Applications of Detrended Fluctuation Analysis for Heart Rate Variability Research

• Main Authors: Rong-Guan Yeh, 葉榮冠
• Other Authors: Jiann-Shing Shieh
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/94727526578955812021

博士 === 元智大學 === 機械工程學系 === 98 === Detrended fluctuation analysis (DFA) is a nonlinear analysis method and has been demonstrated to be useful to distinguish the heartbeat time series signal between healthy subjects and those with severe congestive heart failure (CHF) diseases. However, in the past, DFA has always required long-term data analysis (e.g., 24 hour) for its basis. Since a long-term electrocardiograph (ECG) signal recording in a patient or a late pregnant woman may be impractical and impossible. In this thesis, we consider how a short-term time series can still present useful analysis results. In addition, we investigated the advantage of global (α value) and discrete (short-term, ≤11 beats, α1 and long-term, >11 beats, α2) scaling exponents calculated in DFA of different human groups. Firstly, we calculated time series from 10 minutes to 24 hours to investigate the effect of DFA method. Hence, the measurement groups included 37 healthy (17 young and 20 old), 43 CHF and 9 atrial fibrillation (AF) adults. The scaling exponent α value of young adults from 10 minutes to 24 hours had no statistical significant difference in the different time series. Ten minutes data may be the minimum to reliably predict healthy adults by the DFA method. Although 24 hours time series data of the old, CHF, and AF groups are not all currently available, the same inference from our statistical results of the short-term time point of the α value for the old, CHF and AF groups needed at least 20 minutes, 2 hours and 1 hour time series data in our study. Just as α value, we can predict the α1 and α2 values time series point of four groups in our study. Hence, we found the DFA α values of different groups had a minimum time series for calculation in order to obtain reliable results. Moreover, the α and α2 had lower coefficient of variation (CV) value than α1 in the physiological condition. The α and α2 are better indices than α1 from a convergence view point. Secondly, we used DFA and nonrandomness indices to investigate the output of central physiologic control system under short term (i.e., 1 hour) of heart rate variability (HRV) in surgical intensive care units (SICU). In our DFA study, ten healthy volunteers, ten computes-generates white noise (randomized surrogate) signals, and seventeen patients representing 37 cases undergoing different types of neurosurgery were groups A, B, and C respectively. From C group, 25 cases were selected from 15 patients with brain injury and 12 cases were selected from 2 patients with septicemia. The difference between the DFA and nonrandomness indices study were only in the patients group C. Twenty one patients aged 21-89 years with head injury, septicemia, mechanical ventilation, and unconsciousness were used in the nonrandomness index study. These 21 patients comprised 47 cases. There are 29 cases selected from 16 patients with brain injury, 12 cases selected from 2 patients with septicemia, and 6 cases selected from 3 patients with mechanical ventilator. In our study, it was found that the α value (nonrandomness index) of patients in the SICU was significantly lower (P < 0.05) than that of healthy volunteers and significantly higher (P < 0.05) than white noise (randomized surrogate) signals. Finally, we investigated the DFA and conventional HRV of 16 late pregnant women before and 3 months after delivery, and 16 healthy women in the control group. We found that the late pregnant women had elevated global scaling exponent, elevated short-term scaling exponent and lower HRV measures in the low and high frequency ranges than those of the healthy controls and 3 months after delivery. The deranged measures recovered 3 months after delivery. In addition, the detrended fluctuation scaling exponent did not correlate with most conventional time and frequency domain measures of HRV. In conclusion, our study suggested that the global (i.e., α value) and short-term (i.e., α1value) detrended fluctuation scaling exponents might be a new and independent measures of HRV in SICU or late pregnancy, in addition to those of the conventional time domain and frequency domain measures.