Title: Automatic localisation of epileptic foci in long-term continuous EEG

Authors: K.P. Indiradevi, Elizabeth Elias, P.S. Sathidevi

Addresses: Department of Electronics and Communication Engineering, National Institute of Technology, Calicut, NIT Campus P.O., Calicut-673 601 Kerala, India. ' Department of Electronics and Communication Engineering, National Institute of Technology, Calicut, NIT Campus P.O., Calicut-673 601 Kerala, India. ' Department of Electronics and Communication Engineering, National Institute of Technology, Calicut, NIT Campus P.O., Calicut-673 601 Kerala, India

Abstract: We describe a new algorithm for identifying the signal epochs that contain the epileptiform transients in the long-term continuous EEG and automatic localisation of exact epileptic foci in the human epileptic brain using two-level multi-resolution analysis. Our system consists of three stages: detection of epileptic events using two-level multi-resolution analysis, extraction of spike portion of signal from raw EEG data and localisation of epileptic foci. The key points are: the computation of an appropriate adaptive threshold, the localisation of epileptic foci and computation of number of spikes per second. The results suggest that epileptic foci could be quantitatively identified and accurately localised using two-level multi-resolution analysis. In comparison to visual identification of spike, the new approach is sensitive and provides the possibility of classifying epilepsies into generalised and focal epilepsies. The proposed algorithm has been implemented in MATLAB.

Keywords: adaptive thresholding; epilepsy; epileptic foci; inter-ictal epileptiform transients; IEDs; human brain; multi-resolution analysis; wavelet transform; automatic localisation; long-term EEG; continuous EEG; feature extraction; signal epochs.

DOI: 10.1504/IJMEI.2008.019476

International Journal of Medical Engineering and Informatics, 2008 Vol.1 No.1, pp.134 - 154

Published online: 13 Jul 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article