ANN based adaptive detection of ECG features from respiratory, pleythsmographic and ABP signals

Abstract

This paper presents the prediction of ECG features using artificial neural networks from respiratory, plethysmographic and arterial blood pressure(ABP) signals. One cardiac cycle of ECG signal consists of P-QRS-T wave. This process of feature prediction determines the amplitudes and intervals in the ECG signal for subsequent analysis. The amplitude and interval values of ECG signal determine the functioning of heart for every human. This process is based on artificial neural network (ANN) and other signal analysis technique. In this process a feed forward multilayer perceptron network has been designed using back propagation algorithm. ECG signal is predicted from this network from the application of the respiratory, plethysmographic and ABP data to its input layer. For analyzing the data, a five point differentiation is done on the signal, so as to note the slope change of the resulting graph. Points with zero slopes were considered as the end of respective waves. The algorithm is tested with physionet database. The training and simulation results of the network have been obtained from Matlab7® software.