Authors :
Presenting Author: Shayne Hastings, BA – University of Minnesota
Kat Paige, BS – University of Minnesota
Xinbing Zhang, BS – University of Minnesota
Chris Krook-Magnuson, MS – University of Minnesota
Esther Krook-Magnuson, PhD – University of Minnesota
Rationale:
Neuromodulation therapies, including electrical stimulation, are a promising method for treating drug resistant epilepsy. We previously used Bayesian optimization to demonstrate that electrical stimulation of the cerebellar vermis can - with the correct stimulation parameters – suppress hippocampal seizures (Stieve et al. Brain, 2022). Optogenetic work suggests that excitation or inhibition of cerebellar cortex can stop seizures, but only optogenetic excitation, not inhibition, of the cerebellar fastigial nucleus can stop seizures. Therefore, electrical stimulation targeting the nuclei, rather than the cerebellar cortex, may require very different stimulation parameters or may not be successful at all. Bayesian optimization provides an effective, data-driven approach to explore which stimulation settings may allow for strong seizure suppression via electrical stimulation of the cerebellar fastigial nucleus. Additionally, to prevent adverse motor side effects, the stimulation parameter space for the fastigial must be reduced with lower testable frequency, charge, and train duration compared to the cerebellar cortex. Given these constraints, we aim to test if the fastigial nucleus can serve as an effective target for seizure control using electrical stimulation. Methods:
To generate chronically epileptic mice, we use the intrahippocampal kainic acid mouse model of temporal lobe epilepsy. For on-demand intervention, animals receive electrical stimulation in the fastigial nucleus selectively at the time of detected seizures recorded in the hippocampal local field potential. We varied stimulation charge, frequency, and pulse width, resulting in more than one thousand different parameter combinations. To explore this parameter space efficiently, we used Bayesian optimization with Gaussian process regression. After the optimization process, each mouse underwent additional testing with the optimal, non-optimal and no-stimulation settings to compare the effects of each on seizure duration.
Results:
Preliminary results suggest that electrical stimulation of the cerebellar fastigial nucleus with optimal parameters can robustly inhibit hippocampal seizures (n=6 mice). Stimulation with non-optimal parameters has no significant effect (n=6 mice). Conclusions:
Our preliminary findings suggest that on-demand fastigial nucleus stimulation can strongly suppress hippocampal seizures. Identifying optimal parameters is crucial, as non-optimal parameters yield no benefit. Even brief 50ms fastigial nucleus stimulation proved highly effective at terminating hippocampal seizures. This work underscores the power of Bayesian optimization in refining neuromodulation strategies and highlights the fastigial nucleus as a promising target for seizure control. Funding:
NIH R01-NS112518, University of Minnesota MnDRIVE Brain Conditions Initiative, University of Minnesota McKnight Presidential Fellowship