Independent component analysis and fetal magnetocardiography: a tool for the automatic classification of independent components

Fetal Magnetocardiography (fMCG) allows the non-invasive recording of the weak magnetic field variations associated with the electrical activity of the fetal heart. We used Independent Component Analysis (ICA) for the separation of maternal and fetal signals from fMCG recordings. The identification of fetal components is essential to reconstruct fetal signals. In this work we present a tool for the automatic classification of independent components (ACCT). Its performances were assessed using 66 fMCG data sets of normal fetuses ranging between 22 and 37 weeks. ACCT, whose outcomes were compared with those manually obtained by an expert investigator, showed to be an effective tool. Moreover, ACCT implementation permitted the reconstruction of stable and reliable fetal traces in a completely automatic manner. The SNR of the obtained fetal signals was high, showing that this was a further step forward the use of fMCG in hospital settings. Keywords— Automatic detection, fetal magnetocardiography, independent component analysis, prenatal diagnosis.