“Optimization of Responsive Neurostimulation (RNS)” is a project that aims to provide data-driven guidence to the physicians who examine patients with RNS devices. Specifically, the results of the project will help physicians better understand the patients’ clinical conditions and make better parameter adjustment based on the prediction.

We built a reliable classifier that classify clinical performance based on available background EEG data and examined the neural science intuition behind the machine learning model. We discovered that Sleep ECoG achieved better classification performance when compared to awake ECoG. The reason may be that sleep EcoG shares stronger correlations with ictal activity. We also found that background EEG appears to be equally valuable as pre-ictal EEG in predicting clinical outcome.

Presentation

Background

Some patients with refractory epilepsy are implanted with a responsive neurostimulation device (RNS NeuroPace, Inc., Mountain View, CA). The RNS is an FDA-approved device which delivers brief pulses of electrical stimulation upon detection of abnormal EEG patterns to terminate seizures. Physicians program RNS detection and stimulation parameters to potentially reduce seizure duration and frequency. Current practice is empirically driven, and physicians are limited by a lack of evidence for guiding stimulation settings. A data-driven approach could help physicians reach optimal therapeutic stimulation parameters more quickly, and avoid seizure exacerbation.