Heart rate variability analysis through nonextensive statistics: multiscale q-entropy rate

Author(s)

Luiz Eduardo Virgilio da Silva

• Document type: Doctoral Thesis
• Press: Ribeirão Preto.
• Institution: Universidade de São Paulo (USP). Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto (PCARP/BC)
• Defense date: 2013-02-28

Examining board members

Luiz Otavio Murta Junior; Elbert Einstein Nehrer Macau; Ubiraci Pereira da Costa Neves; Valdo José Dias da Silva

• Advisor: Luiz Otavio Murta Junior

Abstract

Human body is a complex system composed of several interdependent subsystems, interacting at various scales. It is known that physiological complexity tends to decrease with disease and aging, reducing the adaptative capabilities of the individual. In the cardiovascular system, one way to evaluate its regulatory dynamics is through the analysis of heart rate variability (HRV). Classical methods of HRV analysis are based on linear models, such as spectral analysis. However, as the physiological mechanisms regulating heart rate exhibit nonlinear characteristics, analyzes using such models may be limited. In the last years, several proposals nonlinear methods have emerged. Nevertheless, no one is known to be consistent with the physiological complexity theory, where both periodic and random regimes are characterized as complexity loss. Based on physiological complexity theory, this thesis proposes new methods for nonlinear HRV series analysis. The methods are generalization of existing entropy measures, through Tsallis nonadditive statistical mechanics and surrogate data. We defined a method, called qSDiff, which calculates the difference between the entropy of a signal and its surrogate data average entropy. The entropy method used is a generalization of sample entropy (SampEn), through nonadditive paradigm. From qSDiff we extracted three attributes, which were evaluated as potential physiological complexity indexes. Multiscale entropy (MSE) was also generalized following nonadditive paradigm, and the same attributes were calculated at various scales. The methods were applied to real human and rats HRV series, as well as to a set of simulated signals, consisting of noises and maps, the latter in chaotic and periodic regimes. qSDiffmax attribute proved to be consistent for low scales while qmax and qzero attributes to larger scales, separating and ranking groups in terms of physiological complexity. There was also found a possible relationship between these q-attributes with the presence of chaos, which must be further investigated. The results also suggested the possibility that, in congestive heart failure, degradation occurs rather at small scales or short time mechanisms, while in atrial fibrillation, damage would extend to larger scales. The proposed entropy based measures are able to extract important information of HRV series, being more consistent with physiological complexity theory than SampEn (classical). Results strengthened the hypothesis that complexity is revealed at multiple scales. We believe that the proposed methods can contribute to HRV as well as to other biomedical signals analysis. (AU)

• FAPESP's process: 09/17723-0 - Heart rate variability signal analysis through nonextensive statistics: multiscale q-entropy rate and nonlinear spectral q-analysis
• Grantee: Luiz Eduardo Virgilio da Silva
• Support Opportunities: Scholarships in Brazil - Doctorate