A smartphone app has shown its potential in identifying stroke symptoms through a new research study, with its AI capability recognising facial changes in almost all patients.
FAST.AI has been developed to help detect stroke through machine learning algorithms to recognise facial asymmetry, arm weakness and speech changes – all common stroke systems as defined in the FAST acronym.
The app uses a facial video of the patient to examine 68 facial landmark points; sensors that measure arm movement and orientation; and voice recordings detect speech changes. Information from each test was sent to a database server for analysis.
In the preliminary research, results were encouraging, with recognition and treatment being time critical in minimising the long-term impact on the patient. More than 100,000 people have a stroke each year in the UK, and stroke is the main cause of all disability in the US.
The analysis found that FAST.AI – created by Neuronics Medical – accurately detected stroke-associated facial asymmetry in nearly 100 per cent of patients, and accurately detected arm weakness in more than two-thirds of cases.
While the slurred speech module remains to be fully validated and tested, preliminary analyses confirmed that it may be able to reliably detect slurred speech, according to the researchers, who hailed the app as being as accurate at diagnosing stroke as a neurologist.
“Many stroke patients don’t make it to the hospital in time for clot-busting treatment, which is one reason why it is vital to recognise stroke symptoms and call 9-1-1 right away,” said study author Dr Radoslav I. Raychev, a clinical professor of neurology and a vascular neurologist at the University of California, Los Angeles.
“These early results confirm the app reliably identified acute stroke symptoms as accurately as a neurologist, and they will help to improve the app’s accuracy in detecting signs and symptoms of stroke.”
Researchers validated FAST.AI’s performance by testing nearly 270 patients with a diagnosis of acute stroke (41 per cent women; average age of 71 years) within 72 hours of hospital admission at four major metropolitan stroke centres in Bulgaria (St. Anna University Hospital in Sofia; University Hospital Haskovo in Haskovo; University Hospital Pulmed in Plovdiv; and University Hospital ‘Prof. Dr. Stoyan Kirkovich’ in Stara Zagora) from July 2021 to July 2022.
Neurologists who examined the patients tested the app then compared the FAST.AI results with their clinical impressions.
Previous research has found that stroke patients who are treated within 90 minutes of their first symptoms were almost three times more likely to recover with little or no disability, in comparison to those who received treatment more than 90 minutes after symptoms begin.
“This abstract describes a validated approach for an easy assessment of signs of a stroke and the prompt to seek care. The app may help individuals assess the signs of a stroke without the need to recall the warning signs, ” said Daniel T. Lackland, American Stroke Association volunteer expert and EPI and Stroke Council member.
The preliminary research will be presented at the American Stroke Association’s International Stroke Conference 2023, to be held next week.
