Smartphone app uses AI to potentially detect symptoms of stroke
Click Here to Manage Email Alerts
Key takeaways:
- An AI-enabled app accurately detected symptoms of stroke using smartphone video, accelerometer and audio data.
- The FAST AI application also differentiated symptoms of stroke and Bell’s palsy.
An artificial intelligence-enabled smartphone application developed to detect symptoms of stroke accurately evaluated facial asymmetry, arm weakness and speech changes in line with stroke onset, a speaker reported.
“Stroke is very treatable if caught early. ... We also know that most patients do not arrive in time to receive brain-saving treatments, mostly due to poor awareness,” Radoslav I. Raychev Jr., MD, neurologist at the David Geffen School of Medicine at UCLA, said during a presentation at the International Stroke Conference. “Despite the well-known FAST paradigm that has been launched for over 10 years, stroke treatments remain low, and more than two-thirds of the young adults don’t know about FAST. So how can we improve stroke recognition?”
Image: Adobe Stock
Raychev stated that approximately 81.6% of Americans own a smartphone and spend about 5 hours and 24 minutes per day using their device.
The researchers therefore developed a multistage FAST AI smartphone app for detection of stroke signs such as recognition of facial asymmetry, arm weakness and speech changes.
The app used video analysis to detect facial asymmetry and unilateral change of facial movement; accelerometer- and gyroscope-derived analysis of arm strength; and audio recording analysis of sound variability patterns.
The researchers conducted a clinical validation trial of the app at five hospitals in Bulgaria, where neurologists used FAST AI on 400 adult patients with confirmed ischemic or hemorrhagic stroke, Bell’s palsy and healthy controls (median age, 69 years; 45.3% women).
The AI app was cross-validated with neurologist clinical exam and impression and brain imaging data.
Overall, 5.5% of the cohort had hemorrhagic stroke, 66% had ischemic stroke, 10.75% had Bell’s palsy and 17.8% were healthy controls.
Raychev reported that the FAST AI app was able to detect true stroke symptoms with a sensitivity of 0.99 (95% CI, 0.95-1) and a specificity of 0.9 (95% CI, 0.8-1).
Moreover, the app detected Bell’s palsy with a sensitivity of 0.78 (95% CI, 0.76-8) and a specificity of 0.7 (95% CI, 0.68-0.72), according to the presentation.
“FAST AI can potentially identify acute stroke features and differentiate from Bell’s palsy with accuracy comparable to neurologists’ clinical impression,” Raychev said during the presentation. “How do we see this technology being used in the future? Perhaps in clinician-aid settings, like the way the study was conducted, or for [emergency medical services] and telemedicine, for instance. The long-term vision is patient-oriented technology for self-assessment of acute and chronic deficits, and perhaps even in ambient monitoring settings, like [internet-connected] smart home, video conferencing and autonomous driving vehicles. ... You are in an autonomous driving car, and before you know it, FAST AI detects your stroke, and the car will take you to a nearby stroke center.”
Collapse
Raychev R, et al. Closing main event: LB 34. Presented at: International Stroke Conference; Feb. 7-9, 2024; Phoenix.
Read more about
Click Here to Manage Email Alerts