Seizure Forecasting in Epilepsy: From Computation to Clinical Practice

The desirability of accurate seizure forecasts is well understood by most individuals living with epilepsy. Seizure forecasting could be invaluable in a closed-loop neuromodulation approach to suppress seizures before they develop. With inexpensive and widely available computational resources, efforts toward seizure prediction using advanced computational methods on scalp or intracranially recorded electroencephalogram (EEG) data experienced fast growth in the late 1990s. The development of neuromodulation devices for epilepsy and other neurologic disorders paved the way for ambulatory intracranial EEG monitoring. Implanted devices capable of recording intracranial EEG in humans have enabled rigorous confirmation of historical observations that daily and multi-day patterns exist in seizure probabilities, and that these patterns also extend to interictal epileptiform activity. New investigational implanted neuromodulation devices have promising capabilities for clinical application of seizure forecasting, and investigation into noninvasively measured biosignals may make seizure forecasting accessible to patients without invasive implants.