Automatic Detection and Classification of High-Frequency Oscillations in Depth-EEG Signals
IEEE Transactions on Bio-Medical Engineering
|January 24, 2017
Summary
A new universal detector accurately identifies and classifies four types of high-frequency oscillations (HFOs) in epilepsy patients. This method distinguishes HFOs from artifacts, improving epilepsy biomarker analysis.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Interictal high-frequency oscillations (HFOs) are crucial biomarkers in epilepsy diagnosis.
- Existing methods struggle with inter-subject variability and artifact rejection.
Purpose of the Study:
- To develop a universal detector for four types of HFOs: Gamma, high-gamma, ripples, and fast-ripples.
- To enhance robustness against variability and improve specificity by rejecting artifacts.
Main Methods:
- Utilizing Gabor atoms and Gabor transform for HFO detection in intracerebral EEG (iEEG) signals.
- Employing energy ratios and event duration for feature extraction, optimized using a multiclass support vector machine (SVM).
Main Results:
- The proposed method demonstrates high sensitivity and a low false discovery rate on simulated and human iEEG data.
- Successful classification and discrimination of all four HFO types were achieved.
Conclusions:
- The developed detector is robust, universal, and capable of distinguishing between different HFO types and artifacts.
- Experimental validation confirms the feasibility of this advanced HFO detection approach.
