The influence of motor tasks and cut-off parameter selection on artifact subspace reconstruction in EEG recordings
- PDF / 1,054,288 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 13 Downloads / 165 Views
ORIGINAL ARTICLE
The influence of motor tasks and cut-off parameter selection on artifact subspace reconstruction in EEG recordings Phillipp Anders 1
&
Helen Müller 2 & Nina Skjæret-Maroni 1 & Beatrix Vereijken 1 & Jochen Baumeister 2
Received: 19 December 2019 / Accepted: 22 August 2020 # The Author(s) 2020
Abstract Advances in EEG filtering algorithms enable analysis of EEG recorded during motor tasks. Although methods such as artifact subspace reconstruction (ASR) can remove transient artifacts automatically, there is virtually no knowledge about how the vigor of bodily movements affects ASRs performance and optimal cut-off parameter selection process. We compared the ratios of removed and reconstructed EEG recorded during a cognitive task, single-leg stance, and fast walking using ASR with 10 cut-off parameters versus visual inspection. Furthermore, we used the repeatability and dipolarity of independent components to assess their quality and an automatic classification tool to assess the number of brain-related independent components. The cut-off parameter equivalent to the ratio of EEG removed in manual cleaning was strictest for the walking task. The quality index of independent components, calculated using RELICA, reached a maximum plateau for cut-off parameters of 10 and higher across all tasks while dipolarity was largely unaffected. The number of independent components within each task remained constant, regardless of the cut-off parameter used. Surprisingly, ASR performed better in motor tasks compared with non-movement tasks. The quality index seemed to be more sensitive to changes induced by ASR compared to dipolarity. There was no benefit of using cut-off parameters less than 10. Keywords Artifact . Data processing . Electroencephalography
1 Introduction Electroencephalogram (EEG) is one of the most used methods to record activity of the brain in both clinical and applied research (e.g. epilepsy and exergaming, e.g. Acharya et al. [1] and Anders et al. [2]). Recent developments in both hardware and software, such as active electrodes [3] and advanced filter algorithms [4] make it possible to record usable EEG, while participants perform tasks involving physical movements or even in real-world environments. This offers neuroscientists a plethora of novel research designs, such as the concurrent measurement of brain activity during the execution of motor tasks, instead of having to rely on pre-post EEG
* Phillipp Anders [email protected] 1
Department of Neuromedicine and Movement Science, Norwegian University of Science and Technology (NTNU), Trondheim, Norway
2
Department Exercise & Health, Paderborn University, Paderborn, Germany
comparisons or minimal participant behavior, thereby enabling the development of more natural behavior models [5]. However, the analysis of brain activity measured while participants perform motor tasks remains challenging due to the lower signal-to-noise ratio compared with, e.g., resting state analyses. EEG recordings are typically contaminated with non-br
Data Loading...
