Model-Based Cardiac Field Artefact Correction for OP-MEG

Abstract

Each heartbeat creates an electro-magnetic field that propagates throughout the body and is superimposed on measured brain activity in EEG and MEG recordings. Being several orders of magnitude larger than a typical evoked response in the brain, the Cardiac Field Artefact (CFA) is especially problematic in studies where cortical activity time-locked to the cardiac cycle is of interest, such as in interoception research. In this work, we develop a model-based CFA correction method, which exploits the fact that the magnetic field propagates relatively evenly through the mostly diamagnetic tissue of the human body, and that in studies using wearable optically-pumped magnetometers (OP-MEG) participants are free to rotate their head. Free head rotation decouples the heart’s magnetic field from that of the brain, allowing for the creation of a cardiac dipole moment (CDM) estimate, i.e. an estimate of the electrical current flow in the heart, which is minimally confounded by concurrent neuronal activity in the brain. We show in simulations that, given information about the relative position and orientation of head and chest, this CDM estimate can be used to predict and correct the sensor-level CFA by dynamically accounting for the spatial relationship between heart, brain, and sensors. We further provide an empirical proof-of-principle demonstration of the pipeline in a single participant using recordings with intermittent head movement, tracked by optical motion capture, as well as an established auditory response paradigm with well-characterized response pattern.