BCIwiki (bciwiki.com) — A study published in Advanced Science on September 4, 2026, shows that while conventional re-referencing schemes reduce electrocardiogram (ECG) artifacts in signals from the Stentrode endovascular brain-computer interface, they also diminish beta-band activity related to movement; band-limited independent component analysis (BL-ICA) removes ECG artifacts while preserving motor features.
Led by Feldman AK at the Neuroscience Institute, Carnegie Mellon University, the study involved collaborators from the University of Pittsburgh, the Royal Melbourne Hospital at the University of Melbourne, and Synchron, Inc. The researchers compared typical referencing schemes, including monopolar stent-mounted, common average, and Laplacian references, finding that they significantly reduced ECG artifacts but also significantly diminished resting-state beta activity.
They proposed BL-ICA as a spatial filter that weights noise on each electrode differently, effectively separating ECG artifacts from vascular electrocorticography (vECoG) recordings. After reconstructing signals without the ECG component, they evaluated reduced cross-channel correlation and increased relative entropy between rest and go distributions, and assessed beta burst features in both source and reconstructed signal spaces to confirm preservation of low-frequency motor features.