Research Article
Siamese Network-Based Feature Transformation for Improved Automated Epileptic Seizure Detection
Table 4
30 high-ranked features selected by ANOVA.
| Rank | Feature/N. S1 | Rank | Feature/N. S | Rank | Feature/N. S |
| 1 | Spectral En/5 | 2 | Spectral En/6 | 3 | Samp En/10 | 4 | Complexity/10 | 5 | SD/02 | 6 | NLE/9 | 7 | Spectral En/9 | 8 | SampEn/2 | 9 | N. ZC4/5 | 10 | NLE/3 | 11 | N. LE3/1 | 12 | Kurtosis/0 | 13 | Mobility/4 | 14 | Skewness/6 | 15 | N. LE/5 | 16 | SampEn/9 | 17 | Perm. En./10 | 18 | Skewness/5 | 19 | Line length/8 | 20 | NLE/7 | 21 | Spectral En/4 | 22 | Shannon En/8 | 23 | Skewness/2 | 24 | Mobility/1 | 25 | SD/1 | 26 | Line length/4 | 27 | N. ZC/3 | 28 | Sample En/5 | 29 | Mobility/7 | 30 | Complexity/2 |
|
|
1Number of sub-band features extracted. 2Sub-band “0” means raw EEG signal. 3Number of local extrema. 4Number of zero crossing.
|