Characterizing Fetal Heart Rate From A State Space Perspective

Nonlinear features have been widely adopted in var-ious biomedical applications including fetal heart rate analysis, and they have often demonstrated superior discriminatory ability as they can reveal information hidden in time series. However, classical nonlinear features have limited discriminatory ability in fetal heart rate analysis. In this paper, we cast nonlinear features into a state space reconstruction framework and show their intrinsic connection with Takens' theorem. From this perspective, we propose a novel state space reconstruction-based feature that is able to better capture the system variability which is of great importance in fetal heart rate analysis. Our experimental results on an open access intrapartum Cardiotocography database show that the proposed feature achieves better diagnostic performance in pH-based fetal heart rate analysis compared to both classical and state-of-the-art nonlinear features.