Inferring causal connectivity in epileptogenic zone using directed information

Directed information, an information theoretic quantity, is developed in this paper to infer the causal connectivity from electrocorticography (ECoG) recordings of an epileptic patient. The causal connectivity can be used to infer the optimal electrodes for electrical stimulation based treatments of epilepsy. A parametric estimator for directed information between two ECoG signals is also proposed. The estimator estimates entropy and causally conditioned entropy and their difference is the estimate of DI. The estimator is then applied to ECoG data recorded from the electrodes in the epileptogenic zone (EZ) in two patients with focal epilepsy to learn the changes in causal connectivity during seizures.