Abstracts

Rationale: Knowledge of relevant statistical features such as probability distributions (pdf) of energies (E) and inter-seizure intervals (ISI) of clinical and subclinical seizures is lacking in epileptology. This work investigates these features. Methods: Pdf’s and the probability of seizure (Sz) occurrence conditional upon the time elapsed from the previous Sz were estimated using E and ISI obtained from prolonged recordings from 60 subjects with “focal” pharmaco-resistant epilepsy, undergoing surgical evaluation, on reduced doses of anti-seizure medications. Seizure onset and end times and peak intensity were determined using a validated detection algorithm. E corresponded to the product of peak intensity and duration and ISI to the time elapsed between consecutive onsets. Results: E and ISI distributions for seizures appear to follow power laws. In humans there is increased probability of Sz occurrence before and after a seizure and the expected time to next seizure increases with the duration of the seizure-free interval since the last one. Conclusions: The cumulative empirical evidence suggests that, over short time scales, seizures have the inherent capacity to trigger other seizures. Power law distributions of E and ISI raise the possibility that these features lack a typical size/duration and may not be accurate/sufficient criteria for classifying paroxysmal activity as ictal or interictal. The uncovered temporal dependencies and the existence of power law distributions may be harbingers of predictability of time of Sz occurrence and intensity.