Fig. 6. Methods used in the analyses.

a Example of epileptiform discharges (termed ‘spikes’ throughout the text) detected using a patient-specific template matching algorithm. b Two examples of 1 s length iEEG signals (top) and the corresponding autocorrelation function. c An example iEEG signal during a seizure (top) and the corresponding ACFW (bottom). The red region shows the seizure. The ACFW was normalized by the average of a 1 min period prior to the seizure onset (shaded region). d The example system used throughout the text has equilibria given by a manifold (gray surface). The parameters k and r represent two driving parameters that control the state of the system, and how susceptible it is to a state transition. e The synchronization index (SI) is computed by adding complex phases (dashed vectors) and computing the mean resultant length (bold vectors). Shown are two examples, one with a high SI (dark blue) and one with a low SI (green). f Two thresholds were optimized to separate seizure probability into low, medium, and high risk states by maximizing the product of criteria C1 and C2. Regions where criteria C3 and C4 were not achieved were set to zero. In this example, the red dot shows the optimal thresholds where C1 and C2 were maximized.