An Israeli startup has developed a way to predict epileptic seizures an hour before they happen, helping sufferers to stop any symptoms by swiftly taking medication.

Epilepsy is a chronic neurological disorder causing frequent seizures that temporarily disrupt brain function. There are many different types of seizures, each one with a range of symptoms, including uncontrollable jerking (commonly known as a fit) and loss of awareness of one’s immediate surroundings.

The chronic, non-communicable disease can be found in people of all ages, and, according to the World Health Organization, it affects some 50 million people worldwide.

Be’er Sheva-headquartered NeuroHelp was founded three years ago, when its CTO Oren Shriki, a professor in the Department of Cognitive and Brain Sciences at Ben-Gurion University of the Negev, partnered with CEO Nadav Karni to figure out a way to potentially detect even the smallest seizure in patients.

Shriki began with the hypothesis that everyone – even those without epilepsy – operates on the verge of a seizure, and sought to find a way that could detect what he calls the “transition point,” after which a seizure is inevitable.

“If you cross this transition point, you may end up having a hallucination, which is some sort of uncontrolled activity in the brain, or epileptic seizures, which are kind of similar but involve more synchronous [simultaneous] activity of the neural networks,” he tells NoCamels.

Neurohelp created AI-led software that could help detect seizures using electroencephalography (EEG) tests.

An EEG measures electrical activity in the brain with small, metal discs that attach to the scalp. The software the company developed can detect certain measurable fluctuations in brain activity that indicate that a seizure is taking place.

Once the algorithm was able to detect a seizure through monitoring brainwaves, NeuroHelp started work on a way to predict seizures about an hour before they happened. To do so, they took all the brain activity and ran it through the algorithm to identify what brain activity to expect before an epileptic seizure.

The platform currently uses an EEG scanner that can record and monitor brian activity at night via a mobile app, but is working on a “more compact” device that can be worn during the day.

Every few seconds, the NeuroHelp app gets a rating from the ongoing EEG test that tells the AI how close to passing the transition point a patient is. This shows a person if a seizure could soon occur.

Once a seizure is predicted, Shriki first recommends not to do any kind of “risky” activities, such as swimming or attending a public event. The second thing he recommends is to take emergency medication meant to stop seizures.

Because seizures come in clusters, the medication is usually taken to stop the first one and then prevent any others that may follow. However, Shriki explains that taking the medication before the first seizure could prevent all subsequent ones from happening at all.

“The assumption is that if you predict a seizure in advance, even ten to five minutes before, it’s enough to take this kind of medication,” Shriki says.

“We believe that if you take [the medication] even within a timeframe of one hour or 30 minutes before seizure, it will still work.”

He also emphasizes the importance of not constantly taking emergency medication because the brain can adapt to the medication and stop it from its intended purpose.

“These are medications that you should use only when those are concrete alert,” Shriki says.

Furthermore, there are many people for whom medication to prevent seizures is ineffective.

“About a third of patients do not react to drugs. They are considered treatment resistant and they keep suffering from seizures,” says Karni.

Detection of an imminent seizure, Shriki says, is not as straightforward as one might think. Some signs are very clear, but many are much more subtle. And in many instances, movements like eye blinks can seem like a seizure even when they are not.

Using the EEG readings, Shriki and his team also worked on creating an algorithm that learnt what fake seizures look like, thereby ruling out any false alarms.

“Initially, the algorithms had 10 to 20 false alarms per day. Currently there are around one 0.2 false alarms per day, which means one to two false alarms per week,” says Shriki.

NeuroHelp is funded by venture capitalists in the healthcare industry as well as receiving funding from Oazis, Ben-Gurion University’s accelerator program.

Shriki is now developing a system called Neurofeedback, which can provide a patient with information about their brain and train it to stay away from the point of transition.

NeuroHelp is now holding a further set of clinical trials for the seizure detector, and Karni believes that by next year it will be available to consumers for home use.

Based on the assumption that everyone can be on the verge of a seizure, whether they have been diagnosed or not, Shriki hopes the system will be of widespread use.

“We believe we can change the world,” he says. “There is a long way to go, but we’re working on it.”


Source:, Shiri Epstein