CONDITIONING, DELINEATION & STATISTICAL ANALYSIS OF MULTI-CHANNEL ELECTROCARDIOGRAM

Abstract

Multi-channel ECG Signal is recorded at. the Institute of Biomedical Engineering, University of Karlsruhe on healthy volunteers. The aim of this project is to find out the pattern of variation of the QRS complex with varied respiration and heart rates. The recorded ECG is first conditioned in order to remove noise and artifacts. A new method for baseline wander correction based on Discrete Wavelet Transform has been proposed. An algorithm using `Haar' wavelets has been devised to automate the Multi-channel ECG delineation after filtering. The same was tested on MIT-Arrhythmia Database. The resulting sensitivity and positive predictivity was found to be highly satisfactory. Principal component analysis is carried out on the extracted QRS waves to discover the `basic patterns' and the `most important deviations from mean'. Finally, a covariance matrix is calculated both in the temporal and spatio-temporal domain to represent the dependence of QRS morphology on heart rate and respiration in different individuals.