ARTIFICIAL NEURAL NETWORK. BASED QRS DETECTION IN ECG SIGNAL

Abstract

ECG signal is a graphic record of electrical potentials produced in association with heart beat. Analysis of ECG is very important from medical point of view. Analysis of ECG can be manual or computer based . Manual measurements are imprecise suffer from inter & intra observer variability Computer based ECG analysis was made as early as 1957. Main problem before analysing ECG is the detection of QRS complex. QRS complex detection is difficult, not only because of the physiological variability of QRS complexes, but also because of various types of noise that can be present in the ECG signal because of EM Signal, artifacts, due to electrode motion, baseline wander, T waves, power line interference & other reasons. In the present work " Artificial Neural Network" based filter has been designed for detecting QRS complex, which is based upon the fact that each sample in ECG signal can be predicted from its previous samples & if filter is designed such that it does not predict high frequency component i.e. QRS complex, it can be made to filter out all the components ( except for QRS complex). Back propagation algorithm is used for training ANN based filter- ECG signal used for this work is signal recorded at sampling frequency of 500 Hz, at "Military Hospital"' Roorkee,Software is written in C. Initially software was tested for predicting sine wave. Study was made by taking different numbers of I/P units, nonlinearity factors, step sizes for hidden layer & 0/P layer & finally'network was trained with no. of I/P units as 5, nonlinearity factor as 0.2, hidden layer & 1 OR layer step size as 0.2 and 0.1 respectively, momentum factor for hidden & 0/P layer as 0.2.