Fetal development assessed by multiscale multifractal analysis of heart rate variability at the very low frequencies

Abstract

33 Fetal development assessed by multiscale multifractal analysis of heart rate variability at the very low frequencies Jan Gieraltowski , Dirk Hoyer , Uwe Schneider , Jan Żebrowski a Faculty of Physics, Warsaw University of Technology, Warszawa, Poland University Hospital, Biomagnetic Center, Hans Berger Clinic for Neurology, Friedrich Schiller University of Jena, Jena, Germany Objectives: Very-low-frequency and ultra-low-frequency components of human heart rate variability (HRV) account for 95% of signal total power, but their physiological meaning is still very uncertain. The main difficulty lies in the lack of appropriate analysis methods for such low-frequency bands better than the basic power spectral analysis. Fractal methods such as detrended fluctuation analysis or MF-detrended fluctuation analysis have been used to characterize the frequencies below the low-frequency band also regarding fetal heart rate variability [1]. Methods: We applied a new method, the Multiscale Multifractal Analysis (MMA) [2], an extension of multifractal-detrended fluctuation analysis inwhich theHurst surface is calculated. Themethod is designed to analyze correlation properties at such low frequencies. We analyzed one hundred fifty-eight 30-minute magnetocardiographic HRV recordings from fetuses 21 to 38 weeks of age. To analyze such short data sets using MMA, we introduced an overlapping sliding window of analysis and limited the range of the scales considered.We distinguish a quiet and an active state of the fetus (related to the developing behavioral states of quiet and active sleep) irrespective of age and also distinguish young (gestational age b29 weeks) and old (N34 weeks) fetuses. For every single series, we calculated theHurst surface, which represents the properties of the scaling of heart rate fluctuations, depending on their magnitude and frequency range. To show the development stages of HRV in very low frequency, we calculated the mean Hurst surfaces for every week and also a contour plot showing correlations between the Hurst surface and the gestational age. Results: A large difference between the properties of the HRV of the young fetuses and the older ones is well visible and in agreement with Gieraltowski et al [3]. We chose the most significant areas of the Hurst surface (highly correlated with age) and assessed the changes of their mean values statistically. Conclusions:When introduced the MMA method [2] required at least a 104-RR-interval-long series. We modified it and obtained a tool allowing analysis of 4000-RR-interval-long fetal data (30 minutes). (For adults, the requirement for data length is now slightly more than 1 hour.) We observed a difference between the results for small amplitude fluctuations in the signal and for large amplitude fluctuations as well as differences for different scale ranges. In addition, using the results of statistical analysis of Hurst surfaces for the whole group studied, we attempted to prepare criteria to assess the stage of fetal development and gestational age. We hope that our work will help to develop a method of assessing the health status and the exact age of fetuses. Note that MMA is designed as a tool for the analysis of frequencies below the low-frequency band of HRV. http://dx.doi.org/10.1016/j.jcrc.2012.10.049