The assessment of motor performance is of major importance for correct decision making in neurorehabilitation, especially early after stroke . The need for more reliable and valid tools of motor assessment is common to both clinical and research settings. In addition, everyday clinical practice is usually disturbed by scarcity of specialized human resources, which limits the time dedicated to motor assessment and the number of possible motor tests performed.
In this context, a wearable system capable of automatic assessment of motor function is of increased importance. It could allow clinicians to easily document motor performance and would represent a significant upgrade in the management of future rehabilitation plans and clinical trials .
To achieve this goal, we developed a portable motion capture system based on magnetic, angular rate, and gravity sensors to acquire all the relevant three-dimensional kinematics of upper limb movements . These data were then computed through an automatic decision tree classifier, whose features were inferred from the Functional Ability Score (FAS) of the Wolf Motor Function Test. Features comprised the analysis of synergic movements with the shoulder, smoothness, and motor executions out of the plane of action.
Five stroke patients were enrolled and tested on both sides in five selected tasks. The system was compared against a trained clinician, operating simultaneously and blinded.
Results showed, in terms of performance time, a mean difference between the system and the clinician of 0·17 s (standard deviation = 0·14 s) for all trials performed . The systematic delay of the clinician' assessments were probably due to human error, when indicating task conclusion time. For FAS evaluation, the system and the clinician showed agreement in four out of five patients for two motor tasks evaluated .
These results represent an important step toward a system capable of a precise and independent motor evaluation after stroke.