AI

A man having a headache. / Photo by: Getty Images

 

About 65 million people in the world are suffering from epilepsy and one-third of them are resistant to treatments. Seizures from an epilepsy episode often occur without warning, leaving patients susceptible to brain damage. Now, a research team at the University of Sydney developed an algorithm to predict seizures without surgical intervention.

“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 Dr. Omid Kavehei from the Faculty of Engineering and IT at the University of Sydney Nano Institute.

The algorithm has been designed as a generalized, patient-specific, seizures-prediction system to notify patients within several minutes before a potential seizure strikes. The AI system in which the algorithm was included also comprised of nanoelectronics.

In the study, the researchers used three datasets obtained from Europe and the United States. They utilized the datasets to develop the algorithm of the AI system, apply deep learning, and employ data-mining techniques to build a dynamic analytical tool. The tool can read results from electroencephalogram or other similar devices capable of gathering EEG data.

They installed the predictive software in a portable device that can be attached to a wearable device that reads EEG. Once worn by the patient, the system learns the changes in the patient’s brain patterns to predict future seizures more effectively.  

As of now, the sensitivity rate is 81.4 percent in predicting future seizures while its false prediction rate is as low as 0.06 per hour. A confirmed seizure would trigger an alarm between 5 and 30 minutes before the attack occurs, giving the patient enough time to reduce stress or take steps to control the seizure.

The predictive software can relieve the patient’s anxiety and fear over the unknown arrival of an attack. It can also be a helpful tool to warn parents, friends, and workmates associated with the patient.