Characterization and Classification of Heart Rate Signals Usning Linear and Nonlinear Time Series Analysis Techniques

Abstract

The living organisms possess several types of rhythms interacting with each other and the outside dynamic environment, under the control of incalculable feedback systems performing orderly function to enable life. The alternations in the rhythms of physiological system help us to obtain information about the current state of living systems having substantial diagnostic value in context of human health and disease. The human body emits rhythmic alterations in form of recordable signals called biological signals which reflects the, characteristics, state, properties and the information about the physiological parameter such as heart, brain, muscles and genes etc. The large body published literature suggested that heart rate signals are most widely explored biological signals during last four decades. The electrocardiography (ECG) is used to detect abnormalities in the cardiac rhythms during the onset of cardiovascular problems. The rhythms of heart started to change long before the onset of disease, for which long term ambulatory ECG (AECG) recording is required. Therefore, 24 h or 48 h AECG monitoring is becoming vitally important for early detection of abnormal events to prevent onset cardiovascular disease and in various clinical settings. The variations in the beat-to-beat intervals called heart rate variability (HRV) reflects the cardiac autonomic control of the autonomic nervous system (ANS), via its sympathetic and parasympathetic branches. Reduced heart rate variability has been associated with the onset of pathological disturbances, aging and early warning signs of impending disease. During the last three decades, many linear and non-linear HRV analysis techniques have been proposed for the extraction of information from cardiac inter-beat interval time series data. In the recent past, few studies have been conducted to find the relation between heart rate (HR) and HRV. These studies either did not investigate the relationship between HR and HRV quantitatively or only considered linear HRV measures to find quantitative relationship between HR and linear HRV parameters. Under usual physiologic conditions, heart is not periodic oscillator, the linear HRV measures may fail to provide account for transient fluctuations in the RR-interval data. The nonlinear HRV measures have been used in numerous studies to account for transient fluctuations in the heart. The one direction of the study was to investigate the relationship of both linear and nonlinear HRV measures with HR. The result revealed inverse relationship between HRV metrics and HR for human and animal heart rate time series data. Recently, researchers proposed the idea of multiscaling for extracting information from biological signals and validated that biological signals provide dynamically incorrect information at single time scale. The second direction of the study was to assess, how multiscaling procedures affect the relationship between HRV parameters and heart rate. The results revealed inverse correlation between HR and HRV parameter at threshold values 1 to 5. Furthermore, the study focused on improving the classification ability of sign series descriptor acceleration change index (ACI) and proposed novel sign series measures for charactering the dynamics of healthy and pathological subjects. The dynamical information encoded in the interbeat interval time series was examined using scale based ACI measures (MACI and CMACI). The proposed scale base ACI measures were compared with ACI for assessing the computational performance. The ANOVA, Bonferroni post-hoc test, AUC, sensitivity, specificity, PPV, NPV, FDR, FOR and total accuracy were used for assessing the performance of ACI and scale based ACI for classifying healthy and pathological subjects. The results reported in the study depicted that scale based acceleration change index measures showed better classification between pathological and healthy groups at wide range of temporal scales.