CAN MUSCLE ARTIFACT REMOVAL IMPROVE SEIZURE DETECTION IN SCALP EEG? Anneleen Vergult

CAN MUSCLE ARTIFACT REMOVAL IMPROVE SEIZURE DETECTION IN SCALP EEG?

Anneleen Vergult

, Pierre LeVan

, Sabine Van Huffel

, Jean Gotman

1

Electrical Engineering Department, Katholieke Universiteit Leuven, Leuven, Belgium and

Montreal Neurological Institute, McGill University, Montreal, Quebec, Canada This study investigated the value of automatic muscle artifact removal prior to

epileptic seizure detection in scalp EEG, as these artifacts can prevent the detection of the seizure onset.

Scalp EEGs from 7 randomly selected epileptic patients were analyzed, including 32 seizures. Canonical correlation analysis was applied as a blind source separation method for automatic muscle artifact removal (De Clercq et al IEEE Trans Biomed Eng, in press). Saab and Gotman’s seizure detector (Clin Neurophysiol 427:442, 2005) was applied to the original and processed EEGs. Sensitivity, detection delay and false detection rate were compared.

Both detection systems had a sensitivity of 88.3%. Removing muscle artifacts decreased the median detection delay from 14.8 to 12.5 seconds. Preprocessing also increased the false detection rate from 0.83 to 1.09/h by revealing previously obscured short bursts of rhythmic activity or rapid eye blinking that are known to be to main cause of false detections even in the unprocessed EEG. By tuning the seizure detector

threshold, a comparable improvement in detection delay could be achieved on the original EEGs, but at higher false positive rate.

The proposed muscle artifact removal method provides a general way of improving the detection delay in automatic seizure detection.

Research supported by Research Council KUL: GOA-AMBioRICS, CoE EF/05/006;

FWO project G.0360.05 (EEG, Epileptic), FWO- research communities (ICCoS,

ANMMM); Belgian Federal Science Policy Office IUAP P5/22; BIOPATTERN (FP6-

2002-IST 508803); by grant MOP-10189 of the Canadian Institute of Health Research.