The Detection of P-Waves in Ambulatory ECG Recordings by the Use of Wavelet Transforms

Abstract

This research concerns the automated detection of P-waves in ambulatory ECG recordings using wavelet transforms. The P-wave is the electrical signal produced during the depolarisation of the atria. The interest in the detection of P-waves lies in their usefulness in the detection of arrhythmias of the heart. Huntleigh Healthcare have already developed a commercial automatic ambulatory ECG analysis system, which classifies heartbeats using the information obtained from the ventricular activity. Improvements to this software can be obtained by augmenting this information with that obtained from the atrial activity. Initially the Haar, Coiflet and Daubechies wavelets were compared for denoising ECG recordings. The Coiflet and Daubechies wavelets proved to be very similar in their ability to denoise signals, with both being superior to the Haar wavelet. The algorithme 4 trous was identified as a suitable algorithm for implementing the detection of P-waves using wavelet transforms due to its time invariance. The Haar and quadratic spline wavelets were compared using this algorithm and both were able to accurately detect P-waves in real ECG recordings. Work on artificial ECG records however, suggested that the quadratic spline wavelet was superior to the Haar wavelet if a large amount of noise was present in the signal.