Abstracts

Rationale: Visualizing the spatiotemporal dynamics of seizure spread through conventional trace-based review of intracranial electroencephalography (ICEEG) has become more challenging in recent years, owing to practice trends involving high electrode densities and heterogeneous electrode locations. Emerging information visualization (InfoVis) approaches in computer science and digital graphics simplify complex data for human interpretation and can help expose latent patterns. Applying such practices to simplify and/or reduce the dimensionality of intracranial EEG data (e.g., 3-D anatomical space, time, frequency, amplitude, sharpness/waveforms features, etc.) may thus help reveal novel patterns (phenotypes) of ictal spread routes with clinical relevance.

Methods: We used a recently described open-source technology, omni-planar and surface casting of epileptiform activity (OPSCEA), to visualize the onset and spread of 147 seizures recorded from 75 patients with drug-resistant focal epilepsy who underwent intracranial monitoring with subdural electrodes at our center. For each seizure, we created a video showing all electrode locations and a metric of epileptiform activity projected as a heatmap onto a 3-D brain reconstruction. These videos were used to characterize the onset location(s) and spatiotemporal dynamics over the entire seizure duration, solely from this visualization approach.

Results: Using OPSCEA visualization alone, the onset and spread of seizure activity could be grouped into three categories: contiguous spread across the cortex after onset (wildfire-like; 57%), multi-focal at onset (27%), or initial contiguous spread followed by skipping over entire sublobar regions (“spread-skip”; 16%). Preliminary assessment of post-surgical outcomes in patients with at least one year of follow-up available (N=61, four had RNS only) revealed that patients with multi-focal and spread-skip patterns (grouped together) were surprisingly more likely to be free of disabling seizures (Engel I) than those with unifocal contiguous spread (Engel II-IV; p-value = 0.0460, chi-squared statistic = 3.9814). Ictal activity frequently appeared to hesitate at the borders of rolandic cortex (pre- and post-central gyri; 83% of contiguous spread seizures) or skip across this region entirely (23% and 50% of multi-focal and spread-skip seizures, respectively).

Conclusions: Re-interpreting conventional ICEEG tracings using only a 3-D anatomic heatmap-based visualization revealed three distinct seizure spread phenotypes with potential clinical relevance. We speculate that variable involvement or sparing of rolandic cortices during seizure spread might be driven by white matter connectivity differences and help explain the later timing of the bilateral tonic-clonic phase of seizures. Emerging InfoVis approaches may hold accelerating clinical utility for previously unrecognized features of seizure spread as intracranial EEG studies continue to rapidly scale in complexity in the coming decades.

Funding: Please list any funding that was received in support of this abstract.: Dr. Kleen was funded by NIH / NINDS grant K23NS110920.