Detection of premature ventricular contraction using wavelet-based statistical ECG monitoring

Abstract

Automatic detection of premature ventricular contractions (PVCs) is essential to timely diagnosis of dangerous heart conditions. However, accurate detection of PVCs is challenged by multiform PVCs. In this thesis, an ECG monitoring procedure based on wavelet-based statistical process control (SPC) is proposed for diagnosing PVC beats. After ECG signals are decomposed and denoised via discrete wavelet transformation, significant wavelet coefficients are extracted through sparse discriminant analysis for constructing a monitoring statistics following Hotelling $T^{2}$ distribution. The proposed monitoring method alarms when the statistics exceeds the predefined threshold values. We select 22 recordings from the Massachusetts Institute of Technology - Beth Israel Hospital (MIT-BIH) arrhythmia database for evaluating the proposed monitoring procedure, and demonstrate the effectiveness of the proposed method.