Application of linear and nonlinear methods for processing HRV and EEG signals

Biomedical signal processing is crucial for an objective interpretation of physiological systems because it allows unveiling and quantifying the information hidden in the signals that are generated by the system under study. In order to analyze biomedical signals, a large panel of algorithms conceived in different research areas has been applied. In the last few decades, concepts and techniques from nonlinear dynamics and, in particular, from chaos theory integrated the more classical linear approach, mainly based on spectral analysis. The aim of this thesis is to assess results originated by the application of linear and nonlinear analysis methods to specific clinical studies concerning the heart rate variability (HRV) and the electroencephalographic (EEG) signal. In fact, the behavior of these signals can be ascribed to systems whose nature may be linear or not, depending on the conditions they are investigated. First, the two signals (HRV and EEG) and the processing techniques used in this thesis are presented. Chapter 1 describes the physiological meaning, the data acquisition and the pre-processing requirements of both signals. Chapter 2 introduces methods and algorithms used for the characterization of the different experimental conditions in which HRV and EEG were investigated, with particular notice to nonlinear techniques. Then (chapters 3-7), the five applications of HRV and EEG analysis studied during the PhD course are described. To be more specific, the first three applications concern HRV, while the remaining two are about EEG. With respect to HRV, the first investigation assesses age-related spectral and fractal variations of the heart rhythm in healthy subjects; the second one focuses on the usefulness of the nonlinear approach in the HRV analysis of polysomnographic recordings of patients with severe Obstructive Sleep Apnea Syndrome; the third one presents the differences