Fetal QRS complex detection based on three-way tensor decomposition

This study is focused on detection of fetal QRS complexes in multichannel ECG signals recorded from mother's abdomen, containing both fetal and maternal ECGs. Assuming different values for maternal and fetal heart rates, the proposed method relies on a deterministic tensor decomposition method, which aims at deterministic blind separation of sources having different symbol rates. In the ECG context, due to the quasi-periodic nature of ECG signal, maternal ECG R-peaks are firstly detected from the mixture to identify maternal beats as maternal ECG symbols. Then the maternal ECG beats are stacked into a three-dimensional array. Decomposition of this tensor yields three loading matrices that are now used to reconstruct the maternal ECG. The residue of subtraction of the maternal ECG estimate from the original mixture is then used to detect fetal QRS complexes. The obtained average scores of event 4 and 5 on the set B of PhysioNet Challenge 2013 data are 1514.59 and 57.01, respectively.