Imagine, for example, if an epileptic knew with reasonable certainty that his next seizure would not occur for an hour or a day or a week. That might allow him to run to the market or go out for the evening or plan a short vacation with less concern.
Computers and even dogs have been tested in the effort to do this, but now a group of organizations battling epilepsy is employing “big data” to help. They sponsored an online competition that drew 504 entrants who tried to develop algorithms that would detect and predict epileptic seizures.
Instead of the traditional approach of asking researchers in a handful of labs to tackle the problem, the groups put huge amounts of data online that was recorded from the brains of dogs and people as they had seizures over a number of months. They then challenged anyone interested to use the information to develop detection and prediction models.
“Seizure detection and seizure prediction,” said Walter J. Koroshetz, deputy director of the National Institute of Neurological Disorders and Stroke (NINDS), are “two fundamental problems in the field that are poised to take significant advantage of large data computation algorithms and benefit from the concept of sharing data and generating reproducible results.”
On Friday, the groups announced that the prediction contest was won by a team of scientists and engineers who forecast abnormal electrical activity in the brain with 82 percent accuracy.
They were led by Drew Abbot, an engineer, and Philip Adkins, a mathematician, who work at AiLive, a small company in Sunnyvale, Calif. Also on the team were Quang Tien, Simone Bosshard and Min Chen, scientists at the Center for Advanced Imaging at the University of Queensland in Australia.
The sponsors of the challenge — NINDS, the American Epilepsy Society and the Epilepsy Foundation — noted that six other entrants scored within five-hundredths of a percentage point of the winners.
For the full story, click here.