Scientists in Australia have developed a low-cost wearable that utilizes artificial intelligence (AI) system to predict and alert an epileptic patient when a seizure is likely to strike. The predictive algorithm has sensitivity of up to 81.4% and false prediction rate as low as 0.06/hour.
Epileptic seizures are triggered by a disruption in the electrical activity of the brain. There are several different types of epileptic seizures. These seizures strike with little or no warning and almost one third of people living with epilepsy are resistant to treatment that controls these attacks.
“We are on track to develop an affordable, portable and non-surgical device that will give reliable prediction of seizures for people living with treatment-resistant epilepsy,” said Omid Kavehei, University of Sydney in Australia.
Kavehei said that there had been extraordinary developments in AI as well as micro and nano electronics that have enabled such systems. “Just four years ago, you couldn’t process sophisticated AI through small electronic chips. Now it is completely accessible. In five years, the possibilities will be enormous,” he said.
For their study, researchers proposed a generalized, patient-specific, seizure-prediction technique that can alert epilepsy sufferers within 30 minutes of the probability of a seizure.
The team used 3 data sets Europe and the US. The team developed a predictive algorithm with sensitivity of up to 81.4 per cent and false prediction rate as low as 0.06 an hour, based on these data sets.
“While this still leaves some uncertainty, we expect that as our access to seizure data increases, our sensitivity rates will improve,” Kavehei said.
Researchers decided to create a dynamic analytical tool that can read a patient’s electroencephalogram (EEG) data from a wearable cap or other portable device to gather EEG data. For this they utilized deep machine learning.
A low-cost device with wearable technology attached to it could provide a patient a 30-minute warning and likelihood of a seizure. Five to thirty minutes before the seizure onset, an alarm would alert the patient giving them ample time to look for a safe place, reduce stress or start an intervention strategy to control or prevent the seizure. Kavehei said an advantage of their system is that it is not likely to require regulatory approval, and could easily work with existing implanted systems or medical treatments.
The algorithm can produce optimized features for each individual. They do this using ‘convolutional neural network’ – a neural network that is highly attuned to observing changes in brain activity based on EEG readings. A big advantage of the system is that it learns as brain patterns change, requiring minimal feature engineering. Because of this, a faster and more frequent updates on the info can be provided to the patient, giving them maximum benefit from the seizure prediction algorithm.
The study was published in the journal Neural Networks.