Many epilepsy patients respond to medication for preventing the onset of seizures, but for individuals whose seizures remain uncontrolled, everyday activities, like driving a car, can become suddenly dangerous.
For those patients, neurostimulation therapy is an option. There, implanted electrodes first monitor the brain for the abnormal electrical activity that is the signature of an impending seizure, and then deliver electrical pulses of its own to disrupt that activity before it impacts the patient’s motor control.
This potentially life-changing therapy is not without its hurdles, however. Detection of an impending seizure must be both rapid and reliable if patients are to be confident that an unexpected seizure won’t put them in jeopardy. Conversely, being too sensitive results in unnecessary neural stimulation.
“Detecting seizures accurately and as early as possible is the key to building effective devices to treat epilepsy,” says co-primary investigator Brian Litt, a professor of neurology and bioengineering and director of the new Penn Center for Neuroengineering and Therapeutics. “Leveraging the tremendous talent of scientists and the machine learning community worldwide has tremendous potential to help researchers and our patients. It’s an exciting new model for collaboration.”
In addition to Litt, the Penn team is co-led by Zack Ives, an associate professor in the Department of Computer and Information Science, and includes Joost Wagenaar from the Neurology and Bioengineering departments, and Charles Vite, an associate professor of veterinary neurology at the School of Veterinary Medicine.
The contestants will analyze electrical activity datasets previously recorded from electrodes implanted in the brains of 12 epilepsy patients. Penn Vet is involved in the contest because four of those patients are dogs; dogs naturally suffer from epilepsy much like humans, and are also candidates to receive implanted neurostimulation therapy devices for treatment of their seizures. The datasets of the eight human patients were recorded during evaluation for epilepsy surgery.
The goal of the competition is develop a computer algorithm that can identify the brain activity changes within those datasets that led to actual seizures; the algorithm that detects these changes the fastest with the fewest false alarms wins.
The contest runs until Aug. 19 and offers $8,000 in prizes funded by the American Epilepsy Society and the National Institute of Neurological Disorders and Stroke.
After the contest, the datasets will be made freely available to researchers worldwide.
Originally published on June 5, 2014