More than 500 teams of data scientists worldwide participated recently in a competition that examined hundreds of hours of recorded brain electrical activities in two people and five dogs prior to and during epileptic seizures.
The results suggest that seizure forecasting is possible in both canine and human epileptic models.
The results of this study, “Crowdsourcing reproducible seizure forecasting in human and canine epilepsy,” were published in the neurology journal Brain.
Affecting about 50 million people worldwide, epilepsy is a neurological brain disease identified by recurrent episodes of seizures.
The unusual electrical activities occurring in the brain during a seizure is seriously problematic to many patients for reasons that include the potential to become unconscious; which turns ordinary activities, like driving a car or holding a baby, into dangerous situations. The ability to predict when seizures are about to occur could life changing.
“If an algorithm could detect subtle changes in the electrical activity of a person’s brain before a seizure occurs, people with epilepsy could take medications only when needed and possibly reclaim the daily activities many of us take for granted,” said Mayo Clinic data scientist Ben Brinkmann in a press release.
Brickmann was the lead author of the study conducted by Mayo Clinic, the University of Pennsylvania and the University of Minnesota.
Developing accurate predicting algorithms of epileptic seizures is challenging because significant data related to electrical brain impulses prior or during seizures are required. Unfortunately, data collected through implanted electroencephalography (EEG) provide a limited window to seizure under normal occurring conditions because of many factors: limitation of EEG recording, alteration of brain function by administrated medications, confidentiality issues for the release of data, and cost and intellectual property problems.
“In the hope of winning up to $15,000 in prize money and bragging rights in data science circles, hundreds of algorithm developers, most with little or no experience with epilepsy or EEG, worked countless hours to build, test and rebuild algorithms for seizure forecasting,” Brinkmann said.
In the study, data scientists evaluated algorithms of almost 350 seizures recorded over 1,500 days. Results of the four-month competition revealed that more than half of the crowdsourced algorithms achieved better results when compared to the random predictions. Furthermore, when tested on epileptic canine data obtained with the Seizure Advisory System, the algorithms predicted over 70% of seizures.
“These results support our effort to develop the next generation of epilepsy devices designed to continuously monitor brain activity, forecast and prevent seizures,” said Dr. Greg Worrell, a Mayo Clinic neurologist. “These datasets and source code also serve as a benchmark, allowing new algorithms to be compared to each other and to the algorithms developed in this competition.”
In 2015, the Mayo team received a five-year federal grant from the National Institutes of Health to continue research on forecasted seizures. The Mayo team will further evaluate the safety and efficacy of seizure forecasting first on canines then on people.
“The dog EEG data has great bidirectional benefits, both as a comparative model for human epilepsy and also to directly benefit dogs with naturally occurring epilepsy,” said Dr. Ned Patterson, a University of Minnesota veterinarian.
