Time-Frequency Characterization of Tri-Axial Accelerometer Data for Fetal Movement Detection M. S. Khlif 1 , B. Boashash 1,2 , S. Layeghy 1 , T. Ben-Jabeur 2 , M. Mesbah, C. East 3 , and P. Colditz 1 1 The University of Queensland, UQ Centre for Clinical Research, Herston QLD 4029, Australia 2 Qatar University, Department of Electrical Engineering, Doha, Qatar 3 The University of Melbourne, Department of Obstetrics and Gynaecology, Melbourne VIC, Australia Abstract—Monitoring fetal wellbeing is a significant problem in modern obstetrics. Clinicians have become increasingly aware of the link between fetal activity and its well-being. Using data acquired by accelerometry sensors, we use TFDs such as the spectrogram and modified B distribution (MBD) to characterize fetal movements in the time-frequency (TF) domain. This paper reports a fetal activity detection method based on the root-mean- square (RMS) of time series and evaluates its performance against real-time ultrasound imaging, taken as the gold standard. The evaluation showed better performance with the RMS-based detector as compared to maternal perception. The evaluation also showed that the detector performance is age-dependent and that fetal movement is characterized by different TF morphology. Time-frequency distributions (TFDs) with better resolution such as MBD are investigated for TF-based techniques for the detection of fetal movements. Keywords—Accelerometer, detection, fetal movement, time- frequency analysis, quadratic TFDs, Spectrogram, modified B distribution I. INTRODUCTION A widely accepted hypothesis is that fetal conditions during pregnancy significantly affect outcomes after birth [1]. Typically maternal conditions such as preeclampsia or gestational diabetes indicate a high prevalence of fetal growth compromise and increased rates of fetal morbidity and death. Monitoring of the fetus during pregnancy is an important and challenging problem in modern obstetrics. Fetal monitoring techniques are used to detect pathological conditions early enough to enable health care providers to intervene and prevent irreversible damages from occurring [2]. This goal is reachable as most of unfavorable fetal outcomes are caused by events that occur prior to the onset of labor [3, 4]. The continued monitoring of fetuses can provide important data for understanding the unexplained and unexpected stillbirths that happen late in pregnancy. A. Fetal Movement Movement is an important behavior of the fetus that can be monitored. Fetal movement is possibly a result of early neural activity as it is generated spontaneously by the central nervous system [5]. Fetal movement can then be used to monitor the immediate wellbeing of the fetus and to evaluate insight into its neurodevelopment status. In fact, fetal movement is capable of identifying antenatal factors that account for over 60% of neurodevelopment problems recognized in childhood as reported in [6]. In particular, decreases in fetal movement have been linked to fetal distress and placental dysfunction [7]. Abnormal fetal movement has also been linked to fetuses with chromosome abnormalities, anencephaly, prolonged oligohydramnios and cerebral malformations [8-10]. There are two current methods for measuring fetal movement: passive and active. Passive methods include accelerometry, phonography and tocodynamometry; they measure the fetal vibration incident on the maternal abdomen [11-14]. Active methods include the ultrasound which use echoes from high frequency sound waves directed at the fetus to produce signals, displayed as real-time images. Ultrasound techniques are accurate but expensive, require a skilled operator, and have a number of objections to their routine use [5]. Passive fetal monitoring techniques, such as accelerometry, lack the imaging capability of ultrasound but are safe, inexpensive, and simple to implement. Recent advances in solid state technology have allowed the production of new accelerometers that are small, low powered, sensitive, and robust enough to be ideal for longterm monitoring. Fetal movement can also be monitored using mother perception, but this method has been shown to be unreliable [15]. Fig. 1. Spectrogram (hamming - N/8) of a multicomponent fetal movement Automated fetal movement detection can be presented as a signal processing problem. This signal is nonstationary, 978-1-4673-0753-6/11/$26.00 ©2011 IEEE 978-1-4673-0753-6/11/$26.00 ©2011 IEEE 978-1-4673-0753-6/11/$26.00 ©2011 IEEE 466