Detection of Upper Limb Motor Dysfunction for Stroke Patients Using a Wearable Device and a Two-Layer LSTM Model

Upper limb motor dysfunction is prevalent among stroke patients, resulting in impaired movement coordination and reduced dexterity. This paper presents a motor dysfunction detection method for the upper limb based on a wearable device and a neural network model. The device captures the inertial data from patients while they perform finger-nose tests. The data is then preprocessed and fed into a two-layer long short-term memory (LSTM) model for classification. The model uses quaternions and triaxial linear accelerations as features for analysis. The output results include normal and abnormal. We conducted finger-nose tests, built a dataset, and evaluated the accuracy of the model. Results demonstrated that the model achieved an average accuracy of 98.33% on the test set with an F1 of approximately 98.32%. The wearable device has high efficiency and accuracy for assisting medical professionals in hospital diagnosis and patients in early self-examination at home.