Related Experiment Videos
Predicting epileptic seizures from scalp EEG based on attractor state analysis
Hyunho Chu1, Chun Kee Chung2, Woorim Jeong3
1Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Daejeon, Republic of Korea.
Computer Methods and Programs in Biomedicine
|April 11, 2017
Summary
Researchers identified a new seizure precursor using attractor state analysis in scalp electroencephalography (EEG). This method offers a novel way to predict epileptic seizures, improving patient safety and quality of life.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Medical Engineering
Background:
- Epilepsy is a prevalent neurological disorder impacting patients' quality of life due to unpredictable seizures.
- Accurate seizure prediction can enable timely interventions, enhancing patient safety and well-being.
Purpose of the Study:
- To investigate a novel seizure precursor based on attractor state analysis for predicting epileptic seizures.
- To develop and validate a low-complexity seizure prediction algorithm using scalp electroencephalography (EEG).
Main Methods:
- Analyzed the transition from normal to seizure attractor states in EEG data.
- Defined a quantified spectral measure based on Fourier coefficients of EEG frequency bands.
- Computed the spectral measure using half-overlapped 20-second windows from scalp EEG recordings.
Main Results:
- Identified an early-warning indicator in scalp EEG preceding epileptic seizures.
- Observed a relative increase in the power spectral density of low-frequency bands as seizures approached.
- The prediction algorithm achieved 86.67% sensitivity with a false prediction rate of 0.367h⁻¹ and an average prediction time of 45.3 minutes over ~583 hours of EEG data.
Conclusions:
- A novel seizure prediction method utilizing scalp EEG and attractor state analysis shows promise for clinical application.
- This study pioneers the investigation of seizure precursors through attractor-based analysis of macroscopic brain dynamics.
- The proposed spectral feature, focusing on perturbation dynamics, offers a new approach for seizure prediction from EEG.