Epilepsy Research 68 (2006) 9–18 Seizure anticipation, states of consciousness and marginal predictability in temporal lobe epilepsy Dingzhou Li a , Weiping Zhou d , Ivo Drury e , Robert Savit a,b,c,∗ a Physics Department, University of Michigan, Ann Arbor, MI 48109, United States b Biophysics Research Division, University of Michigan, Ann Arbor, MI, United States c Michigan Center for Theoretical Physics, University of Michigan, Ann Arbor, MI, United States d Department of Biostatistics, University of Michigan, Ann Arbor, MI, United States e Department of Neurology, University of Michigan Medical School, Ann Arbor, MI, United States Received 24 May 2004; received in revised form 13 September 2005; accepted 14 September 2005 Available online 13 December 2005 Abstract Purpose: It has recently been shown that differences between the marginal predictability associated with scalp electrodes adjacent to and remote from the site of a seizure focus are able to distinguish between epochs temporally distant from and just prior to (within about 20 min) the onset of a seizure in patients with temporal lobe epilepsy. The purpose of this paper is to show that these differences of marginal predictability intervals are independent of the state of consciousness of the patient. Methods: We have studied a data set encompassing 33 preictal epochs (within 1h prior to a seizure) and 61 interictal epochs (defined as at least 1 h away from any seizure) from 14 patients. Each 30 s interval of each epoch was categorized into one of seven different states of consciousness. Statistical models were used to search for relationships (in aggregated data) between the values of differences of marginal predictabilities and state of consciousness. Results: It was not possible to reject the null hypothesis of no relationship between differences of marginal predictabilities and state of consciousness. Conclusions: The values of the differences between marginal predictabilities on aggregated data are apparently insensitive to the state of consciousness. This conclusion, coupled with the fact that the differences between marginal predictabilities do depend on time to seizure, suggests the potential utility of these measures as the basis for ambulatory, non-invasive methods of seizure anticipation. However, the development of a practical non-invasive method for seizure anticipation requires further extensive study on disaggregated data from individual patients. © 2005 Published by Elsevier B.V. Keywords: Epilepsy; Seizure; Seizure prediction; Seizure anticipation; Nonlinear dynamics; Nonlinear time series ∗ Corresponding author. Tel.: +1 734 764 3426; fax: +1 734 994 4208. E-mail address: savit@umich.edu (R. Savit). 0920-1211/$ – see front matter © 2005 Published by Elsevier B.V. doi:10.1016/j.eplepsyres.2005.09.030