| This work assesses the efficacy of three prediction schemes, wavelet entropy and a connectionist measure for their abilities to successfully anticipate an experimental model of a seizure in a timely manner. Wavelet entropy quantifies the order of a system by ascertaining the relative energy of each frequency band, subsequently, an impending seizure is deduced as this quantifier decreases below a predetermined threshold. It was shown to be a robust method with a potential to adequately forecast a seizure prior to its ictal outburst. A second approach is comprised of a connectionist measure implemented as a neural network with wavelet function nonlinearities. It is able to classify the underlying dynamics of spontaneous in vitro events into interictal, preictal and ictal activities and successfully predict the onset of a seizure. |
