Rajesh Ghongade

A brief performance evaluation of ECG feature extraction techniques for artificial neural network based classification

Electrocardiogram is the most easily accessible bioelectric signal that provides the doctors with reasonably accurate data regarding the patient heart condition. Many of the cardiac problems are visible as distortions in the electrocardiogram (ECG). Normally ECG related diagnoses are carried out by the medical practitioners manually. The major task in diagnosing the heart condition is analyzing each heart beat and co-relating the distortions found therein with various heart diseases. Since the abnormal heart beats can occur randomly it becomes very tedious and time-consuming to analyze say a 24 hour ECG signal, as it may contain hundreds of thousands of heart beats. Hence it is desired to automate the entire process of heart beat classification and preferably diagnose it accurately. In this paper the authors have focused on the various schemes for extracting the useful features of the ECG signals for use with artificial neural networks. Once feature extraction is done, ANNs can be trained to classify the patterns reasonably accurately. Arrhythmia is one such type of abnormality detectable by an ECG signal. The three classes of ECG signals are normal, fusion and premature ventricular contraction (PVC). The task of an ANN based system is to correctly identify the three classes, most importantly the PVC type, this being a fatal cardiac condition. Transform feature extraction and morphological feature extraction schemes are mostly preferred. Discrete Fourier transform, principal component analysis, and discrete wavelet transform are the three transform schemes along with three other morphological feature extraction schemes are discussed and compared in this paper.

A robust and reliable ECG pattern classification using QRS morphological features and ANN

This paper describes electrocardiogram (ECG) pattern classification using QRS morphological features and the artificial neural network. Four types of ECG patterns were chosen from the MIT-BIH database to be classified, including normal sinus rhythm, premature ventricular contraction, atrial premature beat and left bundle branch block beat. Authors propose a set of six ECG morphological features to reduce the feature vector size considerably to make the training process faster, and realize a simple but effective ECG heartbeat extraction scheme. Three types of artificial neural network models, MLP, RBF neural networks and support vector machine were separately trained and tested for ECG pattern classification and the experimental results of the different models have been compared. The MLP network exhibited the best performance and reached an overall test accuracy of 99.65%, while, RBF and SVM network reached 99.1% and 99.5% respectively. The performance of these classifiers was also evaluated in presence of additive white Gaussian noise. MLP network was found to be more robust in this respect.

An effective feature set for ECG pattern classification

In this paper, QRS morphological features and the artificial neural network method was used for Electrocardiogram (ECG) pattern classification. Four types of ECG patterns were chosen from the MIT-BIH database to be recognized, including normal sinus rhythm, premature ventricular contraction, atrial premature beat and left bundle branch block beat. Authors propose a set of six ECG morphological features to reduce the feature vector size considerably and make the training process fast in addition to a simple but effective ECG heartbeat extraction scheme. Three types of artificial neural network models, MLP, RBF neural networks and SOFM were separately trained and tested for ECG pattern recognition and the experimental results of the different models have been compared. The MLP network exhibited the best performance and reached an overall test accuracy of 99.65%, and RBF and SOFM network both reached 99.1%. The performance of these classifiers was also evaluated in presence of additive Gaussian noise. MLP network was found to be more robust in this respect.

Rough set based segmentation and classification model for ECG

Electrical activity in the heart is given by electrocardiogram (ECG) signal. Manual analysis of ECG beat is very time consuming task as it may contain hundreds of thousands of beats for 24 hours of ECG signal. This study gives a robust classification model for ECG using Rough Set Theory (RST). RST generates rules which are simple and more apprehensible for the user causing the extraction of more accurate information from the database. ECG signal is pre-processed by using different digital filters and some essential features such as R-peak, P-wave, QRS complex etc. are extracted using signal processing toolbox in MATLAB. Finally Rough set theory is used to generate reducts and then classify the ECG signal using various classification schemes. To analyze and understand the principles of Rough Set Theory, combination of different attribute selectors, search methods and different classifier is done. We have compared different classifier such as Fuzzy Rough Nearest Neighbor , Multilayer Perception (MLP), Nearest Neighbor (NN) with respect to different parameter such as correctly classified samples, kappa statistics, root mean square error, TP rate, ROC, etc.. Performance of classifier is tested on MIT-BIH arrhythmia database.

Design and simulate an antenna for aqueous glucose measurement

Diabetes is a disease that afflicts millions of people. Blood glucose measurement is a vital task to manage diabetes. The available technologies for blood glucose measurement are both painful and costly. A microwave sensor can be used for the non-invasive determination of blood glucose levels. As the step towards developing a non-invasive blood glucose measurement based on microwave, a sensor is designed and tested with aqueous glucose. A Sonnet Microwave Simulation Studio Suite v 13.52 is used for the design and simulation of sensor. Three types of sensor namely ring resonator, a microwave single spiral ring resonator and double spiral resonator are designed and simulated. The simulation result shows that, the resonant response of the sensor changes as per the change in permittivity of aqueous glucose and single spiral ring resonator exhibits excellent response as it gives maximum measurable frequency shift.

Performance Analysis of Feature Extraction Schemes for Artificial Neural Network Based ECG Classification

Many of the cardiac problems are visible as distortions in the electrocardiogram (ECG). Since the abnormal heart beats can occur randomly it becomes very tedious and time-consuming to analyze say a 24 hour ECG signal, as it may contain hundreds of thousands of heart beats. Hence it is desired to automate the entire process of heart beat classification and preferably diagnose it accurately In this paper the authors have focused on the various schemes for extracting the useful features of the ECG signals for use with artificial neural networks. Arrhythmia is one such type of abnormality detectable by an ECG signal. The three classes of ECG signals are Normal, Fusion and Premature Ventricular Contraction (PVC). The task of an ANN based system is to correctly identify the three classes, most importantly the PVC type, this being a fatal cardiac condition. Discrete Fourier Transform, Principal Component Analysis, and Discrete Wavelet Transform and Discrete Cosine Transform are the four schemes discussed and compared in this paper. For comparison the statistical techniques like linear discriminant analysis and tree clustering are also evaluated.

Arrhythmia classification using morphological features and probabilistic neural networks

Arrhythmia can be detected by carefully studying the electrocardiogram (ECG) and the distortions in the QRS complex. Since the appearance of the distorted beats, the indicators of arrhythmia, may occur randomly with respect to time and span over a large time interval, an automated classification mechanism may reduce the tedium in identifying and isolating these beats. This paper proposes an arrhythmia classifier based on probabilistic neural networks. The data is derived from MIT-BIH arrhythmia database. The classifier is designed to classify ten different types of beats, where the difference is based on morphology of the beat. Ten statistical morphological parameters are computed from the training dataset and they form the feature vector for the PNN training. The proposed classifier performs quite well with an average classification accuracy of 98.1%, average sensitivity of 0.9810, average specificity of 0.9978, average positive prediction rate as 0.981, average false prediction rate of 0.002 and average classification rate of 0.9962. The main advantage of using PNN is that it requires no training and a new class category can be added without major modifications to the network.

A frame work for analysis and optimization of multiclass ECG classifier based on Rough set theory

Detection and delineation of Electrocardiogram has played a vital role in cardiovascular monitoring systems. The enormous database of heart beats which characterize the heart disease, uncertainty, randomness in occurrence of these beats necessitate the use of Rough set theory. Over the years Rough set theory has been effectively used for removal of uncertainties and reduction of dataset. This paper discusses an optimized rough set based algorithm for detection of fiducial points for ten classes of ECG. Fiducial points help determine the peaks, valleys, onset and offset of the waves. Ten morphological features have been identified and investigation of efficiency of Rough set theory to reduce and extract the decision rules from the database has been done. The experimental results show that the proposed method has sensitivity 48%; average specificity 96% and average detection accuracy 91%. Methods involving the use of evolutionary algorithms have also been a powerful tool for dealing with complex optimization problems. Rough-fuzzy approach accompanied with Ant colony optimization, Particle swarm optimization and Genetic algorithm as search methods has also been studied. The results obtained by integrating Multilayer Perceptron or Fuzzy-Rough neural network with fuzzy rough approach for attribute selection as well has shown the highest accuracy of around 96%.

Design of bioimpedance spectrometer

Patient health is monitored by invasive as well as non-invasive methods. As invasive methods are harmful to patients body medical science requires more non-invasive methods. Bioimpedance spectroscopy (BIS) provides information regarding patient health in clinics as well as in home noninvasively. This paper describes design and implementation of the Bioimpedance Spectrometer. It uses magnitude ratio and phase difference detection method as basic technique. Developed bioimpedance spectrometer measures impedance in the 10 KHz to 1 MHz range of frequency which is β-dispersion range. Pathological changes generally occurs in this range.

Noise analysis of ECG signal using fast ICA

ECG(Electrocardiogram) signals get corrupted due to different noises and artifacts such as power line interference, baseline wander, motion artifacts, contact noise that hide lot of required information. This information is required for detecting various cardiac diseases. Hence before processing ECG signal denoising plays important role to obtain significant features of ECG signal. For denoising, Independent Component Analysis (ICA) is applied on contaminated ECG signal. ICA is a blind source separation technique used to find the independent source signal from non-gaussian noisy signal. The ICA technique deals with maximization of non-gaussian source signal using higher order parameters such as kurtosis and Negentropy. The process of maximization is based on the central limit theorem. ICA can be used for ECG denoising only because ECG signal has super Gaussian shape. This technique is capable to remove noise and artifacts even though they have same frequency as original signal. The FastICA with different granularity functions like tanh, power3 isolates noise and original signal using kurtosis values of separated signal. Analysis of FastICA shows tanh granularity has faster convergence as compared to power3 granularity and SNR ratio is of 7db.