Muscle artifacts in multichannel EEG: Characteristics and reduction

ObjectiveTo study the characteristics of unintentional muscle activities in clinical EEG, and to develop a high-throughput method to reduce them for better revealing drug or biological effects on EEG.MethodsTwo clinical EEG datasets are involved. Pure muscle signals are extracted from EEG using Independent Component Analysis (ICA) for studying their characteristics. A high-throughput method called ICA-SR is introduced based on a new feature named Spectral Ratio (SR).ResultsThe spectral and temporal characteristics of the muscle artifacts are illustrated using representative muscle signals. The spatial characteristics are presented at both the group- and the subject-level, and are consistent under three different electrode reference methodologies. Objectively compared with an existing method, ICA-SR is shown to reduce more artifacts, while introduce less distortion to EEG. Its effectiveness is further demonstrated in real clinical EEG with the help of a CO2-inhalation EEG recording session.ConclusionThe characteristics of unintentional muscle activities align with the reported characteristics of controlled muscle activities. Artifact spatial characteristics can be EEG equipment dependent. The ICA-SR method can effectively and efficiently process clinical EEG.SignificanceArmed with advanced signal processing algorithms, this study expands our knowledge of muscle activities in EEG from muscle-controlled experiments to general clinical trials. The ICA-SR method provides an urgently needed solution with validated performance for efficiently processing large volumes of clinical EEG.

► This study expands our knowledge of muscle activities in EEG from muscle-controlled experiments to general clinical trials by studying the temporal, spectral, and spatial signal characteristics of unintentional muscle activities. ► The proposed high-throughput muscle artifact reduction method provides an urgently needed solution with validated performance for processing large volumes of clinical EEG. ► The muscle activity topography generated by the proposed method can be used as a biofeedback tool to reveal the unintentionally stressful spots on a person’s head.