Abstracts

The aim of this study was to develop and investigate the performance of automated muscle artifact removal from ictal electroencephalograms (EEG) based on canonical correlation analysis (CCA). CCA separated a 10s EEG epoch in a set of components (or sources) with a decreasing autocorrelation index (AI). It has been noticed that muscle artifact components have a lower AI than genuine EEG components. A neurologist removed muscle artifact from 317 epochs of ictal EEG of 40 patients by gradually removing the components starting from the one with the lowest AI. The neurologist determined subjectively the separation point where as much muscle activity was removed as possible, without affecting the brain activity significantly. This yielded a gold standard binary classification of the CCA components into muscle artifact components and EEG components. Next a fully automated classification of each CCA component was performed based on the ratio of spectral energy of the frequency band 25-50Hz to the band 10-15Hz. It is anticipated a large value of this feature will be observed when muscle activity is present in a component. The optimal feature threshold was selected by maximally remaining EEG components and maximally suppressing muscle artifact components. The developed automated method was able to remain on average 99[plusmn]0.3 % of the EEG component energy, while removing on average 71[plusmn]33 % of the muscle artifact component energy. An automated muscle artifact removal technique was presented, that could be used as pre-processing step in the early detection of ictal activity or as a filter in the visual evaluation of ictal EEG.Currently the same EEG[apos]s are being processed by other neurologists, to investigate the user dependency of the optimal separation point to obtain a more generalized gold standard which can improve the robustness of the automated muscle artifact elimination. (Supported by European network of excellence BIOPATTERN (FP6-2002-IST-508803) and Fund for Scientific Research Flanders FWO research project nr G.0360.05. Bart Vanrumste is funded by the [apos]Programmatorische Federale Overheidsdienst Wetenschapsbeleid[apos] of the Belgian Government.)