Called FAST AI, it uses artificial intelligence to detect the most common symptoms. It is not yet ready for public use but has already been tested with several patients.
In France and in several countries, cerebrovascular accidents (CVA) are one of the main causes of disability and mortality. Every minute counts for this disease, the diagnosis must be rapid so that the victim can be treated and come out of it without sequelae. An application could precisely help in the detection of the most common symptoms of a stroke, namely facial asymmetry, weakness of the arms and difficulty in speaking, reports the Washington Post .
Called FAST AI, it was designed by researchers and uses machine learning algorithms to detect these symptoms. His name is also an acronym to educate individuals on these warning signs (“facial dropping”, “arm weakness”, “speech difficulty”) with the T for time (“time”) because you must immediately call an ambulance if any of these symptoms occur. “Our goal is very simple. We want to catch the stroke as soon as it happens,” said Radoslav I. Raychev, lead author of the study, which will be presented next week at the American Heart Association’s International Stroke Conference.
An application under development
The application has been tested by neurologists from four major centers specializing in this disease in Bulgaria, with around 270 patients who have had a confirmed diagnosis of an acute stroke. Using video recordings and sensors, FAST AI was able to identify facial asymmetry with over 97% accuracy and arm weakness with over 72% accuracy. The app has also been tested on hundreds of additional patients since the initial research phase, Radoslav I. Raychev said. However, it has not yet been tested for the detection of difficulty of expression.
The ultimate goal of researchers is to go beyond the application, to other devices or platforms. “When someone is video chatting, or talking to ‘a virtual assistant like Alexa’ or driving their self-driving vehicle, we want to be able to have that passive monitoring to capture the patient’s symptoms. when the stroke occurs , ”explained Radoslav I. Raychev. At the moment the application is not ready for public use. No information has been given regarding its availability or to whom it will be marketed.