ECG delineation using multiresolution DWT and relative magnitude and slope comparison

The paper proposes an algorithm for delineation of ECG data based on multiresolution Discrete Wavelet Transform (DWT) and relative magnitude and slope comparison on selected time windows. The data is initially denoised using linear and nonlinear wavelet based filtering techniques to eliminate both low and high frequency ECG artifacts and also reduce the wide-band noises. The delineation algorithm involves combining specific detailed sub-bands of the denoised data and applying proper thresholds and time windows to identify the wave boundaries (QRS, P and T waves). The characteristic points of each wave (peaks, onset and offset) are then detected using relative magnitude and slope comparison within each wave. The algorithm yields sufficiently high detection sensitivity of 99.85% for QRS regions and also efficiently identifies the P and T wave features, as tested with the MIT-BIH arrhythmia database.