M-ECG: extracting heart signals with a novel computational analysis of magnetoencephalography data

Magnetoencephalography (MEG) captures neural activity with high temporal and spatial resolution, but it typically discards other biopotentials, such as cardiac signals, as noise. Here, we demonstrate the feasibility of extracting cardiac signals from MEG recordings using a novel algorithm to compute heart rate variability (HRV), a key autonomic biomarker. Using the Brainstorm MEG auditory dataset and the Open MEG Archive resting-state sample dataset, we developed an approach that isolates MEG-derived electrocardiogram (M-ECG) using either independent component analysis or MEG reference sensors. This algorithm identifies physiologically valid R-peaks, removes outliers, and corrects aberrant RR intervals to enable accurate HRV computation. We evaluated HRV derived from M-ECG against HRV derived from simultaneously recorded electrocardiogram (ECG) using time-domain and frequency-domain measures, along with non-parametric statistical tests and similarity metrics. Results revealed strong temporal and spectral agreement between M-ECG and simultaneously recorded ECG signals, including alignment across HRV bands and minimal bias in RR intervals. These findings highlight the potential of M-ECG for non-invasively assessing autonomic function using existing MEG data. Incorporating HRV into MEG studies could advance our understanding of brain-heart interactions and provide new diagnostic and prognostic insights, particularly in neurological disorders involving autonomic dysregulation.