Artificial Intelligence Can Anticipate an Epileptic Seizure

Early diagnosis of an approachingan epileptic seizure can provide effective relief, and in some cases, prevent the serious consequences of an unexpected attack of the disease. Researchers from the University of Louisiana at Lafayette achieved significant success in creating an algorithm for early diagnosis of seizures by developing a prediction mechanism that allows one to determine the approach of an attack one hour before the exacerbation of the disease.

In testing a new method based onusing artificial intelligence, 22 patients from the Children's Hospital in Boston took part. The experimental results are impressive - seizures were predicted in 99.6% of cases, while the frequency of false alerts reached a negligible amount.

Forecasting Methodology from Scientists from Louisianabased on the use of deep learning and analysis of electroencephalogram (EEG) data. To predict an attack of AI, it compares brain activity in real time with similar historical data on brain activity shortly before an exacerbation of the disease. The technique has one significant drawback. The algorithm needs to be trained individually for each patient, and EEG data are necessary not only at the time of rest or time preceding the crisis, but also at the time of the seizure itself.

Scientists are currently creating compact,wearable devices for EEG. At the same time, it is taken into account that the built-in AI algorithm of such devices should be quite powerful, since information processing in real time will be required. The advent of such a device will greatly facilitate the lives of more than 50 million people with diagnosed epilepsy, allowing them to prepare for an attack, and possibly preventing it by taking the necessary medications.

Source: Engadget