Researchers have developed an app called FAST AI to detect the most common stroke symptoms — facial asymmetry, arm weakness and changes in speech — using machine learning algorithms. The preliminary findings will be presented next week at the American Heart Association’s International Stroke Conference.
“Our goal is very simple. We want to detect stroke at onset,” said Radoslav I. Raychev, lead author of the study and a vascular neurologist at the University of California at Los Angeles.
The app is not ready for public use, and it is unclear when it will be available or to whom it will be marketed.
Eventually, Raychev said, the goal is to go beyond the app to other devices or platforms.
“When someone’s on video chat, or when they’re talking to” a virtual assistant such as Alexa, “or when they’re driving their autonomous driving vehicle, we want to be able to have that passive surveillance to capture the patient’s symptoms as the stroke happens,” he explained.
Using the app as an educational tool
Neurologists at four major metropolitan stroke centers in Bulgaria tested the app with nearly 270 patients who had a confirmed diagnosis of acute stroke.
Using video recordings and device sensors, the app was able to identify facial asymmetry with more than 97 percent accuracy and arm weakness with more than 72 percent accuracy, according to the abstract, which has been peer-reviewed but has not yet been published in a medical journal.
Since the initial research phase, the app has been tested on several hundred more patients, Raychev said. It is also designed to detect abnormal speech, but that function has not yet been tested on patients, he said.
Experts said a smartphone app that can help detect stroke symptoms is a great idea. But, they said, the app should be used to educate people about the signs and motivate them to call 911, not as a diagnostic tool.
Some experts expressed concern that a false negative result may discourage some people from seeking emergency medical care, for instance, which could have devastating consequences.
“This is an exciting and interesting opportunity to see how we can use AI technology to further information about stroke symptoms,” said Andrew Russman, head of the stroke program and medical director of the Comprehensive Stroke Center at the Cleveland Clinic.
“But by no means should this be considered a methodology for deciding whether to call 911 or deciding whether someone definitively is having stroke symptoms. It’s just one piece of information that sounds interesting right now but doesn’t have an actual application yet.”
Early detection is crucial for strokes
Stroke, which occurs when a blood clot blocks blood flow to the brain or when a vessel bursts within the brain, is the fifth leading cause of death and a leading cause of long-term disability in the United States.
Time is critical in diagnosing the condition and initiating treatment protocols, said Matthew Potts, surgical director of the Comprehensive Stroke Center at Northwestern Memorial Hospital. “When you have blood cut off to the brain, every minute that passes, neurons are dying,” he said.
“Clot-busting” medications can break up clots during the first few hours of an ischemic stroke, which accounts for 87 percent of strokes, preventing or minimizing brain damage.
But these medications such as Alteplase IV r-tPA must be given within 4.5 hours of ischemic stroke onset. In some cases, the clot may be removed during a thrombectomy, a procedure to manually remove it from an artery.
Hemorrhagic stroke may be managed with medications or require surgical intervention.
The prognosis depends on the area of the brain and amount of tissue that is affected, but timing is key, and in some cases, quick action can prevent damage, experts said.
Even with transient ischemic attacks, or TIAs, in which the blood clot is temporary and does not cause permanent injury, they can be a warning sign that the patient is at risk of future strokes. That’s why stroke patients — and those around them — should be aware of the symptoms.
Common warning signs of a stroke
The acronym FAST is used to teach people the most common warning signs of a stroke.
- “F” is for facial drooping: The face may droop on one side, particularly when the person smiles.
- “A” is for arm weakness: When both arms are raised, one arm may start to drift downward.
- “S” is for speech difficulty: The person may be slurring words.
- “T” is for time: If someone is experiencing any of these symptoms, call for an ambulance immediately.
There are other symptoms, too, which an app may have a hard time detecting, that could signal a stroke, Russman said. Stroke patients may experience sudden and severe headaches, confusion, vision loss in one or both eyes, numbness or tingling down one side of their bodies or trouble walking because of dizziness or a loss of balance or coordination.
Greg Albers, director of the Stanford Stroke Center at Stanford Medical Center, said stroke symptoms can be disabling.
“The type of vision you lose from a stroke, for example, is typically half of your visual field,” he said. “If you’re bumping into things because you can’t see, you’re waving a hand in front of your eyes and there’s a big chunk of missing vision, that’s very concerning.”
“Symptoms come on very abruptly,” he added.
Other times, however, symptoms are mild, and patients may not take them as seriously and seek medical attention, which could leave them with permanent deficits, experts said.
Albers acknowledged that determining whether someone is having a stroke can be tricky, which is one reason people should have “a low threshold” for calling 911 when symptoms indicate a stroke.
Even in the emergency room, he said, medical professionals must perform brain scans and other tests before making a definitive diagnosis and starting treatment because giving a “clot-busting” drug to a patient who is having a hemorrhagic stroke — in which an artery in the brain leaks or ruptures — could result in “a disaster.”
“They’re going to bleed more,” he explained.
Experts said the goal of smartphone technology that can detect stroke symptoms should be to get stroke patients into the ER more promptly.
“We want people to err on the side of coming to the hospital, not err on the side of staying home,” said Matthew Fink, chief of neurology at Weill Cornell Medical Center and NewYork-Presbyterian Hospital.