Performance of dynamic features in classifying scalp epileptic interictal and normal EEG
Forrest Sheng Bao1, Ya-Liang Li, Jue-Ming Gao
1Department of Computer Science, Texas Tech University, Texas, USA. forrest.bao@gmail.com
Summary
This study explores dynamic features for epilepsy diagnosis using electroencephalography (EEG). Dynamic features show promise for classifying epileptic EEG, potentially improving computer-aided diagnosis systems.
Area of Science:
- Neurology
- Biomedical Engineering
- Signal Processing
Background:
- Epilepsy affects over 50 million globally.
- Computer-aided diagnosis systems for epilepsy often use electroencephalography (EEG) classification.
- Current systems utilize classical spectral features and novel dynamic features.
Purpose of the Study:
- To evaluate the suitability of dynamic features for classifying interictal EEG in epilepsy diagnosis.
- To compare the performance of dynamic features against classical spectral features.
- To identify and rank the most discriminative dynamic features.
Main Methods:
- Developed a computer-aided epilepsy diagnostic system.
- Classified scalp EEG data using both classical spectral and dynamic features.
- Employed F-score to rank the discriminative power of dynamic features.
Main Results:
- Dynamic features demonstrate suitability for epilepsy classification from EEG.
- Comparison indicates potential advantages of dynamic features over classical spectral features.
- Feature ranking identified specific dynamic features with higher discriminative value.
Conclusions:
- Dynamic features are a viable and potentially superior alternative to classical spectral features for epilepsy diagnosis.
- Further research into refining dynamic features could enhance classification accuracy.
- This preliminary study supports the use of dynamic EEG features in epilepsy detection systems.
