The status of nonlinear dynamics in the analysis of heart rate variability

The status of nonlinear dynamics in the analysis of heart rate variability About the time when heart rate variability (HRV) was being increasingly discussed as a tool for risk stratification after myocardial infarc-tion (6a, 13, 19), chaos theory was being propagated in popular science publications. Striking about most of the latter works was that, although heart rate regulation was the focus of nonlinear dynamics (NLD) and was seen as a particularly lucid example of chaos in physiology (see for example Gleick (8)), there were very few scientific papers on which such claims could be based. Notable exceptions were, for example, the work of Kobayashi and Musha (14), Babloyantz and Destexhe (1), and Goldberger and West (9). In the following years increasing efforts were made to determine the chaotic nature of cardiac activity by applying analysis methods from nonlinear systems theory. For instance, on the presumption of an underlying chaotic attractor, scaling properties of the generating dynamics were estimated on the basis of dimensional analysis (4, 12, 20, 28). It was however soon realized that there are fundamental difficulties involved which include the noisy nature of biological signals , the restricted length of the data available and the problems with non-stationarity. This has led to a shift of the notation from chaos to complexity, irregularity, or random-ness and has resulted in the development of measures and analysis techniques which are deemed more appropriate and more practical in application with respect to HRV. Examples of the former include the approximate entropy (22), the renor-malized entropy (15), binary entro-pies (5) and the recently introduced 'information domain' strategies of Porta (25). These all do not assume chaotic dynamics but are based on information theoretical approaches to the high-dimensional dynamics in living organisms. Others focus on fractal-like scaling properties, 1/f spectral characteristics, self-similarity , or heart rate turbulence, which also extend the field of classical NLD analysis. Pioneering and also very important to mention are the approaches developed to detect synchronization or coordination in multivariate physiological data, particularly between heart rate, beat-to-beat blood pressure , and respiratory signals (e.g. The renascence of car-diorespiratory coordination analysis (7), for example, seems to provide a clue to physiologically important sy-nergetic phenomena which play a main part in guaranteeing both stability and flexibility in physiological control processes (11). As the development of these techniques is still in progress and they have rarely been used in clinical practice (e.g. 21), …