Detecting High Frequency Oscillations for Stereoelectroencephalography in Epilepsy via Hypergraph Learning.
Summary
This study introduces a hypergraph-based detector for identifying high-frequency oscillations in stereoelectroencephalography (SEEG) signals. This method aids in precisely locating epilepsy zones, improving surgical outcomes for patients with refractory focal epilepsy.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Epilepsy surgery success relies on accurate pre-operative localization of epileptogenic zones.
- Stereoelectroencephalography (SEEG) is vital for identifying these zones by recording brain activity.
- Detecting high-frequency oscillations (HFOs) in SEEG signals is crucial for pinpointing seizure onset zones.
Purpose of the Study:
- To develop an automated method for detecting high-frequency oscillations (HFOs) in SEEG signals.
- To formulate HFO detection as a signal segment classification problem.
- To create a hypergraph-based detector for improved epileptogenic zone localization.
Main Methods:
- Formulated HFO detection as a signal segment classification task.
- Developed a novel hypergraph-based detector.
- Evaluated the detector on 4,000 SEEG signal segments from 19 epilepsy patients.
Main Results:
- The hypergraph-based detector successfully localized interictal HFOs.
- The method achieved 90.7% accuracy, 80.9% sensitivity, and 96.9% specificity.
- Outperformed several peer machine learning methods in HFO detection.
Conclusions:
- The proposed hypergraph-based detector is effective for automatic HFO detection in SEEG signals.
- This tool can assist human experts in visually reviewing SEEG data for epilepsy localization.
- Accurate HFO detection using this method can enhance pre-operative planning for epilepsy surgery.
