Optimization and Data Mining in Epilepsy Research: A Review and Perspective

During the past century, most neuroscientists have believed that epileptic seizures began abruptly in a matter of a few seconds before clinical onset. Since the late 80’s, there has been an explosion of interests in neuroscience research to predict epileptic seizures based on quantitative analyses of brain electrical activity captured by electroencephalograms (EEGs). Many research groups have demonstrated growing evidence that seizures develop minutes to hours before a seizure clinical onset. The methods in those studies include signal processing techniques, statistical analyses, nonlinear dynamics (chaos theory), data mining, and advanced optimization techniques. Although the past few decades have seen revolutionary of quantitative studies to capture seizure pre-cursors, seizure prediction research is still far from complete. Current techniques still need to be advanced and novel approaches need to be explored and investigated. In this paper, we will give an extensive review and prospective of seizure prediction research including various methods in data mining and optimization techniques that have been applied to seizure prediction research. Future directions of data mining and optimization in seizure prediction research will also be discussed in this paper. Successful seizure prediction research will give us the opportunity to develop implantable devices, which are able to warn of impending seizures and to trigger therapy to prevent clinical epileptic seizures.