Indu Saini

QRS detection using K-Nearest Neighbor algorithm (KNN) and evaluation on standard ECG databases

The performance of computer aided ECG analysis depends on the precise and accurate delineation of QRS-complexes. This paper presents an application of K-Nearest Neighbor (KNN) algorithm as a classifier for detection of QRS-complex in ECG. The proposed algorithm is evaluated on two manually annotated standard databases such as CSE and MIT-BIH Arrhythmia database. In this work, a digital band-pass filter is used to reduce false detection caused by interference present in ECG signal and further gradient of the signal is used as a feature for QRS-detection. In addition the accuracy of KNN based classifier is largely dependent on the value of K and type of distance metric. The value of K = 3 and Euclidean distance metric has been proposed for the KNN classifier, using fivefold cross-validation. The detection rates of 99.89% and 99.81% are achieved for CSE and MIT-BIH databases respectively. The QRS detector obtained a sensitivity Se = 99.86% and specificity Sp = 99.86% for CSE database, and Se = 99.81% and Sp = 99.86% for MIT-BIH Arrhythmia database. A comparison is also made between proposed algorithm and other published work using CSE and MIT-BIH Arrhythmia databases. These results clearly establishes KNN algorithm for reliable and accurate QRS-detection.

Electrocardiogram beat classification using empirical mode decomposition and multiclass directed acyclic graph support vector machine

In this paper, a classifier motivated from statistical learning theory, i.e., support vector machine, with a new approach based on multiclass directed acyclic graph has been proposed for classification of four types of electrocardiogram signals. The motivation for selecting Directed Acyclic Graph Support Vector Machine (DAGSVM) is to have more accurate classifier with less computational cost. Empirical mode decomposition and subsequently singular value decomposition have been used for computing the feature vector matrix. Further, fivefold cross-validation and particle swarm optimization have been used for optimal selection of SVM model parameters to improve the performance of DAGSVM. A comparison has been made between proposed algorithm and other two classifiers, i.e., K-Nearest Neighbor (KNN) and Artificial Neural Network (ANN). The DAGSVM has yielded an average accuracy of 98.96% against 95.83% and 96.66% for the KNN and the ANN, respectively. The results obtained clearly confirm the superiority of the DAGSVM approach over other classifiers.

Delineation of ECG Wave Components Using K-Nearest Neighbor (KNN) Algorithm: ECG Wave Delineation Using KNN

Detection of the boundaries of electrocardiogram (ECG) characteristic waves with a reasonable accuracy has been a difficult task. As a classical statistical pattern recognition algorithm characterized with high accuracy and stability, KNN has been proposed for locating the waveform boundaries (the onsets and offsets of P, QRS, and T waves) in ECG signals. First, the QRS-complex of each beat is detected from the ECG signal. Next, the onset and offset of each QRS complex are located. The P wave and T wave, relative to each QRS complex along with their onset and offset points, are then identified using this algorithm. Further, QRS duration, heart rate, QT-interval, P-wave duration and PR-interval have also been computed using ECG wave fiducial points. This algorithm is tested on the ECG dataset acquired using ATRIA®6100 ECG machine in our own laboratory. The results obtained using the proposed algorithm presented for the assessment of performance, has been compared with the output of inbuilt software based detector of ATRIA machine.

Empirical Wavelet Transform Based ECG Signal Compression

ABSTRACT Transmission of biomedical signals over telephone lines or other communication channels is currently an important issue for the telemedicine applications. An efficient compression algorithm is needed to achieve a reduced information rate, for the storage and transmission purposes. In this paper, empirical wavelet transform (EWT) along with discrete wavelet transform (DWT) has been used for compression and reconstruction of the ECG signals. Key point lies in using different threshold for different modes obtained by applying EWT. Proposed algorithm has been tested on self-acquired (on BIOPAC®MP150) ECG signals of 20 subjects, each of 12 minutes duration and 360,000 samples with sampling rate 500 Hz and average 31.2 compression ratio (CR) and 3.28% percentage ratio distortion (PRD) have been obtained. Algorithm has also been applied on all 48 arrhythmia signals of Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH)77777\+766 database, each of 30 minutes duration and 650,000 samples, sampled at 360 Hz and an average 33.1 CR and 3.3% PRD are obtained.

K-nearest neighbour-based algorithm for P- and T-waves detection and delineation

Abstract The aim of an automated Electrocardiogram (ECG) delineation system is the reliable detection of the characteristic waveforms and determination of peaks and limits of individual QRS-complex, P- and T-waves. In this paper, a classical statistical pattern recognition algorithm characterized with high accuracy and stability, i.e., K-Nearest Neighbour (KNN) has been proposed for locating the fiducial points along with their waveform boundaries in ECG signals. First, the QRS-complex along with its onset and offset points of each beat is detected from the ECG signal. After that P- and T-wave, relative to each QRS-complex along with their onset and offset points, are then identified using this algorithm. The feature extraction is done using the gradient of the ECG signals. The performance of the proposed algorithm has been evaluated on two standard manually annotated databases, (i) CSE and (ii) QT, and also on ECG data acquired using BIOPAC®MP100 system in laboratory settings. The results in terms of accuracy, i.e., 92.8% for CSE database obtained, clearly indicate a high degree of agreement with the manual annotations made by the referees of CSE dataset-3. Further, the delineation results of the CSE and QT database are compared with the accepted tolerances as recommended by the CSE working party. The results for ECG records acquired using the BIOPAC®MP100 system, in terms of QRS duration, heart rate, QT-interval, P-wave duration and PR-interval using KNN algorithm have also been computed.

Adaptive correlation dimension method for analysing heart rate variability during the menstrual cycle

Correlation dimension (CD) is used for analysing the chaotic behaviour of the nonlinear heart rate variability (HRV) time series. In CD, the autocorrelation function is used to calculate the time delay. However, it does not provide optimum values of time delays, which leads to an inaccurate estimation of the HRV between phases of the menstrual cycle. Thus, an adaptive CD method is presented here to calculate the optimum value of the time delay based upon the information content in the HRV signal. In the proposed method, the first step is to divide the HRV signal into overlapping windows. Afterwards, the time delay is calculated for each window based on the features of the signal. This procedure of finding the optimum time delay for each window is known as adaptive autocorrelation. Then, the CD for each window is calculated using optimum time delays. Finally, adaptive CD is calculated by averaging the CD of all windows. The proposed method is applied on two data sets: (i) the standard Physionet dataset and (ii) the dataset acquired using BIOPAC®MP150. The results show that the proposed method can accurately differentiate between normal and diseased subjects. Further, the results prove that the proposed method is more accurate in detecting HRV variations during the menstrual cycles of 74 young women in lying and standing postures. Three statistical parameters are used to find the effectiveness of adaptive autocorrelation in calculating time delays. The comparative analysis validates the superiority of the proposed method over detrended fluctuation analyses and conventional CD.

P- and T-wave Delineation in ECG Signals using Support Vector Machine

Abstract Detection and delineation of QRS-complexes, P and T-waves, are important issues in the analysis and interpretation of Electrocardiogram (ECG) signals. In this paper, a classifier motivated from statistical learning theory, i.e., Support Vector Machine (SVM), has been explored for detection and delineation of these wave components. Digital filtering techniques are used to remove interference present in ECG signal. The feature extraction is done using a modified definition of slope of the ECG signals. The performance of the proposed algorithm is validated using ECG recordings from dataset-3 of the CSE multi-lead measurement library. The results in terms of accuracy, i.e., 94.4%, obtained clearly indicate a high degree of agreement with the manual annotations made by the referees of CSE dataset-3.

Classification of RR-Interval and Blood Pressure Signals Using Support Vector Machine for different Postures

In this paper, the classification of RR-interval and blood pressure series for two different physical activities postures has been performed using support vector machine (SVM). Without understanding the changes in these features from lying to standing posture in the same subject it is not possible to decipher the hidden dynamics of cardiovascular control. Thus classification of the subjects based on their RR-interval and blood pressure series, prior to spectral analysis, is essential. Therefore support vector machine, a classifier motivated from statistical learning theory, is used here for classifying the subjects based on lying and standing postures. The efficiency of SVM lies in the choice of the kernel for a given problem. Here in this paper a comparative study has been performed between Linear, Polynomial and Radial Basis kernel functions, and based on highest classification accuracy linear kernel function is proposed for SVM classifier for deciphering the postural related changes in RR-interval and blood pressure signals.

Detection of QRS-complex using K-nearest neighbour algorithm

The automatic detection of ECG wave is important for cardiac disease diagnosis. A good performance of an automatic ECG analysing system depends upon the accurate and reliable detection of the QRS complex. This paper presents an application of K-nearest neighbour (KNN) algorithm for detection of QRS-complex in ECG. Here, the ECG signal was filtered using a band-pass filter to remove power line interference and baseline wander and gradient of the signal was used as a feature for QRS detection. The accuracy of KNN algorithm is largely dependent on the value of K and type of distance metric. Hence, K = 3 and Euclidean distance metric has been proposed, using five-fold cross-validation. The performance of this algorithm was evaluated on EUROBAVAR database and ECGs recorded using BIOPAC®MP100 system and using Atria®6100 ECG machine. The detection rates of 100%, 99.97% and 100% have been achieved for respective datasets. These results emphasises that KNN is a useful tool for QRS detection.

Support vector machine-based QRS-detection – evaluation on standard databases

Detection of QRS-complex is an important issue in the analysis and interpretation of electrocardiogram (ECG) signals. In this work, a classifier motivated from statistical learning theory, i.e., support vector machine (SVM), has been explored for detection of QRS-complex. Here, a raw ECG signal is band-pass filtered to remove base line wander and power line interference. Further, gradient criterion was used to enhance the QRS-complexes. The performance of the algorithm was tested on MIT-BIH arrhythmia standard database. The numerical results indicated that the algorithm achieved 99.87% of detection rate. This algorithm performs better in comparison to other published works on the same database. Furthermore, the performance of this algorithm was also estimated on EUROBAVAR database and ECGs recorded using BIOPAC®MP100 system and using Atria®6100 ECG machine. The detection rates of 100%, 99.9% and 100% have been achieved for respective datasets. This demonstrates the superiority of SVM algorithm for QRS detection.