Improving N1 classification by grouping EEG trials with phases of pre-stimulus EEG oscillations

Abstract

A reactive brain-computer interface using electroencephalography (EEG) relies on the classification of evoked ERP responses. As the trial-to-trial variation is evitable in EEG signals, it is a challenge to capture the consistent classification features distribution. Clustering EEG trials with similar features and utilizing a specific classifier adjusted to each cluster can improve EEG classification. In this paper, instead of measuring the similarity of ERP features, the brain states during image stimuli presentation that evoked N1 responses were used to group EEG trials. The correlation between momentary phases of pre-stimulus EEG oscillations and N1 amplitudes was analyzed. The results demonstrated that the phases of timefrequency points about 5.3 Hz and 0.3 s before the stimulus onset have significant effect on the ERP classification accuracy. Our findings revealed that N1 components in ERP fluctuated with momentary phases of EEG. We also further studied the influence of pre-stimulus momentary phases on classification of N1 features. Results showed that linear classifiers demonstrated outstanding classification performance when training and testing trials have close momentary phases. Therefore, this gave us a new direction to improve EEG classification by grouping EEG trials with similar pre-stimulus phases and using each to train unit classifiers respectively.

How to Evaluate Phase Differences between Trial Groups in Ongoing Electrophysiological Signals

“…W-W, and never POP or PBI. We thus recommend abandoning these two measures, especially PBI which we had introduced in 2009 (Busch et al, 2009 ) and which has been independently employed in (at least) 13 publications since (Hamm et al, 2012 ; Ng et al, 2012 ; Auksztulewicz and Blankenburg, 2013 ; Hanslmayr et al, 2013 ; Manasseh et al, 2013 ; Rana et al, 2013 ; Diederich et al, 2014 ; Park et al, 2014 ; Li et al, 2015 ; Shou and Ding, 2015 ; Strauss et al, 2015 ; van Diepen et al, 2015 ; Batterink et al, 2016 ). Although there is no indication that PBI would have led authors to erroneous conclusions in any of these studies, it is also clear that more accurate results (or higher statistical power) would have been achieved using other measures.…”

“…Thus, phase opposition measures generally involve a combination (sum or product) of ITC for each trial group, appropriately corrected (by subtraction or division) to remove the overall ITC. For example, the “phase bifurcation index” (PBI) introduced by Busch et al ( 2009 ), and employed several times since (Hamm et al, 2012 ; Ng et al, 2012 ; Auksztulewicz and Blankenburg, 2013 ; Hanslmayr et al, 2013 ; Manasseh et al, 2013 ; Rana et al, 2013 ; Diederich et al, 2014 ; Park et al, 2014 ; Li et al, 2015 ; Shou and Ding, 2015 ; Strauss et al, 2015 ; van Diepen et al, 2015 ; Batterink et al, 2016 ), was based on this principle. Other analogous procedures have been described, however (Drewes and VanRullen, 2011 ; Dugue et al, 2011 , 2015 ; VanRullen et al, 2011 ; Han and VanRullen, 2016 ), and there exists no systematic comparison between these various measures and, consequently, no accepted practice in this field.…”

How to evaluate phase differences between trial groups in ongoing electrophysiological signals

“…W-W, and never POP or PBI. We thus recommend abandoning these two measures, especially PBI which we had introduced in 2009 (Busch et al, 2009) and which has been independently employed in (at least) 13 publications since (Hamm et al, 2012;Ng et al, 2012;Auksztulewicz & Blankenburg, 2013;Hanslmayr et al, 2013;Manasseh et al, 2013;Rana et al, 2013;Diederich et al, 2014;Park et al, 2014;Li et al, 2015;Shou & Ding, 2015;Strauss et al, 2015;van Diepen et al, 2015;Batterink et al, 2016). Although there is no indication that PBI would have led authors to erroneous conclusions in any of these studies, it is also clear that more accurate results (or higher statistical power) would have been achieved using other measures.…”

“…Thus, phase opposition measures generally involve a combination (sum or product) of ITC for each trial group, appropriately corrected (by subtraction or division) to remove the overall ITC. For example, the 'phase bifurcation index' (PBI) introduced by Busch et al (2009), and employed several times since (Hamm et al, 2012;Ng et al, 2012;Auksztulewicz & Blankenburg, 2013;Hanslmayr et al, 2013;Manasseh et al, 2013;Rana, Vaina, & Hamalainen, 2013;Diederich, Schomburg, & van Vugt, 2014;Park, Correia, Ducorps, & Tallon-Baudry, 2014;Li et al, 2015;Shou & Ding, 2015;Strauss et al, 2015;van Diepen, Cohen, Denys, & Mazaheri, 2015;Batterink, Creery, & Paller, 2016), was based on this principle. Other analogous procedures have been described, however Dugue et al, 2011;VanRullen et al, 2011;Dugue et al, 2015;Han & VanRullen, 2015), and there exists no systematic comparison between these various measures and, consequently, no accepted practice in this field.…”

Review of Brain Computer Interface Components and Corresponding Research Topics

“…In terms of instability of EEG signals among users, Reference [99] build a classifier collection based on user-specific spatio-temporal filters and use L1-norm regulated quadratic regression to make the classifier discrete, then the final classifier can be reliably generated to other users. Reference [92] considers that the inter-trial changes are inevitable and it is hard to extract consistent features, so the paper clusters the trials with similar characteristics and uses corresponding adjusted classifiers to eliminate the inert-trial instability. Reference [95] combines several classification methods, such as K Nearest Neighbour, Multilayer Perceptron, etc., for introducing robustness to the classifier.…”