A general framework for extracting fetal magnetoencephalogram and audio-evoked responses.

The analysis of auditory evoked cortical responses in fetal magnetoencephalography (fMEG) can be used as an early marker of functional cerebral development. A major obstacle for this objective is the very low signal-to-noise ratio of the fMEG recordings in presence of other biological contaminants (mainly maternal and fetal cardiac activities). Due to the fMEG nonstationarity and noise, the purpose of the present study is to improve the detection of the fetal auditory evoked response (fAER) by proposing a multi-stage framework for removing maternal and fetal artifacts using quasi-periodicity of cardiac activities, semi-blind source separation methods and detection of fAER using an ad hoc matched filter. The validation stage is performed using synchronous averaging, energy ratio comparison, statistical analysis of signal distribution, and the geometric localization of the fetal head and heart. The validation results show that the method can be effectively used in high precision fMEG and fAER applications.