A Method for Optimizing the Artifact Subspace Reconstruction Performance in Low-Density EEG

Electroencephalogram (EEG) plays a significant role in the analysis of cerebral activity, although the recorded electrical brain signals are always contaminated with artifacts. This represents the major issue limiting the use of the EEG in daily life applications, as the artifact removal process still remains a challenging task. Among the available methodologies, artifact subspace reconstruction (ASR) is a promising tool that can effectively remove transient or large-amplitude artifacts. However, the effectiveness of ASR and the optimal choice of its parameters have been validated only for high-density EEG acquisitions. In this regard, this study proposes an enhanced procedure for the optimal individuation of ASR parameters, in order to successfully remove artifacts in low-density EEG acquisitions (down to four channels). The proposed method starts from the analysis of real EEG data, to generate a large semisimulated dataset with similar characteristics. Through a fine-tuning procedure on this semisimulated data, the proposed method identifies the optimal parameters to be used for artifact removal on real data. The results show that the algorithm achieves an efficient removal of artifacts preserving brain signal information, also in low-density EEG signals, thus favoring the adoption of the EEG also for more portable and/or daily-life applications.