A new method for quantifying the performance of EEG blind source separation algorithms by referencing a simultaneously recorded ECoG signal

Neural Networks : the Official Journal of the International Neural Network Society
Naoya OosugiNaotaka Fujii

Abstract

Blind source separation (BSS) algorithms extract neural signals from electroencephalography (EEG) data. However, it is difficult to quantify source separation performance because there is no criterion to dissociate neural signals and noise in EEG signals. This study develops a method for evaluating BSS performance. The idea is neural signals in EEG can be estimated by comparison with simultaneously measured electrocorticography (ECoG). Because the ECoG electrodes cover the majority of the lateral cortical surface and should capture most of the original neural sources in the EEG signals. We measured real EEG and ECoG data and developed an algorithm for evaluating BSS performance. First, EEG signals are separated into EEG components using the BSS algorithm. Second, the EEG components are ranked using the correlation coefficients of the ECoG regression and the components are grouped into subsets based on their ranks. Third, canonical correlation analysis estimates how much information is shared between the subsets of the EEG components and the ECoG signals. We used our algorithm to compare the performance of BSS algorithms (PCA, AMUSE, SOBI, JADE, fastICA) via the EEG and ECoG data of anesthetized nonhuman primates. The results (B...Continue Reading

Citations

Dec 21, 2018·Healthcare Technology Letters·Seda Senay

❮ Previous
Next ❯

Related Concepts

Related Feeds

Anxiety Disorders

Discover the latest research on anxiety disorders including agoraphobia, panic disorder, obsessive-compulsive disorder, and post-traumatic stress disorder here.