Automatic Prediction of Upper Limb Motor Recovery in Stroke Patients Undergoing Robotic Therapy

The recovery of upper extremity motor functions is crucial after a stroke. Robotic technologies have been employed for over two decades as part of rehabilitation strategies. However, the varying positive responses to robotic treatments emphasize the complexity of each clinical case. Addressing this diversity is a challenge in personalizing rehabilitation paths, underscoring the need for tailored recovery strategies. This study aims to develop a predictive method based on machine learning algorithms to anticipate upper limb motor recovery following therapy with the ALEx exoskeleton. The proposed approach automatically identifies patients more likely to benefit from robotic therapy. Through extensive feature extraction and validation (based on EMG and kinematics features of the upper limbs), we identified the contraction median of the brachioradialis muscle as the most predictive variable, resulting in an F1-Score of 1.0 in the majority of the classifiers. Our findings confirm that integrating EMG data during assessment can provide valuable information, enhancing overall evaluation and predicting motor recovery in stroke survivors undergoing robotic therapy.