Seizure focus localization is the key to control seizures. However, in this paper, we show that the clinically localized seizure focus may be not exactly the positions to abate seizures. Firstly, the reliability of a previously proposed methodology employed to estimate the synchronicity and directionality of information flows over time between EEG signals, is numerically assessed with a coupled mass neural model. Then 10 channels' EEG signals from a patient with focal epilepsy are used to reconstruct the dynamical complex network of pathological seizure. This may facilitate to identify the evolution paths of information flows and localize the potential seizure foci. What's more, based on the controllability and observability principles of complex systems, we can focus on the key nodes which is effective to control the network seizure behaviors and the key ones that can allow us to estimate the state of all other variables. Results show that to fully control the epileptic network may not just be related to the focus zone, it may also involves in other non-focus nodes. In addition, we use the spatiotemporal neural network model connected by our modeled dynamical adjacent matrix to successfully reproduce the original EEG signals which can be effectively abated by applying the normal dis-