File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Multiresolution metrics for detecting single-trial evoked response potentials (ERPS)

TitleMultiresolution metrics for detecting single-trial evoked response potentials (ERPS)
Authors
KeywordsBSS
Issue Date2004
Citation
Proceedings of 2004 International Conference on Machine Learning and Cybernetics, 2004, v. 7, p. 4240-4245 How to Cite?
AbstractIt is desirable to determine from electroencephalography (EEG) or magnetoencephalography (MEG) the time course of brain activation in response to sensory stimulation. Because of the relatively poor signal to noise ratio, evoked responses potentials (ERPs) were typically measured by averaging over multiple trials. While recent applications of blind source separation (BSS) and independent component analysius (ICA) improved the effective signal to noise ratio (S/N) by separating different brain sources and other extra-cranial sources, variations in the background on-going activity of each brain sources makes it difficult to determine whether and when an evoked response potential has occurred. We introduced and combined several new approaches to improve single-trial ERP detection from a previously reported MEG data set with relatively low S/N. First, new metrics based on multiresolution filtering were introduced to better discriminate a ERP against background oscillatory activity. Second, a novel interactive user interface was implemented to use the new metrics to detect single-trial ERPs from an example. Third, time series of brain source activation recovered using BSS were used as inputs to this multiresolution method. We report sharpened average ERPs after aligment using the detected single-trial ERP onset time and a reduction in false detection from the previously reported 26+/-2% to 13+/-2%.
Persistent Identifierhttp://hdl.handle.net/10722/228071

 

DC FieldValueLanguage
dc.contributor.authorLoring, Terry A.-
dc.contributor.authorWorth, David E.-
dc.contributor.authorTang, Akaysha C.-
dc.date.accessioned2016-08-01T06:45:07Z-
dc.date.available2016-08-01T06:45:07Z-
dc.date.issued2004-
dc.identifier.citationProceedings of 2004 International Conference on Machine Learning and Cybernetics, 2004, v. 7, p. 4240-4245-
dc.identifier.urihttp://hdl.handle.net/10722/228071-
dc.description.abstractIt is desirable to determine from electroencephalography (EEG) or magnetoencephalography (MEG) the time course of brain activation in response to sensory stimulation. Because of the relatively poor signal to noise ratio, evoked responses potentials (ERPs) were typically measured by averaging over multiple trials. While recent applications of blind source separation (BSS) and independent component analysius (ICA) improved the effective signal to noise ratio (S/N) by separating different brain sources and other extra-cranial sources, variations in the background on-going activity of each brain sources makes it difficult to determine whether and when an evoked response potential has occurred. We introduced and combined several new approaches to improve single-trial ERP detection from a previously reported MEG data set with relatively low S/N. First, new metrics based on multiresolution filtering were introduced to better discriminate a ERP against background oscillatory activity. Second, a novel interactive user interface was implemented to use the new metrics to detect single-trial ERPs from an example. Third, time series of brain source activation recovered using BSS were used as inputs to this multiresolution method. We report sharpened average ERPs after aligment using the detected single-trial ERP onset time and a reduction in false detection from the previously reported 26+/-2% to 13+/-2%.-
dc.languageeng-
dc.relation.ispartofProceedings of 2004 International Conference on Machine Learning and Cybernetics-
dc.subjectBSS-
dc.titleMultiresolution metrics for detecting single-trial evoked response potentials (ERPS)-
dc.typeConference_Paper-
dc.description.natureLink_to_subscribed_fulltext-
dc.identifier.scopuseid_2-s2.0-6344231603-
dc.identifier.volume7-
dc.identifier.spage4240-
dc.identifier.epage4245-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats