Epilepsy Seizures Prediction with Big Data Analytics

11 December, 2014 - 16:58

Seizure forecasting systems hold promise for improving the quality of life for patients with epilepsy. The learning machines group at SICS has found a solution.

Epilepsy afflicts nearly 1% of the world's population, and is characterized by the occurrence of spontaneous seizures. Even with medication patients with epilepsy experience persistent anxiety due to the possibility of a seizure occurring. A solution to this would be seizure forecasting systems with the potential to help patients with epilepsy lead more normal lives.

In order for EEG-based seizure forecasting systems to work effectively, computational algorithms must reliably identify periods of increased probability of seizure occurrence. If these seizure-permissive brain states can be identified, devices designed to warn patients of impeding seizures would be possible. The primary challenge in seizure forecasting is differentiating between the preictal (prior to seizure) and interictal (between seizures, or baseline) states.

American Epilepsy Society Seizure Prediction Challenge is the name of a competition held in 2014 to demonstrate the existence and accurate classification of the preictal brain state in dogs and humans with naturally occurring epilepsy. The SICS team ended up top 10 %, in fierce competition with the leading medical research teams. The team used a support vector classifier and features based on a combination of signal frequency and channel correlation.

Participating in a competition like this enables us to evaluate our machine learning algorithms on real data and at the same time contribute to the solution of a very important problem, says team leader John Ardelius.

The LearningMachines @ SICS team

Theodore Vasiloudis, PhD student
Erik Ylipää, PhD student
Johan Tjellden, PhD student
Olof Görnerup, PhD
John Ardelius, PhD