Digital models for arterial pressure and respiratory waveforms.

Abstract

Digital models for arterial pressure pulse (APP) and respiratory volume waveforms (RVW) are proposed for efficient representation of these signals. When these signals are discrete cosine transformed (DCT), the pole-zero technique of Steiglitz-McBride (SM) gave system functions of much lower order than those obtained directly from the signals. The DCT of a bell-shaped biphasic wave needed two poles and two zeros. Based on this, the model order is fixed by the number of distinct peaks in the magnitude spectrum of the transformed APP/RVW signal. The partial fraction expansion (PFE) of the system function allowed delineation of component waves present in the time signal. The angles of model poles and zeros enabled easy determination of several important features from both of these signals. The model performance is evaluated using the normalized root mean-square error (NRMSE). A Bayes classifier using the pole angles as the feature vector performed satisfactorily when a limited number of RVW's recorded under deep and rapid maneuver are classified into normal (n) and abnormal (ab) categories of respiratory pathways.