Adaptive Neurostimulation with Multi-Stage Control Loop: Application to Drug Resistant Epilepsy

presentation, poster

English

Responsive neural stimulation (RNS) delivers electrical stimulation in a brain region in response to brain activity to treat neurological disorders. RNS directed at seizures and abnormal epileptiform activity proved to be successful in reducing seizures in drug resistant epilepsy, but efficacy optimization can be slow. There is ongoing interest in systems where stimulation paradigms and network targets are optimized adaptatively over time. Here we present a system where a cascade of detectors in implanted neural sensing and stimulation (INSS) device and a handheld device enabling machine learning seizure detection algorithms to adaptively control stimulation parameters and brain targets. We developed a system for multi-stage control over adaptive neurostimulation paradigms. The system utilizes an INSS with continuous real-time intracranial EEG (iEEG) recording and bi-directional communication with a handheld for off-device programmable control. The adaptive RNS settings on the INSS device engage multiple different network targets. In this application, the initial electrical stimulation targets detected hippocampal seizures. If the seizure activity continues despite the responsive stimulation for a specified time interval, the algorithm adaptively switches to a second network target with different stimulation parameters. The detected events are further evaluated by detection algorithms running on the handheld device and can alter the stimulation paradigm. We tested the detector cascade system in benchtop settings on 200 seizures and 6 interictal days using hippocampal iEEG previously recorded from 6 patients undergoing evaluation for epilepsy surgery. We measured performance of the second cascade detector which is engaged after initiation of the first detector and achieved 96% sensitivity and 99.5% specificity with 10.62 seconds average detection lag from seizure onset. The off-device detector was developed on large dataset from the NeuroVista study (5 dogs and 10 humans with more than 10 years of data). The detector was tested using data from 4 patients (181 seizures over 496 days in total). Our off-device detector improved the specificity up to 99.9% which enables more control over adaptive network stimulation. We developed and tested the RNS system with multi-stage control loop for adaptive network stimulation. The system enables hypersensitive RNS targeting abnormal epileptiform activity that runs concurrently with a second INSS cascade detector for detecting seizures of a specified length and triggers RNS of a different network target. Deployment of the system to an INSS might prove useful in disrupting epileptogenic networks.

epilepsy; implantable neural stimulator

SLADKÝ, V.; KŘEMEN, V.; MÍVALT, F.; BRINKMANN, B.; LUO, R.; YOTTER, C.; GUNAWARDANE, N.; LITVINOV, B.; JOBST, B.; BLUMENFELD, H.; WORRELL, G.

8. 11. 2021

Society for Neuroscience

Chicago, the United States of America

1

1

1