Performance of epileptic single-channel scalp EEG classifications using single wavelet-based features.

Classification of epileptic scalp EEGs are certainly ones of the most crucial tasks in diagnosis of epilepsy. Rather than using multiple quantitative features, a single quantitative feature of single-channel scalp EEG is applied for classifying its corresponding state of the brain, i.e., during seizure activity or non-seizure period. The quantitative features proposed are wavelet-based features obtained from the logarithm of variance of detail and approximation coefficients of single-channel scalp EEG signals. The performance on patient-dependent based epileptic seizure classifications using single wavelet-based features are examined on scalp EEG data of 12 children subjects containing 79 seizures. The 4-fold cross validation is applied to evaluate the performance on patient-dependent based epileptic seizure classifications using single wavelet-based features. From the computational results, it is shown that the wavelet-based features can provide an outstanding performance on patient-dependent based epileptic seizure classification. The average accuracy, sensitivity, and specificity of patient-dependent based epileptic seizure classification are, respectively, 93.24%, 83.34%, and 93.53%.