Authors :
Presenting Author: Jivesh Patel, BS candidate – University of California at Davis
Kira Kiviat, MS – University of California at Davis
Zachary McNaughton, BS – University of California at Davis
Caren Armstrong, MD PhD – UC Davis
Sheela Toprani, MD, PhD – University of California at Davis
Rationale:
Epilepsy affects over 50 million people worldwide, with nearly 3.4 million in the United States alone. Epilepsy patients face unpredictable seizures, impacting daily life by limiting activities such as driving, working, and socializing. This unpredictability contributes to significant morbidity. We are striving to develop noninvasive biomarkers to assess seizure risk, empowering patients to make informed decisions. Identifying low-risk periods allows for normal activities, while high-risk periods could prompt precautionary measures such as avoiding certain activities, using preventive medications, or taking immediate protective actions. However, seizure prediction is highly individualized. Current population-trained detection algorithms lack the precision needed for personalized forecasts. The goal of this project is to identify a high risk state minutes before a seizure and to distinguish high-risk vs low-risk days, ideally with only the limited data available from commercial neurostimulator devices.
Methods:
We are using intracranial EEG (iEEG) data collected from depth electrodes implanted in patients undergoing pre-surgical epilepsy monitoring. Electrodes are strategically placed within seizure onset zones (SOZ) and in non-SOZ regions to record electrical activity from adjacent brain tissue. We quantify the effective connectivity of the brain over time by calculating coherence between all pairs of recording channels in physiological frequency bands. Using principal component analysis (PCA) we reduce this data to a low-dimensional state space that can be visualized and classified via logistic regression. We have performed this analysis using recordings from 100+ iEEG contacts. We have also preliminarily applied it a subset of four contacts, either in the SOZ (as captured by closed-loop Responsive Neurostimulation (RNS) devices) or in areas outside the seizure onset zone (as recorded by open-loop Deep Brain Stimulations (DBS) devices).
Results:
Analysis of patients (n=11) using all 100+ iEEG contacts has shown that we can distinguish pre-seizure brain states seconds to minutes before a seizure. In preliminary work (n=1), we have found similar results using only four contacts, either in the SOZ or outside of the SOZ. We have also found (n=1) distinct clusters in state space during non-seizure baseline periods on days with many seizures compared to days without seizures.
Conclusions:
These results could provide patients early insight on high-risk days and greater predictive power minutes before seizure onset. This would shift epilepsy care from reactive to proactive, improving patient independence and quality of life. It also lays the groundwork for adapting iEEG models to simpler systems like RNS, DBS, or wearables for real-world use.Funding:
National Center for Advancing Translational Sciences, National Institutes of Health UL1 (# TR001860) and linked UC Davis Clinical Translational Science Center (CTSC) KL2 Award (# TR001859)