Advanced maternal ECG removal and noise reduction for application of fetal QRS detection

The fetal heart rate (fHR) is commonly used as an indirect indicators of fetal condition. Noninvasive fHR monitoring uses electrodes placed on the mother's abdomen. However it is challenging to detect fetal QRS (fQRS) complexes from the signals measured in this way because of different types of noise and overlapping frequencies between maternal and fetal ECG. In this paper, we introduce an augmented multi-lead principal component regression (PCR) approach for maternal ECG removal and multichannel correlation based fHR detector. Using this method we participated to Computing in Cardiology, PhysioNet Challenge - and final results were for event 4: 28.893 bpm
²
and event 5: 4.844 ms. The proposed algorithm succeeded to remove maternal ECG from the abdomen signals with high accuracy.