ABSTRACT
Over the last few decades, several advancements in experimental techniques have allowed for the measurement of health parameters at a high sampling frequency [3,5]. Although a frequency analysis of the data often reveals major anomalies and helps in the diagnostic process, a great part of the information is hidden in fine properties of the measured time series. Among these properties, extreme fluctuations of health parameters may trigger irreversible processes and result in acute crisis. In the ambit of cardiovascular disease, blood pressure fluctuations may trigger acute hypotensive (hypertensive) episodes and, in some cases, cardiac crisis. For a series of independent and identically distributed (iid) variables, a traditional extreme value analysis straightforwardly gives the probability of observing extremely low (or high) fluctuations of health parameters [13]. However, blood pressure data have internal correlations originating from the quasiperiodic biological processes responsible for blood circulation. Therefore, in order to provide effective warnings against cardiac crisis, the traditional techniques must be accompanied by methods that preserve the dynamical information contained in the data.
