An improved detection algorithm in fetal electrocardiography

Fetal electrocardiogram signals from abdominal recordings were digitized and processed by a personal computer. An averaged maternal signal was derived from the cross-correlation function and nonrecursive digital filtering. The cross-correlation function was calculated from the cross-spectrum and the fast Fourier transform algorithm. The maximum value of the cross-correlation function and the time location of that value were found by (1) searching for the similarity between the waveforms (for elimination of artifacts) and (2) measuring the relative time delay (for the waveform's alignment in the averaging process) between the ECG waveforms. Prior to this procedure each, of the ECG waveforms was filtered through a nonrecursive digital high-pass filter. With this procedure a template signal corresponding to one complete maternal ECG signal is obtained. The fetal ECG and the electromyogram (EMG) are suppressed in the maternal template waveform, since they are not correlated with the maternal ECG. The averaged maternal ECG was then subtracted from the abdominal signals. The subtraction results in, complete elimination of the maternal signal, thus it is possible to detect all of the fetal QRS complexes in spite of their coincidence with the maternal ECG. An averaged fetal ECG is then extracted using the same averaging process to improve the signal-to-noise ratio, making it possible to recognize fetal P and T waves. The performance of this method was assessed using simulated and recorded (real) abdominal ECG signals. This method shows a significant improvement with respect to detection of fetal heart rate, evaluation of short-term variation in the heart rate and detection of arrhythmia disturbances in the fetal ECG. The averaging procedure to obtain the averaged fetal waveform can be used to study long-term alterations in the fetal ECG.