Novel AI System Can Predict Seizures with 99.6% Accuracy Up to an Hour Before They Occur

Epilepsy is a debilitating disorder that causes havoc for those who suffer from it. Luckily technology has come a long way to helping with the condition creating a night watch to help with the attacks.

RELATED: EPILEPSY WARNING BRACELET FEATURES TECHNOLOGY TO SAVE LIVE

99.6% accuracy

Now, there is a new innovation that offers even more accuracy when seeking to predict seizures at any time of the day, as IEEE Spectrum reports. A novel artificial intelligence system has been found to detect epileptic seizures with 99.6% accuracy up to an hour before they occur.

The new system, developed by Hisham Daoud and Magdy Bayoumi of the University of Louisiana at Lafayette, combines EEG (electroencephalogram) technology and predictive modeling. Previous versions used both those techniques but in a two-step process.

By bringing together extraction and classification processes into a single automated system, Daoud and Bayoumi were able to get earlier and more accurate seizure predictions. This is of great help to epileptic patients who can in many cases control seizures through medications.

The researchers tested their system on 22 patients at the Boston Children’s Hospital. They found a 99.6% accuracy rate with 0.004 false alarms per hour.

Trained on individual patients

The system, however, does need to be trained on each individual patient.

“In order to achieve this high accuracy with early prediction time, we need to train the model on each patient,” said Daoud to IEEE Spectrum.“This recording could be [done] off-clinic, through commercially available EEG wearable electrodes.”

Now, Daoud said his team is working on a customized computer chip to process the algorithms.

“We are currently working on the design of an efficient hardware [device] that deploys this algorithm, considering many issues like system size, power consumption, and latency to be suitable for practical application in a comfortable way to the patient,” he added.

The system is described in a study published in IEEE Transactions on Biomedical Circuits and Systems.