted for feature extraction and classification of ECG signals by pooling embedded into layers and double channel connection. At the same time, the wavelet adaptive threshold denoising algorithm is used to complete the high signal-to-noise filtering of ECG signals. Then, the alternate pooling residual network (APRN) is compared with the convolutional neural network (CNN), CNN with one residual unit (CNN-R), and the deep residual network (ResNet-18) using ECG datasets from the American MIT-BIH arrhythmia and ST segment abnormality database, European ST-T database, and sudden cardiac death ambulatory ECG database. The results are as follows: The average classification accuracy of the APRN on the four datasets is 97.89%, while the accuracies on CNN, CNN-R, and ResNet-18 are 97.17%, 97.53%, and 97.73%, respectively. In addition, compared with ResNet-18, the classification accuracy of our APRN on each class of data improves by 16.44% in total.