A fast and scalable framework for automated artifact recognition from EEG signals represented in scalp topographies of Independent Components

Electroencephalography (EEG) measures the electrical brain activity in real-time by using sensors placed on scalp. Artifacts, due to eye movements and blink, muscular/cardiac generic disturbances, have be recognized eliminated allow a correct interpretation of useful signals (UBS) EEG. Independent Component Analysis (ICA) is effective split signal into independent components (ICs) whose re-projections 2D scalp topographies (images), also called topoplots, recognize/separate artifacts UBS. Until now, IC topoplot analysis, gold standard EEG, has been carried visually human experts and, hence, not usable automatic, fast-response We present completely automatic framework for EEG artifact recognition based Convolutional Neural Networks (CNNs), capable divide topoplots 4 classes: 3 types The setup described results are presented, discussed compared with those obtained other competitive strategies. Experiments, public datasets, shown an overall accuracy above 98%, employing 1.4 sec PC classify 32 that drive system sensors. Though real-time, proposed efficient enough used EEG-based Brain-Computer Interfaces (BCI) faster than methods ICs.