A Wavelet Based Algorithm for Detecting Ventricular Tachyarrhythmia

Wavelet transform has emerged over recent years as a powerful time-frequency analysis tool favoured for the interrogation of complex nonstationary signals. In this paper a new wavelet based algorithm for detection of Ventricular Tachyarrhythmia (VT) by analyzing ECG is presented. The proposed algorithm uses generalized regression neural networks (GRNNs) for detecting wavelet preprocessed ECG signals. A MATLAB routine using built in library functions for preprocessing removes high frequency noise. The preprocessed signal is applied to the spectral algorithm (SPEC) which works in frequency domain and analyses the energy content. If the algorithm decides that the ECG part contains VT, the result is accepted as true and no further investigation is required. On the other hand a further investigation is carried out to confirm the result or to disprove it. The terminal parts of the ECG signal are processed with a continuous wavelet transform, which leads to a time-frequency representation of the signal. The diagnostic feature vectors are obtained by subdividing the representations into several regions and by processing the sum of the decomposition coefficients belonging to each region. The neural network is then trained with these feature vectors. Based on this information GRNN classifies the ECG into presence or absence of VT. With this method, underlying features within the VT waveform are made visible in the wavelet time-scale half space. The proposed algorithm overcomes the non-sensitivity of SPEC algorithm utilizing its highly specific nature to the fullest, enabling the cardiologists and electro physiologists to detect VT with accuracy of more than 85%.