Performance Evaluation of Interference Removal Methods Based on Subspace Projection With Wearable OPM-MEG

Magnetoencephalography (MEG) is an important development in the field of noninvasive brain imaging. Nevertheless, MEG recordings are prone to contamination from background noise and other artifacts. The advantage of wearability and mobility of the optically pumped magnetometer (OPM) system has resulted in higher requirements for reducing interference. However, under the new OPM-MEG measurement system with limited channels, the effects of different interference suppression methods on background noise suppression have not been fully studied. We investigate six background noise suppression methods, including signal space projection, common temporal subspace projection (CTSP), signal space separation (SSS), spatiotemporal SSS (tSSS), dual signal subspace projection (DSSP), and extended SSS (eSSS). The relationships and differences between these methods are summarized. In addition, the sensitivity to its parameter and its robustness to noise of different intensities for each algorithm are verified via adequate simulation experiments. Finally, the performance of these methods is analyzed and evaluated by the semi-physical simulation experiments and the somatosensory evoked experiments of OPM-MEG for four subjects. We proposed, meanwhile, a new evaluation method for real OPM-MEG experiments. The results show that the improved CTSP method has the strongest noise reduction robustness based on the comprehensive evaluation of waveform reconstruction degree, signal-to-noise ratio (SNR), root mean square error (RMSE), power spectral density reconstruction degree, source localization error, and expected source location deviation (ESLD). This study provides a reference for the anti-jamming and evaluation of OPM-MEG.