ABSTRACT
Seizure prediction through analysis of the electroencephalogram (EEG) remains a challenging and thus far unsuccessful task (Mormann et al. 2007). Algorithms employing measures such as accumulated energy and correlation density showed promise to predict seizures on select EEG segments, but have failed to show reproducibility when extended to unselected, long-term EEG (Lehnertz and Elger 1998; Litt et al. 2001). The general seizure prediction approach is based upon continuously calculating a parameter of interest using a moving window, with a seizure being predicted whenever the calculated metric moves outside a prede–ned range of values. This chapter discusses the circadian rhythm and its effects on seizure occurrence, and raises the question as to whether seizure prediction could be enhanced by taking the known information about the in§uence of the circadian rhythm into account. For instance, algorithms may bene–t by simply varying-in circadian-like fashion-the threshold a measure must meet to predict a seizure. Recent research has provided evidence that seizures are more likely to occur at particular times of the day, depending on the region of the brain in which the epileptogenic focus is found (Durazzo et al. 2008; Hofstra et al. 2009; Pavlova, Shea, and Brom–eld 2004; Quigg and Straume 2000; Quigg et al. 1998; Hofstra et al. 2009). Circadian-like distributions of seizures have been observed at both the individual and group levels, generating interest in the possibility of incorporating information of the circadian rhythm into seizure prediction algorithms.
