Usefulness of the Scalp Electroencephalography over Bone Defect for the Decoding of Muscle Activity.

Abstract

Artificial neural connections (ANCs) included in brain-machine interfaces translate neural activity into control commands for external stimulators to restore motor function in individuals with paralysis. While invasive methods such as stereotactic electroencephalography (EEG), electrocorticography (ECoG), and intracortical microelectrodes provide high bandwidth, information-rich signals capable of controlling ANCs, noninvasive techniques like EEG are often limited by the skull's attenuation of cortical activity. In this study, we tested whether EEG recorded over a bone defect following decompressive craniectomy for brain injury retains rich neural information sufficient for effective ANCs control. In this cross-sectional study, we recorded scalp EEG signals from patients who had undergone hemicraniectomy and analyzed neural activity, including high-gamma frequencies (65-300 Hz), during a simple hand-grip task. Using these EEG signals, we predicted hand muscle activity through regression-based decoding models. For comparison, ECoG signals were recorded from patients undergoing awake tumor resection, and muscle activity was similarly decoded. The prediction accuracy of EEG over the bone defect was then compared with that obtained from ECoG. EEG over the craniectomy site captured substantial neural activity, enabling decoding of muscle activity with moderate to high accuracy (R = 0.74 ± 0.21, N = 5). In comparison, decoding accuracy from ECoG signals did not significantly differ (R = 0.55 ± 0.18, N = 4). EEG recorded over a bone defect preserves rich cortical signals comparable to invasive ECoG, providing a minimally invasive platform for developing effective ANCs-based therapy in patients with brain injury.