Comparing advanced interference suppression algorithms for OPM-MEG

On-scalp MEG using optically pumped magnetometers (OPMs) have become more widespread with the recent introduction of the first whole-head systems that offer comparable spatial sampling to conventional MEG systems. For these systems to become truly useful in neuroscience and clinically they need to be able to deal with the kind of interference seen in typical lab and hospital environments - both environmental and participant related (e.g., dental wires). Conventional MEG uses sophisticated interference suppression algorithms to deal with this, but these tend to become instable at lower sensor counts - which makes up the majority of OPM-MEG systems to date. Alternative methods for OPM-MEG are needed. Several alternative algorithms have been proposed but a clear comparison between them is lacking. Here, we test and compare 5 of the most promising candidates, namely homogenous field correction (HFC), adaptive multipole models (AMM), signal space projection (SSP), signal space separation (SSS) and its iterative implementation (SSSit) using simulations and data recorded with a whole-head system. Recorded data includes a participant with dental wire - an especially difficult to deal with source of interference. We find that all algorithms tested achieve good interference suppression with dual-axis data. With single-axis data, SSS exhibited stability problems and several algorithms showed large signal loss (up to ~25%). SSP, temporal SSS and temporal AMM were able to suppress the dental artifact. Finally, we provide recommendations for researchers to determine help decide how to deal with interference in their OPM-MEG studies.