Stereo-EEG is being increasingly used for pre-surgical epilepsy monitoring. Standard methods of evaluation with these recordings rely on visually identifying EEG changes in specific contacts to indicate the seizure onset zone. The objective is to apply a novel quantitative measurement, non-linear information transfer (NTE), from such recordings to identify additional information that may not be apparent with conventional linear measures or on visual evaluation of EEG.

Epilepsy is one of the most common chronic neurological disorders and affects approximately 3 million people in the USA. Surgery is the most effective treatment option for patients with epilepsy with drug-resistant epilepsy (DRE). As part of the surgical evaluation, patients may undergo intracranial EEG monitoring (IEEG), which is considered a diagnostic procedure for selected patients with treatment-resistant epilepsy. The purpose of IEEG is to directly record EEG suspected brain areas in order to precisely identify the region of seizure onset and early spread. The major challenge in epilepsy surgery is that even with invasive IEEG monitoring and extensive pre-surgical evaluation, a significant portion of patients continues to have seizures post-operatively. This might be due to improper identification and resection of the dysfunctional network.

We have developed an automated method to apply non-linear information transfer entropy on patients who already underwent intracranial EEG. These patients met the following inclusion criteria: 1) Patient’s with known medically refractory localization-related epilepsy. 2) Surgical outcome available at six-months post-surgery. 3) Accurate co-registration of intracranial electrode available for correct data analysis.

As part of preliminary process, we analyzed two patients who met the aforementioned criteria. We found that our automated method found and in some cases predicted the ictal onset and early spread. It also accurately determined the dysfunctional network. 

Based on our preliminary results, we have found that our method automatically identify the dysfunctional network.