Epileptogenic Foci Localization with Intracranial EEG Using Effective Connectivity Network Based on Phase Transfer Entropy

One of the greatest challenges of achieving successful surgical outcomes in patients with epilepsy is the correct localization of the epileptogenic foci. Clinicians rely heavily on electroencephalogram (EEG) recordings, especially intracranial EEG (iEEG), to localize the epileptogenic zone and make surgical decisions. In this paper, we propose a method based on a Phase Transfer Entropy (PTE) metric targeting accurate automatic epileptic foci localization using the iEEG recordings. The phase of an oscillation has been shown to be critical in the coordination of anatomically distributed processing. We suggest using a PTE metric for the quantification of directed phase interactions between channels with high computational efficiency and a single parameter. First, infer causal connectivity using PTE between the multi-channel iEEG time series to establish the effective connectivity network of the brain with sliding windows. Then, compute features of that time-varying network, and eventually determine the epileptogenic foci. The proposed approach is evaluated on the HUP iEEG dataset, which contains patient data collected during surgical treatment of drug-resistant epilepsy. The results show the effectiveness as well as computational efficiency of the phase-based method in epileptogenic foci localization.