Independent component analysis of auditory evoked potentials
Marko Perkušić · 2017
Cochlear implant (CI) uses electric pulses to stimulate the auditory nerve in deaf people, allowing them to have a partial sense of hearing. If these people with cochlear implants undergo electroencephalography (EEG) or other types of measuring bioelectric potentials such as auditory evoked potentials, recordings are typically contaminated with artifacts caused by electrical stimulation in CI. In this work, we used independent component analysis (ICA) for detection and removal of potential CI artifacts. ICA is based on data measured across many spatially distributed sensors (electrodes) which, when subjected to the process of orthogonalization may lead to independent components representing potential signal generators. Since CI artifacts possess specific temporal and spatial characteristics, we assume that ICA components containing these characteristics represent CI artifacts. ICA allows identification and removal of undesirable generators which can be used for reconstructing signal without artefactual components. Based on this approach, in this graduation thesis we used independent component analysis to identify potential artifacts in auditory evoked potentials recorded as brain responses of single CI user listening to large number of short auditory stimuli such as vowels and simple syllables.