Analysis of epileptic seizure count time series by ensemble state space modelling

We propose an approach for the analysis of epileptic seizure count time series within a state space framework. Time-dependent dosages of several simultaneously administered anticonvulsants are included as external inputs. The method aims at distinguishing which temporal correlations in the data are due to the medications, and which correspond to an unrelated background signal. Through this method it becomes possible to disentagle the effects of the individual anticonvulsants, i.e., to decide which anticonvulsant in a particular patient decreases or rather increases the number of seizures.