Automatic Epileptic Seizure Detection Using PSO-Based Feature Selection and Multilevel Spectral Analysis for EEG Signals

Table 7

Comparisons with state-of-the-art seizure detection methods in the CHB-MIT database.

Authors	Methodology	No. of patients-channels	(%)-(%)-(%)
Kiranyaz et al. [57]	Time domain, frequency domain, TF, nonlinear features	21-18	89.0-94.7-NR
Zabihi et al. [58]	Seven features from intersection sequence, LDA	23-23	89.1-94.8-94.6
Li et al. [48]	CE-stSENet	21-5	92.41-96.05-95.96
Chandel et al. [36]	Orthonormal triadic wavelet-based features, -NN	18-22	98.36-99.62-99.45
Tian et al. [59]	Multiview features, CNN	24-23	96.66-99.14-98.33
Peng et al. [49]	Stein kernel-based SR	20-5	97.85-98.57-98.21
Chen et al. [30]	Discrete wavelet transform, seven wavelet and two statistical features, SVM	18-22	91.71-92.89-92.3
Jiang et al. [31]	Symplectic geometry eigenvalues, SVM	22-5	97.17-99.72-99.62
Hassan et al. [38]	EMD, multilayer perceptron neural network (MLPNN) classifier	23-22	NR-NR-99.57
This work	46 features from time and frequency domain and information theory, PSO-based feature selection, SVM	20-5	96.79-98.64-98.14