Seizure prediction systems and methods include measuring impedance within a brain of a patient to determine whether the brain is in a state indicative of a possibility of an onset of a seizure. In some embodiments, the measured impedance is compared to a predetermined threshold in order to determine whether the brain is in a state indicative of a possibility of a seizure. In other embodiments, a trend of the impedance measurements is correlated to a template. In other embodiments, a frequency component of a waveform of the impedance measurement amplitudes over time is correlated to frequency components of a template waveform. Upon detecting a state in which a seizure is likely to occur, a seizure indicator may be generated, which, in some embodiments, may be used to activate therapy delivery to the patient or, in other embodiments, activate an alarm.
