Automated Estimation of Foetal Heart Rate Variability Using Frequency-Based Floatingline Method

Özet

The assessment of foetal heart rate (FHR) variability plays a crucial role in diagnosing foetal well-being. However, the lack of a standardized definition of FHR variability (FHRV) and agreement on evaluation methodologies remains a challenge. Existing approaches often compute FHRV in segments devoid of accelerations and decelerations, making it challenging to evaluate signals with frequent and closely spaced events of this nature. In this study, we propose an automated method for estimating FHRV by defining it as the difference between the FHR signal and the floatingline. The floatingline represents an imaginary line that follows accelerations and decelerations while considering their frequency characteristics. We evaluated the developed software on both simulated and real FHR signals comprising sets of 62 signals. For simulated signals, the mean square error between the estimated and simulated floatinglines was only 0.039 bpm². In the case of real signals, without a gold standard reference, a team of five expert obstetricians visually assessed the estimated floatinglines, finding them matching the defined criteria in 95.84% of cases. Regarding FHRV evaluation, we compared the estimated values with reference values of short-term variability (STV) and sympathovagal balance (SVB) using simulated FHR signals. The error was below 1.68% for the STV index, while the SVB index was underestimated with an error of approximately 4.32%. Furthermore, we compared our proposed floatingline estimation method with traditional filters such as moving average and FIR with Hamming window. On average, our method outperformed these filters, exhibiting mean square errors up to five times lower.