Uso de redes complexas ordinais para análise de variabilidade da frequência cardíaca

Abstract

Introduction: Efforts to apply the analysis of complex ordinal networks are being expanded and recognized as effective in the evaluation and interpretation of non-linear biological signals, including those obtained from electrocardiograms (ECGs) and electroencephalograms (EEGs). Ordinal symbolic analysis, that is, the analysis of ordinal representations, has some practical advantages. First of all, it is generally recognized as being conceptually simple and computationally fast. Furthermore, ordinal patterns, being defined by inequalities, are relatively robust against observational noise. For this reason, ordinal symbolic analysis has remained a popular method in biology and medicine, especially when it comes to distinguishing normal and abnormal health conditions in real time. Objective: To analyze the quantification measures of complex networks to apply them in the differentiation of time series of RR intervals of a group of healthy individuals and those with coronary artery disease. Material and Methods: This is a computational study that involves analysis of time series of RR intervals of healthy young people with coronary artery disease from a database. The time series was mapped for transformation into complex networks. From the complex networks, several parameters were obtained, such as entropy calculation, number of network edges, average degree of network, network density, and global clustering coefficient of the network. In the implementation of the mathematical analyses, Python and SPSS version 20.0 were used for the statistical analyses. In the first part of the work, the impact of dimension and delay parameters on entropy measurements was analyzed. In the second part, the total sample corresponding to 40time series of RR intervals, divided into two groups, were compared by applying the network quantification measures. Results and Discussion: The parameters had a directly proportional effect on entropy measurements up to a certain limit. Findings in the literature showed that the limitation in the analysis of ordinal networks is caused by the dissipation in the mapping process or by the lack of evidence of some types of signals. It was observed that to reduce the degenerations, the dimensions need to be large enough without, however, derail the computational analysis or excessively increasing the noise in the signals. In the quantifiers of the network, it was possible to evaluate the impact of the parameters on the efficacy, for the classification of healthy and coronary patients. Conclusion: Shannon entropy stood out in all dimensions for delay 6, showing better results in dimensions 5 and 6, with special emphasis on dimension 6 in several of them. Enabling the differentiation of time series of RR intervals of a group of healthy and coronary diseased individuals.