Abstracts

Rationale: Intracranial EEG (iEEG) used to localize seizure focus in patients with medically intractable epilepsy for surgical evaluation, sometimes fails to satisfactorily narrow down the seizure onset zone due to its inherent narrow bandwidth and low spatial resolution. In some epilepsy centers, additional microelectrode grids and hybrid depth electrodes are implanted, which have identified electrographic activity over sub-millimeter scale tissue volumes, pathological high-frequency oscillations and epileptiform activity localized to microdomains before it is detected on the macroelectrodes. In this work, we aim to study the pathological evolution of seizures captured by ictal propagation across microdomains. Methods: We analyzed micro-iEEG seizure data recorded from 2 patients, each of which consisted of a period of 1-2 minutes before the clinical seizure onset and ended 20-40 s into the seizure. A multivariate autoregressive model was fitted to the micro-iEEG data of 5 s segments filtered in the 2-40 and 70-110 Hz range, using an arfit algorithm. The model order was chosen using a Bayesian information criterion. The model parameters were used to calculate a generalized partial directed coherence (gPDC), a frequency domain measure of causality between pairs of signals in a multivariate dataset. The gPDC values in the 2 frequency bands calculated from the ictal period were used to determine any significant propagation by comparing them with the 98th percentile of the corresponding null distributions estimated from baseline segments of a few minutes that were temporally separated from the seizure. Results: In the first patient (Fig 1), one microelectrode grid(AMIC) was close to the clinical seizure focus while the other (PMIC) was in between the macroelectrodes involved early(1,2,11) and those involved later with seizure spread(29,30). The majority of the 2-40 Hz ictal propagation in AMIC is initially towards the focus and then it rotates away from it, while that in the PMIC changes direction from being towards the region involved later towards the focus. The 70-110 Hz propagation in both the microelectrodes is mostly towards the focus. In the second patient (Fig 2), there were 3 microelectrode grids surrounding the resected region, with microelectrode 3 being closest to the involved macroelectrodes. The 2-40 Hz propagation in 3 and 2 are towards the focus while that in 1 is away from it. In 3, we also saw similar patterns for both the frequency bands in the initial phase. Conclusions: In the two patients considered, we saw changing propagation of the micro-iEEG, lower frequency components showing higher and longer range flows during the initial seizure phase that decrease into the ictal event. Higher frequency components, specifically over regions close to the focus, tend to have a more stable flow direction. Study of µiEEG recordings of ictal activity can thus reveal potentially useful seizure dynamics over micro scales. This might help in future seizure localization for a precise targeting for improved surgical outcomes and for anti-epileptic neuromodulation systems.