IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL.53, NO.1, JANUARY 2006 43 Abstract—We investigated how complexity-based estimators of heart rate variability can detect changes in cardiovascular autonomic drive with respect to traditional measures of variability. This was done by analyzing healthy subjects and paraplegic patients with different autonomic impairment due to low (vascular impairment only) or high (cardiac and vascular impairment) spinal cord injury, during progressive autonomic activations. While traditional techniques only quantified the effects of the autonomic activation, not distinguishing the effects of the lesion level, some recently proposed complexity estimators could also reveal the pathologic alterations in the autonomic control of heart rate. These estimators included the detrended fluctuation analysis coefficient (sensitive to both low and high autonomic lesions), sample entropy (sensitive to low-level lesions) and the largest Lyapunov exponent (sensitive to high-level lesions). Thus complexity-based methods provide information on the autonomic function from the heart rate dynamics that cannot be obtained by traditional techniques. This finding supports the combined use of both complexity-based and traditional methods to investigate the autonomic cardiovascular control from a more comprehensive perspective Index Terms—heart rate variability, entropy, self-similarity, Hurst, DFA, Lyapunov exponents, spinal cord injury I. INTRODUCTION EART rate variability (HRV) is increasingly used to assess autonomic dysfunction in different pathological conditions, either of cardiac (myocardial infarction, congestive heart failure, life threatening arrhythmias) [1-4] or non-cardiac origin (diabetes, neuropathies, obesity, etc.) [5-9]. Manuscript received October 5, 2004; revised April 26, 2005. Asterisk indicates corresponding author. G. Merati and A. Veicsteinas are with the Institute of Physical Exercise, Health and Sports (IEFSAS), University of Milan, Milan, Italy (e-mail: giampiero.merati@unimi.it). M. Di Rienzo is with the Centro di Bioingegneria of the Fondazione Don C. Gnocchi ONLUS, Milan, Italy (e-mail: mdirienzo@cbi.dongnocchi.it ). G. Parati is with the Istituto Scientifico Ospedale San Luca, Istituto Auxologico Italiano and University of Milano-Bicocca, Milan, Italy (e-mail: gianfranco.parati@unimib.it ). * P. Castiglioni is with the Centro di Bioingegneria of the Fondazione Don C. Gnocchi ONLUS, via Capecelatro 66, I 20148, Milan, Italy (e-mail: pcastiglioni@cbi.dongnocchi.it) Currently, the assessment of cardiovascular autonomic impairment is mostly based on statistical indexes of variability derived from heart rate variance or power spectrum [10], whereas complex and nonlinear components of heart rate dynamics are largely neglected to this aim, although several authors provided evidence of their clinical importance and potential ability to identify cardiac autonomic dysfunction [11- 18]. The limited attention to complexity-related components of HRV in the assessment of autonomic control of circulation is possibly due to the lack of a general consensus on the validity and reliability of these components. This scenario may be determined by a poor knowledge of the settings in which estimators of complex dynamics can be applied to physiologic data. It may also be determined by the lack of an unequivocal biological interpretation of the results obtained from the complexity-based analyses. In an effort to provide a contribution to clarify these issues, this study investigated the ability of different complexity-based estimators of HRV to detect changes in the autonomic drive of the cardiovascular system in health and disease. This was done under 3 behavioral conditions which are known to induce progressively increasing degrees of autonomic activation (supine rest, sitting at rest and mild physical exercise). Recordings have been performed in healthy subjects and in patients with various degree of autonomic impairment due to traumatic spinal cord injury. The performances of the complexity estimators in detecting autonomic changes have been also compared with the performances of some frequency- domain and time-domain parameters traditionally derived from HRV analysis. Given that each estimator offers a particular perspective on the HRV phenomena, our study was aimed at investigating whether the new complexity-based perspectives are more suitable than the traditional approaches to explore autonomic control of circulation in conditions characterized by different degrees of autonomic activation and/or different levels of autonomic dysfunction. Assessment of the Autonomic Control of Heart Rate Variability in Healthy and Spinal-Cord Injured Subjects: Contribution of Different Complexity-Based Estimators Giampiero Merati, Marco Di Rienzo, Member, IEEE, Gianfranco Parati, Arsenio Veicsteinas, and Paolo Castiglioni* H