ABSTRACT NUMBER: ESOC2026OT48 NEUROLOGICAL OBSERVATIONS ASSISTANT IN HEALTHCARE: WEARABLE SENSORS FOR STROKE EVALUATIONS

Abstract Background and aims Wearable devices are increasingly used for health monitoring, offering novel opportunities to gain more detailed insights into motor behaviour in neurological disorders such as stroke. Given the limitations of observation-based clinical evaluations, including subjectivity and inter-rater variability, wearable sensors embedded with Inertial Measurement Units (IMUs) may assist in providing more precise and objective characterisations of motor impairment. Methods These approaches could streamline stroke evaluations and derive digital biomarkers from wearable sensors that may improve prognostic accuracy and facilitate remote evaluations. This can also strengthen our understanding of post-stroke motor outcomes and support the development of tailored rehabilitation pathways. Results This study uses a 17-sensor full-body motion capture system (Xsens MTw Awinda), to obtain three-dimensional skeletal reconstructions and detailed kinematic data including joint angles, acceleration, and angular velocity. Commerically available smartwatches will also be used to evaluate simpler, scalable measurement approaches. Stroke survivors at Charing Cross Hospital (UK) will be recruited. Participants will wear both the Xsens system and four smartwatches (one on each wrist and ankle) while completing the Fugl-Meyer Assessment of Upper Extremity (FMA-UE) and Action Research Arm Test (ARAT). Conclusions Manual feature extraction methods and decision tree models, as well as deep learning models, will be used to predict clinical scores solely from sensor-derived data at baseline, as well as longitudinal outcomes (at 9-month follow-up). The most informative assessment tasks, defined as those most predictive of total scores, will also be identified. Conflict of interest