Spatial Quantification of Facial Electromyography Artifacts in the Electroencephalogram

Abstract

The Electroencephalogram (EEG) has been the most preferred way of recording the brain activity due to its noninvasiveness and affordability benefits. Information estimated from EEG has been employed broadly, e.g., for diagnosis or as input signal to Brain Computer Interfaces (BCI). Nevertheless, the EEG is prone to artifacts including non-brain physiological activities, such as eye blinking and the contraction of the muscles of the scalp. Some applications such as BCI systems may occasionally be associated with frequent contractions of muscles of the head corrupting the EEG-based control signal. This requires the application of a number of filtering techniques. However, standard gold techniques for signal filtering still contain limitations, such as the incapacity of eliminating noise in all EEG channels. For this reason, besides studying and applying filtering techniques, it is necessary to understand the contamination from electromyogram (EMG) along the scalp. Several studies concluded that EMG artifact contaminates the EEG at frequencies beginning at 15 Hz on the topographic distribution of the energy that encompasses practically the entire scalp. Thus, the present work aims to quantitatively estimate EMG noise in 16 bipolar channels of EEG distributed along the scalp according to the 10–20 system. This estimation was based on an experimental protocol considering the simultaneous acquisition of EEG and EMG of five facial muscles sampled at 5 kHz. The protocol consisted in activating facial muscles while listening to 15 beep sounds. The evaluated muscles were occipitofrontalis (venter frontalis), masseter, temporalis, zygomaticus major, orbicularis oculi and orbicularis oris. The mean power of the EEG contaminated by EMG of facial muscles contractions was compared between the periods of muscle contraction and non-contraction. The results show that occipitofrontalis and masseter muscular contamination is present over the scalp with increase from 63.5 μV to 816 μV and from 118.3 μV to 5,617.9 μV, respectively.