Detection of Focal and Non-Focal Electroencephalogram Signals Using Fast Walsh-Hadamard Transform and Artificial...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Neurology

Background:

• Accurate localization of the epileptogenic zone (EZ) is crucial for neurosurgery in epilepsy treatment.
• Conventional diagnosis relies on time-consuming and error-prone visual inspection of long-term electroencephalogram (EEG) signals.
• Distinguishing between non-focal class (NFC) and focal class (FC) EEG signals is a key challenge.

Purpose of the Study:

• To develop an automated system for classifying focal class (FC) and non-focal class (NFC) EEG signals.
• To improve the efficiency and accuracy of epileptogenic zone (EZ) localization.
• To reduce the diagnostic burden on neurologists.

Main Methods:

• Utilized Fast Walsh-Hadamard Transform (FWHT) for frequency-domain analysis and decomposition of EEG signals into Hadamard coefficients.
• Extracted five nonlinear features (approximate entropy, log-energy entropy, fuzzy entropy, sample entropy, permutation entropy) from the Hadamard coefficients.
• Employed an artificial neural network (ANN) classifier with 10-fold cross-validation for classifying NFC and FC EEG signals.

Main Results:

• Achieved high classification performance on two public datasets (University of Bonn and Bern-Barcelona).
• Maximum sensitivity of 99.70%, accuracy of 99.50%, and specificity of 99.30% were obtained using the Bern-Barcelona dataset.
• Demonstrated superior classification performance compared to existing techniques on both datasets.

Conclusions:

• The proposed automated method using FWHT, entropy features, and ANN effectively discriminates between NFC and FC EEG signals.
• This approach offers a significant advancement in the automated diagnosis of epilepsy and localization of the EZ.
• The method shows potential for enhancing diagnostic accuracy and efficiency in clinical practice.