Adaptive Fourier series modeling of time-varying evoked potentials: study of human somatosensory evoked response to etomidate anesthetic

Abstract

Evoked potentials (EPs) have traditionally been analyzed in time domain, with amplitude and latency of various signal components used in clinical interpretation. A new approach, called adaptive Fourier series modeling (FSM), is presented here. Dynamic changes in magnitudes of Fourier coefficients are analyzed for diagnostic purposes. In order to estimate the time-varying changes in the Fourier coefficients of noisy signals, a least mean-square filtering algorithm is applied. Results of computer simulations as well as experimental data are presented. Time varying trends are presented in a new compressed evoked spectrum format. These techniques are applied to the study of alterations in human somatosensory EPs caused by the intravenous administration of etomidate during neurosurgical procedures. Amplitude increases of the order of 200–500% occurring within a time span of about 100 sec were captured. Due to its superior convergence properties, the adaptive FSM technique estimates more rapid changes in amplitude and latency than exponentially weighted averaging or moving window averaging schemes.