A robotic test of proprioception within the hemiparetic arm post-stroke

BACKGROUND: Proprioception plays important roles in planning and control of limb posture and movement. The impact of proprioceptive deficits on motor function post-stroke has been difficult to elucidate due to limitations in current tests of arm proprioception. Common clinical tests only provide ordinal assessment of proprioceptive integrity (eg. intact, impaired or absent). We introduce a standardized, quantitative method for evaluating proprioception within the arm on a continuous, ratio scale. We demonstrate the approach, which is based on signal detection theory of sensory psychophysics, in two tasks used to characterize motor function after stroke. METHODS: Hemiparetic stroke survivors and neurologically intact participants attempted to detect displacement- or force-perturbations robotically applied to their arm in a two-interval, two-alternative forced-choice test. A logistic psychometric function parameterized detection of limb perturbations. The shape of this function is determined by two parameters: one corresponds to a signal detection threshold and the other to variability of responses about that threshold. These two parameters define a space in which proprioceptive sensation post-stroke can be compared to that of neurologically-intact people. We used an auditory tone discrimination task to control for potential comprehension, attention and memory deficits. RESULTS: All but one stroke survivor demonstrated competence in performing two-alternative discrimination in the auditory training test. For the remaining stroke survivors, those with clinically identified proprioceptive deficits in the hemiparetic arm or hand had higher detection thresholds and exhibited greater response variability than individuals without proprioceptive deficits. We then identified a normative parameter space determined by the threshold and response variability data collected from neurologically intact participants. By plotting displacement detection performance within this normative space, stroke survivors with and without intact proprioception could be discriminated on a continuous scale that was sensitive to small performance variations, e.g. practice effects across days. CONCLUSIONS: The proposed method uses robotic perturbations similar to those used in ongoing studies of motor function post-stroke. The approach is sensitive to small changes in the proprioceptive detection of hand motions. We expect this new robotic assessment will empower future studies to characterize how proprioceptive deficits compromise limb posture and movement control in stroke survivors.