Abstracts

Rationale: The critical point in epilepsy surgery is to localize the epileptogenic zone (EZ). This may require invasive intracranial electroencephalography and long-term monitoring (LTM) in order to observe ictal activities. Computational analysis of interictal data may be able to define the EZ without the need to record ictal events. To this end, Granger’s causality (GC) could be a particularly promising option. Moreover, different global physiological states, such as sleep versus wakefulness, may change the connectivity map revealed by the GC method.  Methods: This study of 10 consecutive patients who underwent stereoelectroencephalography (sEEG) implantation for LTM aimed to quantify the congruence of GC-defined putative EZ, using the outdegree metrics and based on the analysis of 20-minute interictal recordings from each state for each patient, with the clinically-defined EZ from conventional visual analysis of prolonged (days) recordings comprising both ictal and interictal epochs. We compared the performance of GC analysis from sleep and wakefulness recordings. One patient was excluded from the congruence study as no EZ was defined by the neurologists but kept for the sleep versus wakefulness study.  Results: The GC predicts the EZ defined by conventional analysis with high probability and GC analysis of either state, sleep versus awake, showed similar congruence with this clinically-defined EZ (combined p-values <10^-31, exact rank order sum test by restricted partitions of integers) and also significant correlation (Spearman’s ? of 0.75, p-value <10^-15). Individual and additional results are shown in Table 1. and Table 2. Conclusions: GC proves promising ability to define the EZ. Both sleep and wakefulness recordings are suitable for GC analysis. Future perspectives such as assisting interictal EEG reading or enhancing surgical decision-making will require its optimization, comparison to other methods and ultimate validation. Funding: No funding