A Stochastic Analysis of Ictal-Interictal Transitions During Experimental Seizures: Seizure Duration Depends on the Duration of Preceding Ictal and Interictal States
Abstract number :
1.174
Submission category :
Year :
2000
Submission ID :
2880
Source :
www.aesnet.org
Presentation date :
12/2/2000 12:00:00 AM
Published date :
Dec 1, 2000, 06:00 AM
Authors :
Sridhar S Sunderam, James F Watkins, Mark G Frei, Ivan Osorio, Flint Hills Scientific, LLC, Lawrence, KS; Univ of Kansas Medical Ctr, Kansas City, KS.
RATIONALE: Little is known about the effect of seizure duration on the length of the subsequent interictal interval. Investigation of this ignored aspect of epilepsy dynamics has heuristic and predictive value. This study correlates ictal and interictal duration in rats, and uses the results to build a model and to test its predictive value. METHODS: Seizure (Ts) and interictal (Ti) durations were measured visually via ECoG from eleven anesthetized Sprague-Dawley rats treated with 3-MPA. Mean length of recordings was 130 min. Using these data, the following were to be predicted: Ts given the preceding interval (Ti1), and duration of the post-seizure interval (Ti2) given Ts. Suitable probability density functions (pdf) were proposed for Ts and Ti. The correlations between Ti1 and Ts and between Ts and Ti2 were both modeled as bivariate distributions, and conditional expectations E(Ts?Ti1) and E(Ti2?Ts) were used for predicting Ts and Ti2 respectively. RESULTS: Ts and Ti were mixtures of two underlying lognormal classes: short (S) and long (L) based on whether the duration was less or more than 32 s for Ts and 48 s for Ti. Scatter plots of Ti1 vs. Ts and Ts vs. Ti2, showed clusters corresponding to the transitions S?S, S?L, L?S, and L?L, and their relative frequencies were tabulated. S?S and L?L transitions were significantly paired (p < 0.001). Each cluster was modeled as bivariate lognormal, with an associated transition probability for the Ti1?Ts and Ts?Ti2 transitions. Given Ti1 (or Ts), Ts (or Ti2) was estimated. A model based on half the data randomly selected from a total of 83 transitions, gave a median error between observed and predicted values for the other half, of 2.1 s for Ts (given Ti1), and 6.4 s for Ti2 (given Ts). In comparison, ranges for observed values were 2.3 < Ts < 120 and 6.6 < Ti < 782. CONCLUSIONS: In this epilepsy model, seizure duration determines the length of the subsequent interictal interval, which in turn determines the duration of the next seizure. This temporal correlation between Ts and Ti allowed satisfactory prediction of the system's behavior using a stochastic model.