Introduction: Intrauterine fetal death is a leading cause of fetal death global-ly. Fetal heart rate variability (FHRV) can readily indicate causes of com-promise. Recently ECG has emerged as an accurate means of measuring FHRV. Thus, we propose a computationally low-cost algorithm capable of locating R-peaks, and subsequently calculating FHRV developed based on online open-source dataset.
Methods: An opensource abdominally recorded ECG dataset from physio-net.org was acquired with 5-minute recordings from 5 patients. Fetal scalp electrocardiography acted as the gold standard. Spectral analysis of the rec-orded signal is performed, and a final 15 Hz – 45 Hz IIR bandpass filter is applied, emphasizing fetal ECG while attenuating maternal ECG. We then developed the following 5 step R-Peak detection algorithm below: Step 1: Averaging of 4 ECG electrodes into a single array of data. Step 2: Squaring of data to emphasis peaks associated fetal QRS. Step 3: Reducing dataset to samples greater than the upper tail of the mean plus three times the standard deviation. Step 4: Threshold for R-peak is mean minus standard deviation of the re-duced dataset. If the standard deviation is greater than half the mean, then half the mean is used as the threshold. Step 5: Final R-peak detection performed based on steps 1-4.
Results: All subjects enrolled for recording demonstrated legitimate perfor-mance, with precision, recall, and F1 scores ranging from 97%–100% in sub-jects 1–4, with 5 still performing modestly well. For FHRV, subjects 1-4 yielded MAPEs < 3%. Bland Altman plot of FHRV performance is provided above. Subject 5s data demonstrating significant fluctuations in voltages, potentially due to maternal body habitus, or high variations in FHRV itself.