Authors :
Presenting Author: Xinbing Zhang, BS – University of Minnesota
Zachary Sanger, PhD, MS – University of Minnesota
Thomas Lisko, BA – University of Minnesota
Robert McGovern, MD – University of Minnesota
Theoden Netoff, PhD – University of Minnesota
Rationale:
Deep brain stimulation (DBS) in the anterior nucleus of the thalamus (ANT) has been shown to be effective for treating drug-resistant epilepsy, but its long-term effect on ANT neural activity remains unclear. Through long-term monitoring of ANT local field potential in drug-resistant epilepsy patients, we identified that circadian and multiday cycles in LFP are phase-locked to patient-reported seizure events, and the circadian cycle was modulated over time in some patients.
Methods:
15 patients were implanted with Medtronic’s Percept PC in the anterior nucleus of the thalamus as part of a clinical trial (#NCT05493722), approved by the University of Minnesota’s IRB (2022). The Medtronic Percept records 10-minute LFP pre-defined band power by averaging across 60 Fourier transform results. A cumulative total of 21 years of theta/alpha (5-10 Hz) LFP power data recorded was analyzed. A Wavelet transform was applied to identify long-term LFP cycles compared to 1,000 simulated white-noise surrogates in each patient. The phase and amplitude of each cycle were estimated using a Hilbert transform. The synchronization index that quantifies how well seizures cluster at a certain phase of a cycle was calculated to determine if seizures were phase-locked to any periodic cycle using a Rayleigh test. Then, a response surface of calculated seizure likelihood modulated by the amplitude and phase of all cycles that are phase-locked to seizures is fitted using a Gaussian process regression. Additionally, we used Wilcoxon pair-wise comparison to detect significant change of circadian cycle amplitude over time.
Results:
Among 7 open-label patients who provided consistent seizure timepoints throughout the trial, we found significant circadian and various multiday cycles in all 7 patients compared to 1,000 simulated surrogates (p < 0.05). In 6 out of 7 patients who provided seizure diaries, we found that seizures are phase-locked to their circadian and/or multi-day cycle(s) (p < 0.05, Rayleigh’s test). Models for all 6 patients achieved forecasting performance better than chance (AUC >