JBHI presents

Automated Estimation of Fetal Cardiac Timing Events From Doppler Ultrasound Signal Using Hybrid Models

Featured Articles

F.Marzbanrad, Y. Kimura, K. Funamoto, R. Sugibayashi, M. Endo, T. Ito, M. Palaniswami, and A. H. Khandoker

Automated Estimation of Fetal Cardiac Timing Events From Doppler Ultrasound Signal Using Hybrid Models

Fetal cardiac monitoring is aimed to evaluate antepartum fetal risks. Although Fetal Heart Rate (FHR) monitoring is commonly used for this purpose, it is not enough for a thorough fetal assessment. There are more sensitive markers for evaluating the cardiac performance, which characterize the electromechanical coupling as a fundamental and clinically significant part of the heart physiology. The opening and closure timings of the cardiac valves are the main bases for estimating these indices. One way to obtain these timings is by using fetal echocardiography which is highly specialized and only performed in particular maternal and fetal conditions. A simpler technique is to use Doppler Ultrasound (DUS) signal and fetal electrocardiogram (fECG) as reference.  Valve motions are linked to a high frequency component of the DUS signal. Earlier studies applied band pass filters to extract this component, then the valve motions were identified manually. Considering the transient nature of the DUS signal and wide changes in its content and spectral characteristics, in this paper it is proposed to apply Empirical Mode Decomposition (EMD) to decompose the DUS signal. Furthermore a novel automated technique based on hybrid support vector machines – Hidden Markov Models (SVM/HMM) is proposed to identify cardiac valve opening and closing from the peaks of the DUS component. This automated method can continuously identify beat-to-beat valve motion timings by the rate of 91% which is higher than previous methods.

Read more at IEEE Xplore

Tags: Doppler ultrasound (DUS), empirical mode decomposition (EMD), fetal cardiac intervals, fetal monitoring, hidden Markov models (HMM), hybrid SVM/HMM, support vector machine (SVM)

Related Articles