In a recent presentation at the Society of NeuroInterventional Surgery’s annual meeting, researchers shared exciting news about a new smartphone app that can reliably detect the physical signs of a stroke using the power of machine learning.
The study involved researchers from the UCLA David Geffen School of Medicine and several medical institutions in Bulgaria. They gathered data from 240 stroke patients across four metropolitan stroke centers.
Within 72 hours of the patients experiencing symptoms, the researchers used smartphones to record videos of the patients and assess their arm strength to look for signs of facial asymmetry, arm weakness, and speech changes, which are typical indicators of a stroke.
To evaluate facial asymmetry, the researchers utilized machine learning to analyze 68 specific points on the face. For testing arm weakness, they used data from the smartphone’s built-in 3D accelerometer, gyroscope, and magnetometer.
To detect speech changes, the team employed a sound recognition method that transformed sound waves into images, allowing them to compare normal and slurred speech patterns. When they tested the app’s accuracy using reports from neurologists and brain scan data, they found that it could accurately diagnose strokes in nearly all cases.
The potential impact of this app and the emerging technology of machine learning is enormous. It could help more people identify stroke symptoms as soon as they appear, allowing for quicker and more accurate assessments. This, in turn, is crucial for ensuring that stroke patients receive prompt medical attention and have a better chance of survival and regaining independence.
The researchers and experts involved in this development are hopeful that the app’s deployment will significantly improve stroke care and make a positive difference in the lives of many patients.
- Eurekalert
