Automatic Classification of Heartbeats using Neural Network Classifier based on a Bayesian Framework

Abstract

This paper presents a method of automatic processing the electrocardiogram (ECG) signal for the classification of heart beats. Data were obtained from 48 records of the MIT-BIH arrhythmia database (only one lead). Five types of arrhythmic beats were classified using our method, Premature Ventricular Conduction beat (PVC), Atrial Premature Conduction beat (APC), Right Bundle Branch Block beat (RBBB), Left Bundle Branch Block beat (LBBB), and Paced Rhythm Beat (PRB), in addition to the Normal Beat (NB). A learning dataset for the neural network was obtained from a five records set (124, 214, 111, 100, and 107) which were manually classified using MIT-BIH Arrhythmia Database Directory and documentation, taking advantage of the professional experience of a cardiologist. Feature set was based on ECG morphology and time intervals. Our system resulted in a minimal sensitivity of 86% and minimal specificity of 90%

Classifying heart sounds using multiresolution time series motifs

“…On the other hand, Strunic [19] extracted signals on certain band to reduce anomalies and then set a amplitude threshold to pick out the spikes and realize the segmentation [8]. To achieve classification, Karraz extracted the QRS complex from the signal as features and used them in a Neural Network Classifier based on a Bayesian framework [14]. Strunic integrated all the segmented heart cycles into one average heart cycle and used it to train the Artificial Neural Network (ANN) to classify heartbeat into categories.…”

Electrocardiogram Classification Using Reservoir Computing With Logistic Regression

“…This step consists of distinguishing the different types of beats to be classified. Many authors have already studied the heartbeat classification problem using several different techniques, such as self-organizing networks (SON) [7], selforganizing maps with learning vector quantization (SOM-LVQ) [8], linear discriminants (LD) [9], [10], signal modeling (SM) [10], support vector machine (SVM) [11], [12], discrete wavelet transformation (DWT) [13], Bayesian artificial neural networks (BANN) [14], local fractal dimension [15] and delay differential equations (DDE) [16], obtaining different performance measures. Comparing results is difficult though, because of the different measures that were used, as well as the different partitions of the available data into training, testing and validation subsets.…”

Classifying Heart Sounds - Approaches to the PASCAL Challenge

“…Strunic extracted signals on a certain band to reduce anomalies and then set an amplitude threshold to pick out the spikes and perform the segmentation (Strunic et al, 2007). To achieve classification, Karraz extracted the QRS complex from the signal as features and used them in a Neural Network Classifier based on a Bayesian framework (Karraz and Magenes, 2006). Strunic integrated all the segmented heart cycles into one average heart cycle and used it to train the Artificial Neural Network (ANN) to classify heartbeat into categories.…”