Extracting epileptic features in EEGs using a dual-tree complex wavelet transform coupled with a classification...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Epileptic seizure detection from electroencephalogram (EEG) signals is traditionally a manual, time-consuming, and error-prone process.
• Expert visual inspection of EEG data requires extensive training and is subjective.
• There is a need for automated, accurate, and efficient methods for seizure detection.

Purpose of the Study:

• To develop and evaluate a novel automated method for classifying epileptic seizures from EEG signals.
• To improve the accuracy and efficiency of seizure detection compared to traditional methods.
• To provide a tool that can assist neurologists in diagnosing neurological disorders.

Main Methods:

• Utilized a dual-tree complex wavelet transform (DT-CWT) to decompose EEG sub-segments into detailed and approximation coefficients.
• Applied fast Fourier transform (FFT) to DT-CWT coefficients to identify relevant frequency bands.
• Employed a least square support vector machine (LS-SVM) classifier for seizure classification based on extracted features.

Main Results:

• Achieved high average accuracies of 97.7% on the Bonn University database and 96.8% on the Bern University database.
• Demonstrated that the combined DT-CWT and FFT feature extraction effectively captures discriminative information from brain signals.
• Outperformed k-means and Naïve Bayes classifiers, as well as previous methods for epileptic seizure detection.

Conclusions:

• The proposed DT-CWT and FFT based feature extraction method is highly effective for classifying epileptic seizures from EEG signals.
• This automated technique offers a reliable alternative to manual inspection, improving diagnostic efficiency.
• The method has potential applications in aiding clinical diagnosis and developing early seizure warning systems.