Can Robots Improve Arm Movement Recovery after Chronic Brain Injury?: a Rationale for Their Use Based on Experimentally Identified Motor Impairments

Significant potential exists for robotic and mechatronic devices to deliver therapy to individuals with a movement disability following stroke, traumatic brain injury, or cerebral palsy. We performed a series of experiments in order to identify which motor impairments should be targeted by such devices, in the context of a common functional deficit – decreased active range of motion of reaching – after chronic brain-injury. Our findings were that passive tissue restraint and agonist weakness, rather than spasticity or antagonist restraint, were the key contributors to decreased active range of motion across subjects. In addition, we observed striking patterns of abnormal contact force generation during guided reaching. Based on these results, we suggest that active assistance exercise is a rational therapeutic approach to improve arm movement recovery after chronic brain injury. We briefly discuss a simple, cost-effective way that such exercise could be implemented using robotic/mechatronic technology, and how such exercise could be adapted to treat abnormal muscle coordination. BACKGROUND Recently there has been a surge of interest in bringing robotic and mechatronic technology to bear on rehabilitation of movement after brain injury [1]. Stroke is currently the leading cause of severe disability in the U.S., and arm and hand movements are often preferentially impaired after stroke. A significant amount of recent research has therefore been focused on devices for therapy of the arm after stroke. Such devices could ultimately benefit approximately 300,000 new stroke survivors per year, as well as the more than 1.5 million chronic stroke survivors with movement disability in the U.S. A current difficulty in designing appropriate robotic technology for movement therapy of brain-injured individuals is that the optimal therapy techniques are unknown. More fundamentally, it is unclear what induces the observed movement impairments. Brain injury is often accompanied by a series of motor impairments, including weakness, spasticity, impaired movement range and impaired motor coordination. These impairments are mediated, in part, by changes to neural pathways, reflex systems, muscle, and connective tissue. Physical rehabilitation – and robotic therapy devices – could be targeted at any of these impairments. The goal of this study was therefore to identify the role of three motor impairments to a common functional Proceedings 6 International Conference on Rehabilitation Robotics, Stanford University, Stanford, California, U.S.A., July 1-2, pp. 9-15 deficit – decreased active range of motion of reaching (or decreased active “workspace”). Briefly, the three impairments were: 1. Increased passive tissue restraint, which may arise due to disuse and persistent abnormal posture of the spastic arm [2], and could cause an increased resistance to voluntary movement of the arm. 2. Antagonist muscle restraint, which could arise from reflex activation of antagonists (spasticity), or abnormal antagonist coactivation [3]. 3. Agonist muscle weakness, arising from destruction of key motor centers and outflow pathways and potentially by disuse atrophy [4]. METHODS To distinguish these three motor impairments, detailed mechanical measurements were made of the arms of five spastic hemiparetic subjects during reaching along a motorized guide. The device, which was used in the configuration shown in Fig. 1, allowed measurement of hand position and multiaxial force generation during guided reaching movements in the horizontal plane, and application of motorized stretches to the arm. After establishing workspace deficits along the device by the subjects, two tests were performed to elucidate the causes of these deficits. Each test was applied following individual reaches by each subject, across a set of twelve reaches: 1. Passive Restraint Test: To evaluate the level of passive tissue restraint at the workspace boundary, the ARM Guide returned the subject’s hand to the position from which the most recent reach was initiated. The arm was then moved slowly (< 4 cm/sec) back to the workspace boundary achieved by the most recent reach, and the force needed to hold the passive arm at the boundary was measured (Fig. 2, top). For comparison, the passive force generated by the contralateral arm (which was ostensibly normal) at a matched position was also evaluated. During these slow passive movements, EMG recordings of seven muscles surrounding the shoulder Figure 1: The Assisted Rehabilitation and Measurement Guide (“ARM Guide”). The subject’s forearm/hand was attached to a handle/splint that slid along a linear constraint via a low-friction, linear bearing. A six-axis force/torque sensor sensed contact forces between the hand and the constraint in the coordinate frame shown. A computercontrolled motor attached to a chain drive was used to drive the hand along the constraint. An optical encoder measured the position of the hand along the constraint. Proceedings 6 International Conference on Rehabilitation Robotics, Stanford University, Stanford, California, U.S.A., July 1-2, pp. 9-15 and elbow were used to verify that muscles were inactive. 2. Active Restraint Test: We hypothesized that any active restraint arising from activation of antagonist muscles during reaching would manifest itself as an increased stiffness following reaching, while the subject was still activating muscles and trying to move beyond the boundary. To evaluate this stiffness, a small stretch (the “terminal stretch”, 4 cm amplitude, bell-shaped velocity trajectory with a peak velocity of 15 cm/sec) was applied to the arm when hand velocity had dropped and remained below 1 mm/sec for 150 msec. An identical small stretch was applied following the slow passive movement of the arm through the same range (Fig. 2 top). The restraint force measured following the passive movement was then subtracted from the restraint force measured following reaching, in order to subtract out any passive forces common to the two conditions, such as those arising from passive stiffness, inertia, and damping. The result was the restraint force due solely to coactivation of muscles at the workspace boundary (Fig. 2 middle). For comparison, the terminal stiffness of the contralateral arm following matched, targeted, reaching movements, and following slow passive movement through the same range, were evaluated in a similar fashion. Five subjects were tested, each having suffered a hemispheric brain injury (four ischemic stroke, one traumatic brain injury) at least two years previously. The subjects had a wide range of movement ability as gauged by a standard clinical exam. The two subjects with the greatest movement ability exhibited workspace deficits during free movement, yet had a full active range of motion during reaching along the ARM Guide. To induce a workspace deficit along the ARM Guide, these subjects (D and E) 11 12 13 14 15 −15 −10 −5 0 5