Assessment of Post-Stroke Patients Using Smartphones and Gradient Boosting

Abstract

Stroke is the second leading cause of death and a major cause of disability worldwide. Surviving a stroke will likely result in impairments that will need special care, with only 10%of stroke patients making a full recovery. The taxing amount required to assist in recovery and the scarcity of physiotherapists makes it hard and very expensive for stroke patients to seek treatment. This has caused a shift towards robotizing the process and using rehabilitation robotics for therapy and diagnosis of stroke patients, despite the high cost. This paper demonstrates a cheap diagnosing technique that uses common tools and an open-source machine learning algorithm. By using the built-in inertial measurement unit of smartphones and open-source gradient boosting, it was possible to diagnose a patient’s score when performing 3 tasks of the Fugl-Meyer upper-extremity assessment. The accuracy of the model was evaluated using 5-fold cross-validation and yielded an accuracy of 95.56%.