Abstracts

Rationale: Recent advances in the robotics technology has enabled precise and extensive stereo EEG (SEEG) electrode placement for intracranial EEG monitoring, which has become an extremely helpful strategy to identify epileptogenic zones. This has paved the way for generation of enormous amounts of data that needs to be analysed before determination of the epileptogenic focus. This has become extremely challenging with naked eye, as the number of electrodes implanted has increased. Appropriately designed deep neural network based tools may be used in addressing this challenge.

Methods: In this study we report the tools developed in identifying the epileptogenic focus from SEEG data using deep neural network. We used retrospective data from 10 patients who underwent successful epilepsy surgery (seizure outcone: ILAE 1-3), after SEEG monitoring. The ages of the patients ranged from 20 to 55 years. The suspected cause of the seizures included trauma, infection or prematurity. All the patients underwent initial video EEG evaluation in epilepsy monitoring unit where the typical clinical seizures were captured. In addition, they underwent MRI, SPECT, PET, MEG and Neuropsychological evaluation for localization of the epileptogenic zones. Epileptogenic zone was finally determined based on the SEEG monitoring by evaluating typical seizures. The interictal EEG data from the SEEG monitoring was used to develop a multilayer model to predict the seizure onset zones. The higher frequency components of the SEEG (60-600 Hz) was used for developing the model.

Results: A model was developed using conventional deep neural network layers (using keras package). Data from each patient was parsed into one-second epochs and a shuffled database containing sEEG from both epileptogenic and non-epileptogenic zones was generated which was used in training the model. Testing the model from the same database revealed 80-90% accuracy in predicting the channels which are potentially epileptogenic.

Conclusions: Our experience suggest that deep learning methods on interictal SEEG data can be used as an effective strategy for localizing epileptogenic zones which can be combined with other strategies of decision making for epilepsy surgery.

Funding: Please list any funding that was received in support of this abstract.: No specific funding source.