Differences in source analysis accuracy of AEP generators following FastICA and TDSEP-ICA denoising

Different factors can affect the accuracy of the source localization of evoked potentials: the SNR of the data, the head model, and the number of sources, mention but a few. Another fundamental factor is the correct application of independent components analysis (ICA) to both filter out artefacts and select the independent components (IC) related to the neurological response. In this paper we assess the differences in the source analysis accuracy of the auditory evoked potential (AEP) generators between a temporal ICA (TDSEP-ICA) algorithm and a statistically based algorithm (FastICA) in EEG, from children with cochlear implants (CIs). The parameters used to compare both ICA algorithms and validate the source localization include: residual standard deviation, confidence ellipsoids, localization and time of best dipole fit. Here we show that pre-processing using TDSEP- ICA facilitates the identification of the AEP peaks and the source analysis procedure. The results obtained in this research could be used as objective technique for a general evaluation of the performance of children with CIs.