Related Experiment Video
Updated: Aug 27, 2025
05:19
Using Neuron Spiking Activity to Trigger Closed-Loop Stimuli in Neurophysiological Experiments
Published on: November 12, 2019
7.1K
Online Learning Koopman Operator for Closed-Loop Electrical Neurostimulation in Epilepsy.
IEEE Journal of Biomedical and Health Informatics
|September 28, 2022
Summary
This study introduces a novel Koopman-MPC framework for adaptive electrical neuromodulation to control epilepsy. This advanced system improves seizure prediction and suppression using a deep Koopman operator model for real-time, efficient treatment.
Area of Science:
- Neuroscience
- Control Theory
- Machine Learning
Background:
- Electrical neuromodulation is a growing palliative treatment for epilepsy.
- Current methods lack adaptive stimulation capabilities, relying on predetermined strategies.
Purpose of the Study:
- To develop a real-time, closed-loop electrical neuromodulation framework for epilepsy control.
- To enable self-adaptive adjustment of stimulation inputs for improved seizure management.
Main Methods:
- Proposed a Koopman-MPC framework integrating a deep Koopman operator model for predicting epileptic EEG.
- Utilized a model predictive control (MPC) module for optimal seizure suppression strategies.
- Embedded the Koopman operator in an autoencoder's latent space for online approximation and updating.
Main Results:
- The deep Koopman operator model demonstrated superior predictive accuracy over baseline models (VAR, kernel, RNN) on synthetic and real epileptic EEG data.
- Koopman-MPC achieved better computational efficiency in suppressing seizure dynamics compared to RNN-MPC in Jansen-Rit and Epileptor models.
Conclusions:
- The Koopman-MPC framework offers a new approach for model-based closed-loop neuromodulation in epilepsy.
- This framework provides insights into nonlinear neurodynamics and effective feedback control policies for seizure management.
