Head and eye movements influence the decoding of different reaching directions from EEG.

Electroencephalography (EEG)-based brain-machine interfaces (BMI) have been proven effective for motor rehabilitation of severely paralyzed patients. The brain activity is classified and translated into a go vs no-go feedback (i.e., mobilizing, or not, the paralyzed limb). Patients performing the same movements but unrelated to their brain activity showed poorer or no recovery, which suggests that an accurate feedback expedites motor recovery. Being able to decode different movements from the EEG would allow providing a more accurate feedback, maximizing the rehabilitative potential. However, a dynamic rehabilitative environment with different types of movements would likely be accompanied by involuntary motions with the eyes and the head, which can contaminate the measured EEG signals. In this study we analyze how external movements associated with the task (i.e., eye or head movements) influence the performance of an EEG-based decoder of reaching movements. Our results reveal that different reaching directions could only be decoded when eye and head movements occur and only using low frequency features (delta band). In summary, this paper highlights the importance of carefully designing protocols to avoid eye and head movements to contaminate EEG signals.