Affecting millions of people worldwide, strokes occur when the blood supply to part of the brain is cut off or reduced, depriving brain tissue of oxygen and nutrients. A delay of even a few minutes can cause permanent damage to brain cells.
A team of biomedical engineers from RMIT University developed the AI capabilities behind the software technology and are Computer methods and programs in biomedical sciences.
Guilherme Camargo de Oliveira, a PhD student at RMIT and Sao Paulo State University, led the research under the supervision of team leader Professor Dinesh Kumar.
Early stroke detection is crucial, as prompt treatment can significantly improve recovery outcomes, reduce the risk of long-term disability and potentially save lives.”
Dinesh Kumar, Professor of Engineering, RMIT University
“We developed a simple smartphone tool that allows paramedics to instantly determine if a patient is post-stroke and notify the hospital before the ambulance leaves the patient’s home.”
The smartphone tool, which is 82% accurate at detecting strokes, does not replace comprehensive clinical diagnostic tests for stroke, but it could help identify people who need treatment much earlier.
“Our facial screening tool boasts stroke detection rates comparable to paramedics,” Kumar says.
Stroke is hard to detect
Symptoms of a stroke include confusion, partial or complete loss of motor control, speech impairment, and reduced facial expression.
“Studies have shown that around 13 per cent of strokes are missed in emergency departments and community hospitals, and 65 per cent of patients who do not have a recorded neurological exam experience an undiagnosed stroke,” Kumar said.
“The signs are often very subtle. Plus, when paramedics are working with people of a different race or gender than themselves, especially women and people of color, they’re more likely to miss the signs.”
“This rate is likely to be even higher in smaller, regional centres. With many strokes occurring at home and initial treatment often provided by paramedics under less-than-ideal conditions, there is an urgent need for real-time, easy-to-use diagnostic tools.”
How the technology works
This new AI-driven technology harnesses the power of facial expression recognition to analyze facial symmetry and the movement of specific muscles, called action units, to detect stroke.
First developed in the 1970s, the Facial Action Coding System (FACS) categorizes facial movements by contraction or relaxation of facial muscles, providing a detailed framework for analyzing facial expressions.
“One of the important factors affecting stroke patients is that the facial muscles are usually unilateral, meaning that one side of the face moves differently from the other,” de Oliveira said.
“We have AI tools and image processing tools that can detect if there is a change in smile asymmetry, which is key to detection in our cases.”
The study used video recordings of facial expression tests from 14 post-stroke patients and 11 healthy controls.
Next steps
The team plans to work with healthcare providers to develop the smartphone tool into an app that can detect other neurological conditions that affect facial expressions.
“We want to increase the sensitivity and specificity as much as possible. We are currently working on developing the AI tool with additional data and will consider other diseases as well,” Kumar said.
“Collaboration with healthcare providers will be key to integrating this app into existing emergency response protocols and providing emergency personnel with an effective means of early stroke detection.”
RMIT researchers partnered with Sao Paulo State University in Brazil on the study.
sauce:
Journal References:
Oliveira, G.C., etc (2024). Facial Expressions for Identifying Post-Stroke: A Pilot Study. Computer methods and programs in biomedical sciencesdoi.org/10.1016/j.cmpb.2024.108195