Seizure Onset Localization From Ictal Intracranial EEG Data Using Online Dynamic Mode Decomposition

Epilepsy is one of the most common neurological diseases. In cases where patients do not respond to medications, resective surgery is often the next best option to obtain seizure freedom. Intracranial EEG analysis is the current gold standard for resective surgery planning. However, clinical marking is subjective, and many seizures are complex with ambiguous onset locations. The objective, in this proof-of-concept study, was to determine whether quantification with dynamic mode decomposition (DMD) may assist in localizing seizure onset. We analyzed one seizure each from five patients with epilepsy and identified channels with maximal involvement in the leading dynamic mode. In three of the five cases, the area of activity identified by our method showed statistically significant correlation with clinically identified channels. We conclude that DMD effectively captures the seizure onsets and is ready for future study in larger cohorts.