Bayesian hierarchical model with wavelet transform coefficients of the ECG in obstructive sleep apnea screening

Abstract

The Wavelet Transform allows one to analyze the properties of a variety of signals: one being able to emphasize changes in either the time or the frequency domain once the appropriate scale is chosen. Since a signal can be expressed in terms of coefficients from wavelet functions, the behavior of this signal could be sparsely represented in these functions, expressing possible properties behind nonstationary signals. Recently, methods based on hierarchical Bayes analysis have been found to be a feasible tool in the approach of physical science and engineering applications. In order to participate in the apnea screening event at the Computers in Cardiology Challenge 2000 and estimate a model that could bring one to an adequate classification between groups the authors developed the present methodology.