Aiming to perform an extraction of features which are strongly related to hemiparesis, this work describes a case study involving the efforts of patients in upper-limb rehabilitation, diagnosed with such pathology. Expressed as data (kinematic and dynamic measures), patients' performance were sensed and stored by a single InMotion Arm robotic device for further analysis. It was applied a Knowledge Discovery roadmap over collected data in order to preprocess, transform and perform data mining through machine learning methods. Our efforts culminated in a pattern classification with the abilty to distinguish hemiparetic sides with an accuracy rate of 94%, having 8 features of rehabilitation performance feeding the input. Interpreting the obtained feature structure, it was observed that force-related attributes are more significant to the composition of the extracted pattern

No abstract

Background A detailed sensorimotor evaluation is essential in planning effective, individualized therapy post-stroke. Robotic kinematic assay may offer better accuracy and resolution to understand stroke recovery. Here we investigate the added value of distal wrist measurement to a proximal robotic kinematic assay to improve its correlation with clinical upper extremity measures in chronic stroke. Secondly, we compare linear and nonlinear regression models. Methods Data was sourced from a multicenter randomized controlled trial conducted from 2012 to 2016, investigating the combined effect of robotic therapy and transcranial direct current stimulation (tDCS). 24 kinematic metrics were derived from 4 shoulder-elbow tasks and 35 metrics from 3 wrist and forearm evaluation tasks. A correlation-based feature selection was performed, keeping only features substantially correlated with the target attribute (R > 0.5.) Nonlinear models took the form of a multilayer perceptron neural network: one hidden layer and one linear output. Results Shoulder-elbow metrics showed a significant correlation with the Fugl Meyer Assessment (upper extremity, FMA-UE), with a R = 0.82 (P < 0.001) for the linear model and R = 0.88 (P < 0.001) for the nonlinear model. Similarly, a high correlation was found for wrist kinematics and the FMA-UE (R = 0.91 (P < 0.001) and R = 0.92 (P < 0.001) for the linear and nonlinear model respectively). The combined analysis produced a correlation of R = 0.91 (P < 0.001) for the linear model and R = 0.91 (P < 0.001) for the nonlinear model. Conclusions Distal wrist kinematics were highly correlated to clinical outcomes, warranting future investigation to explore our nonlinear wrist model with acute or subacute stroke populations. Trial registration http://www.clinicaltrials.gov. Actual study start date September 2012. First registered on 15 November 2012. Retrospectively registered. Unique identifiers: NCT01726673 and NCT03562663.

No abstract