Figure 1.
Flowchart illustrating the system architecture.
Figure 2.
Exemplars of EEG from the three conditions, normal, spike and seizure displayed in both unipolar and bipolar montages.
The right middle figure shows phase reversal of spikes in a bipolar montage.
Figure 3.
The feature extraction of the 6 neighboring channels.
Figure 4.
The magnitude and longevity of a burst.
Table 1.
The top-ranked feature types and associated frequency bands.
Figure 5.
ROC Curve of different classification methods (SVM only, GA+SVM, GA+SVM+Post Spike Matching).
Figure 6.
The accuracy histogram (normal, spike, and seizure) of cross validation with different methods (SVM only, GA+SVM, GA+SVM+Post Spike Matching).
The trend of accuracy in normal EEG decreases slightly, but in spike EEG raises noticeably. In addition, the accuracy of seizure is stable for different feature selection and classifier.
Table 2.
Comparison studies in literature with our approach.