The identification of the epileptogenic network, the ensemble of brain regions involved in the genesis and propagation of epileptic seizures, is crucial for surgery. The most important feature of the seizure onset zone (SOZ) is that its activity is correlated with neighboring and distant sites. Quantitative measures of the degree of epileptogenicity typically characterize the electrical activity in the frequency domain. In the present work, we set the foundations of a new Epileptogenicity Index by estimating the amount of directional correlations originated in the SOZ before and during a seizure. Our method relies on both the signal frequency spectrum and shape. The estimation method is evaluated with synthetic data and then applied to recordings from a patient with temporal lobe epilepsy.