Analysis of non-stationary heart rate series by evolutionary periodogram

The study of heart rate variability (HRV) in the frequency domain provides information to evaluate the sympathetic and parasympathetic activities of the autonomic nervous system (ANS) as related to the cardiovascular system. Many diagnostic procedures require that patients will be submitted to tests and manoeuvres able to provoke well known cardiac answers associated to specific ANS behaviour. As result of these provocative tests a non-stationary heart rate time series is often produced. Classical spectral analysis, performed by FFT or AR modelling, can be applied under steady-state condition only. Methods capable of detecting time variations of spectral components can be adopted to overcome these shortcomings. Among these methods, in this paper the Evolutionary Periodogram is evaluated. This approach decomposes the original non-stationary signal by means of basic functions of oscillatory form; these functions are non-stationary but the notion of frequency is still dominant. The method was tested on non-stationary simulated signals and real HR series recorded during tilt test, Valsalva manoeuvres, baroreflex evocation by intravenous phenylephrine administration. The same series were submitted to the STFT for comparison of the results.