Patient Specific Seizure Prediction System Using Hilbert Spectrum and Bayesian Networks Classifiers
Figure 5
Sample demonstration of thresholding. Upper row shows the probability of having a seizure in near future. Middle row shows the classification results using 0.25 as threshold. For this preictal recordings, seizure is accepted, predicted 35 minutes prior to the seizure; that is, detection latency is 35 minutes. For the interictal recordings number of false positives is recorded as 5; consecutive alarm in a SPH + SOP window is counted as one. Lower row shows the classification results using 0.5 as threshold and the detection latency now 10 minutes and number of false positives is 1.