Academic Editor: Rafael Franco
Background: Magnetoencephalography (MEG) based on optically pumped magnetometers (OPMs) opens up new opportunities for brain research. However, OPM recordings are associated with artifacts. We describe a new artifact reduction method, frequency specific signal space classification (FSSSC), to improve the signal-to-noise ratio of OPM recordings. Methods: FSSSC was based on time-frequency analysis and signal space classification (SSC). SSC was accomplished by computing the orthogonality of the brain signal and artifact. A dipole phantom was used to determine if the method could remove artifacts and improve accuracy of source localization. Auditory evoked magnetic fields (AEFs) from human subjects were used to assess the usefulness of artifact reduction in the study of brain function because bilateral AEFs have proven a good benchmark for testing new methods. OPM data from empty room recordings were used to estimate magnetic artifacts. The effectiveness of FSSSC was assessed in waveforms, spectrograms, and covariance domains. Results: MEG recordings from phantom tests show that FSSSC can remove artifacts, normalize waveforms, and significantly improve source localization accuracy. MEG signals from human subjects show that FSSC can reveal auditory evoked magnetic responses overshadowed and distorted by artifacts. The present study demonstrates FSSSC is effective at removing artifacts in OPM recordings. This can facilitate the analyses of waveforms, spectrograms, and covariance. The accuracy of source localization of OPM recordings can be significantly improved by FSSSC. Conclusions: Brain responses distorted by artifacts can be restored. The results of the present study strongly support that artifact reduction is very important in order for OPMs to become a viable alternative to conventional MEG.