Developing algorithms to predict seizures

Epilepsy is a neurological disorder characterized by unpredictable, recurrent seizures that can pose a risk to a patient’s safety. One research team created a machine-learning algorithm that was "very good" at predicting seizures: It predicted all seizures in their dataset at least two minutes before their onset with 3.9 false positives per hour. The team then transferred this prediction algorithm to custom hardware that runs on patient data to predict seizures in real time. If a seizure were about to occur, the hardware would then communicate back to electrodes implanted in the brain to apply electrical neurostimulation, which can actually stop the seizure before it occurs.

Provided by William Marsh Rice University

Runtime: 2:24

Get Science360's video of the day in your mailbox each weekday.

Sign up now!