Search

In this paper, we propose an end-to-end approach to addressing QRS complex detection and measurement of Electrocardiograph (ECG) paper using convolutional neural networks (CNNs). Unlike conventional detection solutions that convert images to digital data, our method can directly detect QRS complex in images using Faster-RCNN, then the R-peak can be located and measured through a CNN. Validated by clinical ECG data in the St.-Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database and real ECG paper from Peking University People's Hospital, the proposed method can achieve the recall of 98.32%, the precision of 99.01% in detecting and 0.012 mv of mean absolute error in measuring. Experimental results demonstrate the superior performance of our method over conventional solutions, which would pave the way to detect and measure ECG paper using CNNs.

Constructing statistical models of electrocardiogram (ECG) signals, whose parameters can be used for automated disease classification, is of great importance in precluding manual annotation and providing prompt diagnosis of cardiac diseases. ECG signals consist of several segments with different morphologies (namely the P wave, QRS complex and the T wave) in a single heart beat, which can vary across individuals and diseases. Also, existing statistical ECG models exhibit a reliance upon obtaining a priori information from the ECG data by using preprocessing algorithms to initialize the filter parameters, or to define the user-specified model parameters. In this paper, we propose an ECG modeling technique using the sequential Markov chain Monte Carlo (SMCMC) filter that can perform simultaneous model selection, by adaptively choosing from different representations depending upon the nature of the data. Our results demonstrate the ability of the algorithm to track various types of ECG morphologies, including intermittently occurring ECG beats. In addition, we use the estimated model parameters as the feature set to classify between ECG signals with normal sinus rhythm and four different types of arrhythmia.

Arrhythmia (i.e., irregular cardiac beat) classification in electrocardiogram (ECG) signals is an important issue for heart disease diagnosis due to the non-invasive nature of the ECG exam. In this paper, we analyze and criticize the results of some arrhythmia classification methods presented in the literature in terms of how the samples are chosen for training/testing the classifier and the impact this choice has on their performance (i.e., accuracy/sensitivity/specificity). From our implementation, we also report new accuracies for these methods, establishing a new state-of-the-art method, in terms of results.

The Amplitude of Dominant T-Wave Alternans (ADTWA) is a recently introduced index which quantifies the presence of microvolt T Wave Alternans (TWA) on surface ECG recordings. In this paper we investigate the reliability of ADTWA and its robustness against broadband noise. At this regard, we generated synthetic 12-leads ECG recordings through a forward electrophysiological model and we added TWA, at different extent, by modulating the variation of the repolarization times of transmembrane action potentials across even and odd beats. Using a stochastic model, we derived an analytical relationship between the repolarization variation injected into the model and TWA at the surface, thus offering a strategy to evaluate lead sensitivity. In terms of robustness, the results of the simulations show that ADTWA correctly measured the amplitude of synthetic TWA with an average error of 3.3% ± 5.8% in absence of noise. When a 100 μV peak-to-peak broadband noise is present, its effects on estimation errors were kept limited by singular value decomposition on which ADTWA builds.

The problem of detecting T-wave alternans (TWA) in ECG signals has received considerable attention in the biomedical community. This paper introduces a Bayesian model for the T waves contained in ECG signals. A block Gibbs sampler was recently studied to estimate the parameters of this Bayesian model (including wave locations, amplitudes and shapes). This paper shows that the samples generated by this Gibbs sampler can be used efficiently for TWA detection via different statistical tests constructed from odd and even T-wave amplitude samples. The proposed algorithm is evaluated on real ECG signals subjected to synthetic TWA and compared with two classical algorithms.

In this paper we propose a novel channel identification scheme for solving permutation indeterminacy introduced by Independent Component Analysis (ICA) for artifacts removal from recorded three channel ECG signals within the remote health monitoring environment. The proposed scheme does not depend on the definition of any specific artifact which is the case with the existing approach and therefore leads to more robust and generic solution to this problem. The proposed scheme has been validated using nine practical case studies and its robustness has been proved by comparison with the existing approach. Simulation results show that the proposed scheme works successfully for all the nine cases whereas the existing approach fails to identify the correct channel for four cases.

Analogue domain implementations of the Continuous Wavelet Transform (CWT) have proved popular in recent years as they can be implemented at very low power consumption levels. This is essential for use in wearable, long term physiological monitoring systems. Present analogue CWT implementations rely on taking mathematical a approximation of the wanted mother wavelet function to give a filter transfer function that is suitable for circuit implementation. This paper investigates the use of standard filter approximations (Butterworth, Chebyshev, Bessel) as an alternative wavelet approximation technique. This extends the number of approximation techniques available for generating analogue CWT filters. An example ECG analysis shows that signal information can be successfully extracted using these CWT approximations.

This paper describes the performance of beat detection and heart rate variability (HRV) feature extraction on electrocardiogram signals which have been compressed and reconstructed with a lossy compression algorithm. The set partitioning in hierarchical trees (SPIHT) compression algorithm was used with sixteen compression ratios (CR) between 2 and 50 over the records of the MIT/BIH arrhythmia database. Sensitivities and specificities between 99% and 85% were computed for each CR utilised. The extracted HRV features were between 99% and 82% similar to the features extracted from the annotated records. A notable accuracy drop over all features extracted was noted beyond a CR of 30, with falls of 10% accuracy beyond this compression ratio.

Traditional approach to the automated ECG analysis assumes the calculation is triggered for each data point acquired in a uniform time interval. This approach reveals the variability of selected diagnostic parameters differing by more than three orders. Distributed systems using global wireless digital communication may benefit from this difference using variable reporting interval or data-adaptive report content. More natural consequence is a request-driven interpretation described in this paper. Our approach assumes the processing of the acquired ECG is triggered by the data validity period expiry. Such solution significantly reduces unnecessary computation and is particularly interesting in aspect of wearable devices autonomy. Correct definition of data validity for each particular diagnostic parameter guarantees the completeness of the patient status and its convergence to the result of traditional approaches.

Pharmaceutic studies require to analyze thousands of ECGs in order to evaluate the side effects of a new drug. In this paper we present a new support system based on the use of probabilistic models for automatic ECG segmentation. We used a bayesian HMM clustering algorithm to partition the training base, and we improved the method by using a multi-channel segmentation. We present a statistical analysis of the results where we compare different automatic methods to the segmentation of the cardiologist as a gold standard.

This paper is concerned to the cardiac arrhythmia classification by using Hidden Markov Models and Maximum Mutual Information Estimation (MMIE) theory. The types of beat being selected are normal (N), premature ventricular contraction (V), and the most common class of supra-ventricular arrhythmia (S), named atrial fibrillation (AF). The approach followed in this paper is based on the supposition that atrial fibrillation and normal beats are morphologically similar except that the former does not exhibit the P wave. In fact there are more differences as the irregularity of the RR interval, but ventricular conduction in AF is normal in morphology. Regarding to the Hidden Markov Models (HMM) modelling this can mean that these two classes can be modelled by HMM's of similar topology and sharing some parameters excepting the part of the HMM structure that models the P wave. This paper shows, under that underlying assumption, how this information can be compacted in only one HMM, increasing the classification accuracy by using MMIE training, and saving computational resources at run-time decoding. The algorithm performance was tested by using the MIT-BIH database. Better performance was obtained comparatively to the case where Maximum Likelihood Estimation training is used alone.

In this paper, we developed the novel algorithm for cardiac arrhythmia classification. Until now, back propagation neural network (BPNN) was frequently used for these tasks. However, general gradient based learning method is far slower than what is required for their application. The proposed algorithm adapts Extreme Learning Machine(ELM) that has the advantage of very fast learning speed and high accuracy. In this paper, we classify beats into normal beat, left bundle branch block beat, right bundle branch block beat, premature ventricular contraction, atrial premature beat, paced beat, and ventricular escape beat. Experimental results show that we can obtain 97.45% in average accuracy, 97.44% in average sensitivity, 98.46% in average specificity, and 2.423 seconds in learning time.

This paper presents the results of morphological heart arrhythmia detection based on parameters which are obtained from modeling of the cumulants of the electrocardiography, ECG, signals. Cumulants possess many properties that make them effective tools to describe morphological variations of non-stationary signals. Among these properties, the two most attractive founded for analysis of ECG arrhythmia detections are the ability of suppressing morphological variations of different beats of ECG signals belonging to a specific class of heart arrhythmia and reducing the effect of Gaussian noise on the classification significantly. The proposed method combines these properties in conjunction with Hermitian model to perform an efficient classification method for five different heart arrhythmias. We achieved the sensitivity of 98.59% and specificity of 99.67% which are comparable to previous works. This novel combination has made the classification method much more accurate in discriminating different morphological based heart arrhythmias as well as making a good degree of robustness to remove additive Gaussian noises from ECG signals.

This paper presents a method for blind source separation (BSS) of electrocardiographic signals disturbed by noise, using the Lyapunov stability conditions for linear time invariant systems. We used records from the MIT Arrhythmia Database and in-vivo records from cardiac voluntary patients to test our signal extraction method. We quantitatively assessed the performance of our separation procedure. The method presented here compares favorably to other BSS algorithms.

A novel QRS detection algorithm based on multi-scale mathematical morphology (3M) and multi-frame differential modulus cumulation is proposed in this paper. The algorithm introduces the multi-stage filtering based on mathematical morphology from image processing field into ECG analysis to suppress the impulsive noise, and adopts multi-frame differential modulus cumulation to remove the baseline drift and enhance the signal. An average QRS detection rate of 99.67%, a sensitivity of 99.86% and a positive prediction of 99.80% have been achieved against the MIT/BIH Arrhythmia Database. Comparison with existing methods reveals that the proposed algorithm achieves better detection accuracy.