Integrated Automatic Detection, Classification and Imaging of High Frequency Oscillations With
Baotian Zhao1, Wenhan Hu1,2,3,4, Chao Zhang1
1Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, Beijing, China.
Frontiers in Neuroscience
|June 26, 2020
Summary
This study introduces an automated pipeline for detecting and classifying high frequency oscillations (HFOs) using SEEG data. The developed system accurately identifies the epileptogenic zone (EZ) in epilepsy patients, aiding surgical planning.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Computational Neuroscience
Background:
- High frequency oscillations (HFOs) are increasingly recognized as biomarkers for the epileptogenic zone (EZ) in focal epilepsy.
- Accurate delineation of the EZ is crucial for successful epilepsy surgery.
- Bridging the gap between quantitative HFO analysis and clinical application requires robust automated tools.
Purpose of the Study:
- To develop and validate an integrated automatic pipeline for HFO detection, classification, and imaging using stereoelectroencephalography (SEEG).
- To improve the precision of EZ identification in presurgical epilepsy evaluation.
- To facilitate the clinical translation of HFO analysis.
Main Methods:
- A pipeline involving channel selection, HFO detection, and classification using four convolutional neural network (CNN) classifiers was developed.
- The system was evaluated on a simulated dataset and validated in a cohort of 20 epilepsy patients.
- HFO classification performance was assessed by comparing results with seizure onset zone (SOZ) channels and calculating receiver operating characteristic (ROC) curves.
Main Results:
- The initial HFO detector performed well on simulated data.
- CNN classifiers achieved over 95% accuracy on validation data.
- HFO classification significantly enhanced EZ localization accuracy, with improved area under the curve (AUC) values in patient cohorts.
Conclusions:
- The automated HFO analysis pipeline provides robust results for individual EZ identification.
- This technology has the potential to significantly advance presurgical evaluation and surgical planning for epilepsy patients.
- The findings support the integration of HFO analysis into routine clinical practice.
