Abstract
Background: After a stroke, a wide range of deficits can occur with varying onset latencies. As a result, predicting impairment and recovery are enormous challenges in neurorehabilitation. Body function and structure, as well as activities, are assessed using clinical scales. For functional deficits of the upper extremities these include the Fugl-Meyer Assessment for the Upper Extremity (FM-UE), the Chedoke Arm and Hand Activity Inventory (CAHAI) and Barthel Index (BI), administered by clinicians. Although these scales are generally accepted for the evaluation of the motor and functional impairment of the upper-limbs, they are time-consuming, show high inter-rater variability, have low ecological validity, and are vulnerable to biases introduced by compensatory movements and action modifications. For these reasons, alternative methods need to be developed for efficient and objective assessment. Computer-based motion capture and classification tools have the potential to collect and process kinematic data to estimate impairment, function and recovery while overcoming these limitations.Methods: We present a method for estimating clinical scores from movement parameters that are entirely extracted from kinematic data recorded during unsupervised rehabilitation sessions performed with the Rehabilitation Gaming System (RGS). RGS is a rehabilitation technology that uses image-based motion capture, goal-oriented individualised training, virtual reality content delivery, and restricts compensatory trunk movements through feedback. The main protocol considered in this study asks patients to use their upper limbs to intercept spheres that are presented in a 3 dimensional virtual reality display. RGS maps the planar physical arm movements onto matching movements by an avatar presented in a first-person perspective. In this analysis, we performed a multivariate regression using clinical data from 98 stroke patients who completed a total of 191 sessions with RGS.Results: Our multivariate regression model reaches an accuracy of R2 : 0.38, with an error (σ : 12.8), in predicting FM-UE scores. We analyse our model by assessing reliability (r = 0:89 for test-retest), sensitivity to clinical improvements (95% true positive rate) and generalisation to other tasks that involve planar reaching movements (R2 : 0.39). The model achieves comparable accuracy also for the CAHAI (R2 : 0.40) and BI scales (R2 : 0.35).Conclusions: Our results highlight the clinically relevant predictive power of kinematic data collected during unsupervised goal-oriented motor training combined with automated inference techniques and provide new insight into factors underlying recovery and its biomarkers.
Methodological considerations for kinematic analysis of upper limbs in healthy and poststroke adults Part II: a systematic review of motion capture systems and kinematic metrics
