UC Irvine

Stroke is one of the leading causes of disability worldwide because it often creates both sensory and motor deficits, which impact the ability to complete activities of daily living such as walking. Rehabilitation therapy can promote recovery of movement after stroke, but response to rehabilitation is highly variable. Understanding this variability would help optimize treatment. Most assessments focus on the motor effects of stroke – such as hemiparesis – but there is emerging evidence that proprioceptive deficits play an important role in determining the response to movement rehabilitation after stroke. However, at present, there is a large diversity of techniques to measure proprioception, ranging from crude clinical assessments to complex robotic assessments. Further, it has been hypothesized that each assessment measures a different aspect of proprioception, such that generalization is minimal. Therefore, it is unclear which technique is best for gaining insight into movement rehabilitation. Here we focused on the ankle, a key joint for propulsion and balance. We designed and built an innovative robot for testing ankle proprioception and then used it to investigate the following questions: 1) How do proprioceptive errors depend on assessment parameters such as range of motion (ROM) and speed of movement? 2) Do different ankle proprioceptive assessments generalize in persons with stroke? 3) Is there site (i.e. joint) specificity to proprioception? 4) How well do ankle proprioceptive assessments predict gait function in persons with stroke? To answer these questions, we implemented two robotic proprioceptive assessments, joint position reproduction and Crisscross. Joint position reproduction is a well-established proprioceptive assessment, and Crisscross is a novel assessment that has implementation advantages for people with stroke. First, we found that proprioceptive acuity depends on the assessment parameters, with anticipatory errors increasing at slower speeds and with ROM. Second, we found generalization between the assessments in older unimpaired and stroke impaired individuals, but not younger unimpaired individuals. Third, we found proprioceptive processing has a body-general attribute that is shared across the ankles and fingers, particularly for young unimpaired participants. Lastly, we found that proprioceptive impairment weakly predicted gait speed after stroke, even though proprioceptive impairment was independent of motor impairment. As a side note, we also validated the robot for use to measure ROM and strength after stroke, showing that it has comparable reliability to experienced therapists but advantages in terms of resolution for strength measurement. This work therefore validates a novel robotic assessment of ankle proprioception (Crisscross), confirms its generalizability, and demonstrates its ability to quantify the effect of proprioceptive impairment on gait function after stroke. This work also provided the infrastructure to predict response to gait training after stroke using quantitative measures of proprioception acuity.