Preictal bursting activity may predict seizure onset

Introduction: Drug-resistant epilepsy (DRE) outcomes have been largely stagnant over the past two decades. Although recent advances have greatly improved seizure control, there is still much work to be done. Our group has recently shown that single unit bursting activity is uniquely stereotyped across ictal events, suggesting that bursting activity may reveal patterns in seizure activity that have not yet been explored. Here, we show that preictal bursting patterns begin to uniquely organize in a low-dimension manifold up to one-hour prior to ictal onset, suggesting a potential utility in advanced seizure prediction and intervention.

Methods: We obtained intracranial recordings from epilepsy patients implanted with Utah Intracortical Electrode Arrays (UIEA) undergoing neuromonitoring at the NIH Clinical Center. We bandpass filtered signals (0.5-5 kHz) using a 4th-order Butterworth filter and subsequently smoothed signals using a gaussian kernel (σ=25 ms). We then detected baseline, preictal ( < 1 hr from seizure), and ictal bursting activity by applying a z-score threshold of three. We then used Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) to represent bursts in two dimensions.

Results: We visualized bursts on a UMAP manifold and quantified similarity of pre-ictal bursts to baseline and ictal bursts by measuring distance to baseline centroid. Preictal bursts localized spatially near ictal bursts on the UMAP manifold, starting ~1 hr prior to ictal onset, whereas baseline bursts are randomly distributed. Furthermore, preictal bursts deviate significantly from baseline centroid (P=3.00e-8), akin to ictal bursts (P=1.92e-138), suggesting that preictal bursting patterns are similarly uniquely stereotyped.

Conclusion : These findings suggest that changes in bursting activity may begin much earlier than electrographic ictal onset. Ultimately, our findings portend the possibility of predicting seizures well in advance of ictal onset, opening doors for significantly improving preexisting treatments, such as closed-loop neuromodulation, but also paving the way for novel therapeutics.