Real-time computer aided analysis of the electroencephalogram: A two-dimensional approach

This thesis is concerned with the application of computer aided measurement techniques to the analysis of electroencephalographic (EEG) signals. In particular, it deals with the detection of spike-and-slow-wave complexes (SAWCs) which are characteristic of the onset of an epileptic absence. Numerous algorithms have been developed to provide the above detection and a review of some of the more important of these is presented. From this review, an alternative model and algorithm are developed based on an FIR (finite impulse response) filter, that represents a differentiator in series with a Hilbert transformer, followed by a cubic filter, which effectively re-duces the background noise. This algorithm, like its predecessors, suffers from the difficulty of where the detection levels should be set in order to achieve no missed features with no false alarms. An alternative approach is then investigated based on the Wigner Distribution, which provides a simultaneous time and frequency domain analysis. The results of this work are presented and appear to offer significant advantages over the more usual one-domain analyses. Finally, some discussion of the results obtained and suggestions for further work are made, mainly in the area of improving the computational speed of the Wigner Distribution and associated SAWC detection algorithms