HMN 2025: How A brand new instrument may also help predict brain-damaging seizures and liberate steady EEG machines for high-risk sufferers

In hospital intensive care models, neurologists typically use a easy scorecard to shortly consider a critically unwell affected person’s probability of getting a brain-damaging seizure to allow them to stop it. The scorecard saves lives, lowers prices and helps docs make much more environment friendly use of vital, costly medical gear.

But it wasn’t at all times that manner. Until about 2019, hospitals had no option to shortly monitor a affected person’s digital mind exercise and precisely predict whether or not a seizure was lurking. Often, a affected person wanted steady monitoring for twenty-four to 48 hours on a steady EEG (cEEG) machine to know for certain.

Those machines are so costly that almost all hospitals solely have just a few, which suggests solely a handful of sufferers at a time may gain advantage from them. Often, the monitor can be used on a affected person at low-seizure danger whereas extra high-risk sufferers must wait.

In 2016, a staff at MIT led by present Duke pc scientist Cynthia Rudin with then Ph.D pupil Berk Ustun and neurologist colleagues Aaron Struck and Brandon Westover started work on a brand new instrument for docs to make use of. The staff crunched information from 1000’s of sufferers utilizing artificial intelligence to create a greater option to shortly consider mind damage sufferers.

The ensuing scorecard—known as “2HELPS2B,” assigns factors to sufferers primarily based on the patterns discovered of their cEEG information. It shortly estimates a likelihood {that a} affected person would have a seizure in a spread from 5 to 95%.

“There was no model to do that earlier than,” stated Rudin. She is a Duke professor of engineering and pc science who beforehand labored at MIT; it was throughout her time at MIT that she started work on this venture.

Rudin’s staff used a machine {learning} algorithm that evaluated about 70 various factors derived from information from greater than 5,000 sufferers, a sophisticated course of that took a number of years to excellent.

Now, a mind monitoring course of that when took as much as two days is completed as quick as 60 minutes, stated Westover, the Harvard neurologist.

The cEEG monitor derives information on a affected person’s mind exercise {that a} specialist evaluates utilizing the scorecard. The scorecard assigns factors to components resembling whether or not a sure electrical sample is current, or whether or not the affected person has had a previous seizure. A rating of zero means a affected person could be very unlikely to have a seizure; a rating of two suggests a probability of a seizure is 27%; a rating of 6 or extra is the utmost—which helps docs allocate assets successfully.

“It’s a really, very useful gizmo,” Westover stated. “In critical care, the one that wants mind monitoring is often confused or comatose, so you don’t have any concept whether or not they’re having seizures now. Seizures can injury the mind, and if they’ve a brain damage already, a seizure will make it worse.”

A March 2025 study by UNC Chapel Hill researchers discovered that it has produced important advantages. Their study, published within the journal Neurology Clinical Practice, discovered that the scorecard may also help hospitals shortly decide which sufferers had been low danger and transfer their assets to different, extra essential ones. Doing so made extra efficient use of the displays with out compromising the power to precisely predict seizures, the review concluded.

One of the scorecard’s finest options is its simplicity; it would not require any sophisticated arithmetic, only a fast tally of small numbers, solely 1s and 2s. That’s one purpose it is now extensively used, Westover stated.

“It’s ubiquitous now,” he stated. “My estimate can be that at the very least 90% of hospitals, if they’ve any ICU EEG companies in any respect, will use this scorecard.”

It is especially satisfying to Rudin that the instrument, publicly funded by taxpayers via the federal authorities, is accessible on the Internet for anybody to make use of.

“You cannot do that should you work at a private company,” Rudin stated. “The solely manner I could make my models and algorithms public is as a result of I’m funded in a manner that lets me do one thing for the general public good. My objective is to make instruments that everybody can use.”