Topographic analysis of ERPs using artificial neural networks

C. R. Clark

    Research output: Contribution to journalMeeting Abstractpeer-review

    Abstract

    Recent immunohistochemical work provides strong empirical support for the view that the neocortical substrates of cognitive function are organised as parallel and widely distributed neural networks. Event-related, scalp electrical potentials (ERPS) offer one of the few ways in
    which activity patterns in these networks may be investigated. With only few exceptions, however,
    topographic analysis of multi-electrode ERP activity focuses on the distribution of amplitude maxima and minima and in so doing limits the scope for identification of patterns associated with cognitive function.
    Artificial neural networks (ANNs) represent powerful analytical tools for pattern analysis of multidimensional data. One of the problem domains in which they have been used most successfully is that of spatiotemporal pattern recognition and recently it has been shown that
    they are able to extract the principal components or features resident in multidimensional datasets. As such, they offer considerable promise for the identification of
    parallel and distributed features in the spatiotemporal profile of ERP activity. This paper will (a) report on results which demonstrate the capacity of one ANN design to learn and encode consistent features in singletrial ERPs and (b) examine how these features are encoded and their potential use in visualising parallel and distributed information processing in cortical networks.
    Original languageEnglish
    Pages (from-to)60-60
    Number of pages1
    JournalBrain Topography
    Volume5
    Issue number1
    Publication statusPublished - Sep 1992
    EventPan Pacific Workshop on Brain Electric and Magnetic Topography - Melbourne, Australia
    Duration: 17 Feb 199218 Feb 1992

    Fingerprint

    Dive into the research topics of 'Topographic analysis of ERPs using artificial neural networks'. Together they form a unique fingerprint.

    Cite this