McSharry, P.E., Smith, L.A. and Tarassenkeo, L., Nature Medicine Vol 9 (3): 241-242 2003
Abstract
Epilepsy is one of the most common serious neurological disorders, affecting 1% of the population at some time. Reliable and robust detection of seizure precursors would improve the quality of life of many epilepsy sufferers. It is likely that the processes underlying the electroencephalogram (EEG) signal are nonlinear, but there is little, if any, evidence that such signals reflect deterministic chaos. Regardless of fundamental dynamics, the relevant operational question is whether or not the information reflected in a proposed test statistic justifies its use (given its complexity). Can a complicated, novel and potentially nonlinear method systematically out-perform traditional 'linear' methods such as analysis variance, or provide independent and complementary precursor information?