000185106 001__ 185106
000185106 005__ 20190316235616.0
000185106 0247_ $$2doi$$a10.1088/1741-2560/10/3/036014
000185106 022__ $$a1741-2560
000185106 02470 $$2ISI$$a000319510800017
000185106 037__ $$aARTICLE
000185106 245__ $$aSingle trial analysis of slow cortical potentials: A study on anticipation related potentials
000185106 260__ $$bInstitute of Physics$$c2013$$aBristol
000185106 269__ $$a2013
000185106 300__ $$a12
000185106 336__ $$aJournal Articles
000185106 520__ $$aObjective. Abundant literature suggests the use of Slow Cortical Potentials (SCPs) in a wide spectrum of basic and applied neuroscience areas. Due to low signal to noise ratio, often these potentials are studied using grand-average analysis, which conceal trial-to-trial information. Moreover, most of the single trial analysis methods in literature are based on classical- electroencephalogram (EEG) features ([1–30] Hz) and are likely to be unsuitable for SCPs that have different signal properties (such as signal’s spectral content in the range [0.2–0.7]Hz). In this paper we provide insights into the selection of appropriate parameters for spectral and spatial filtering. Approach. To this end, we study anticipation related SCPs recorded using a web-browser application protocol using full-band EEG (FbEEG) setup from 11 subjects on two different days. Main results. We first highlight the role of a bandpass with [0.1–1.0]sHz in comparison with common practices (e.g., either with full DC, just a lowpass, or with a minimal highpass cut-off around 0.05Hz). Second, we suggest that a combination of spatial-smoothing filter (SSF) and common average reference (CAR) is more suitable than the spatial filters often reported in literature (e.g., re-referencing to an electrode, Laplacian or CAR alone). Third, with the help of these preprocessing steps, we demonstrate the generalization capabilities of linear classifiers across several days (AUC of 0.88 ± 0.05 on average with a minimum of 0.81 ± 0.03 and a maximum of 0.97 ± 0.01). We also report the possibility of further improvement using a Bayesian fusion technique applied to electrode-specific classifiers. Significance. We believe the suggested spatial and spectral preprocessing methods are advantageous for grand-average and single trial analysis of SCPs obtained from EEG, MEG as well as for electrocorticogram (ECoG). The use of these methods will impact basic neurophysiological studies as well as the use of SCPs in the design of neuroprosthetics.
000185106 700__ $$0242177$$g176513$$aGaripelli, Gangadhar
000185106 700__ $$aChavarriaga, Ricardo
000185106 700__ $$aMillán, José del R.$$0240030$$g149175
000185106 773__ $$j10$$tJournal of Neural Engineering$$k3$$q036014
000185106 8564_ $$uhttps://infoscience.epfl.ch/record/185106/files/Garipelli_JNE_2013.pdf$$zn/a$$s1251601$$yn/a
000185106 909C0 $$xU12103$$0252018$$pCNBI
000185106 909C0 $$pCNP$$xU12599$$0252517
000185106 909CO $$qGLOBAL_SET$$pSTI$$particle$$ooai:infoscience.tind.io:185106
000185106 917Z8 $$x137762
000185106 917Z8 $$x137762
000185106 917Z8 $$x137762
000185106 937__ $$aEPFL-ARTICLE-185106
000185106 973__ $$rREVIEWED$$sPUBLISHED$$aEPFL
000185106 980__ $$aARTICLE