BACKGROUND: Monitoring of fetal heart rate (FHR) is important during labor since it is a sensitive marker to obtain significant information about fetal condition. To take immediate response during cesarean section (CS), we noninvasively derive FHR from maternal abdominal ECG. METHODS: We recruited 17 pregnant women delivered by elective cesarean section, with abdominal ECG obtained before and during the entire CS. First, a QRS-template is created by averaging all the maternal ECG heart beats. Then, Hilbert transform was applied to QRS-template to generate the other basis which is orthogonal to the QRS-template. Second, maternal QRS, P and T waves were adaptively subtracted from the composited ECG. Third, Gabor transformation was applied to obtain time-frequency spectrogram of FHR. Heart rate variability (HRV) parameters including standard deviation of normal-to-normal intervals (SDNN), 0V, 1V, 2V derived from symbolic dynamics of HRV and SD1, SD2 derived from Poincaree plot. Three emphasized stages includes: (1) before anesthesia, (2) 5 minutes after anesthesia and (3) 5 minutes before CS delivery. RESULTS: FHRs were successfully derived from all maternal abdominal ECGs. FHR increased 5 minutes after anesthesia and 5 minutes before delivery. As for HRV parameters, SDNN increased both 5 minutes after anesthesia and 5 minutes before delivery (21.30+/-9.05 vs. 13.01+/-6.89, P < 0.001 and 22.88+/-12.01 vs. 13.01+/-6.89, P < 0.05). SD1 did not change during anesthesia, while SD2 increased significantly 5 minutes after anesthesia (27.92+/-12.28 vs. 16.18+/-10.01, P < 0.001) and both SD2 and 0V percentage increased significantly 5 minutes before delivery (30.54+/-15.88 vs. 16.18+/-10.01, P < 0.05; 0.39+/-0.14 vs. 0.30+/-0.13, P < 0.05). CONCLUSIONS: We developed a novel method to automatically derive FHR from maternal abdominal ECGs and proved that it is feasible during CS.