Poststroke Upper Extremity Rehabilitation: A Review of Robotic

21 downloads 0 Views 309KB Size Report
Pittsburgh, and a member of the Human Engineering Research. Laboratories, VA Pittsburgh Healthcare System, Pittsburgh,. Pennsylvania. Sharon K. McDowell ...
Poststroke Upper Extremity Rehabilitation: A Review of Robotic Systems and Clinical Results Bambi R. Brewer, Sharon K. McDowell, and Lise C. Worthen-Chaudhari

Although the use of robotic devices to address neuromuscular rehabilitative goals represents a promising technological advance in medical care, the large number of systems being developed and varying levels of clinical study of the devices make it difficult to follow and interpret the results in this new field. This article is a review of the current state-of-the-art in robotic applications in poststroke therapy for the upper extremity, written specifically to help clinicians determine the differences between various systems. We concentrate primarily on systems that have been tested clinically. Robotic systems are grouped by rehabilitation application (e.g., gross motor movement, bilateral training, etc.), and, where possible, the neurorehabilitation strategies employed by each system are described. We close with a discussion of the benefits and concerns of using robotics in rehabilitation and an indication of challenges that must be addressed for therapeutic robots to be applied practically in the clinic. Key words: brain injury, hemiparesis, motor control, motor recovery, robotic rehabilitation, robotic therapy, stroke, upper extremity rehabilitation

T

he field of stroke rehabilitation is changing rapidly. One major recent change is the increasing availability of robotic applications for rehabilitation. Clinicians hear about robotic applications from articles in clinical journals, vendors anxious to sell products, news coverage of amazing recoveries attributed to a new device, patients searching for a miracle cure, and other rehabilitation professionals. The large amount of information available, including information that is not peer reviewed, makes it difficult to become involved in this new aspect of rehabilitation. Rehabilitation seeks to develop compensatory strategies as well as to induce neural plasticity and recovery. Robotics is a tool that can be used for both these goals— to assist an individual in activities of daily living or to assess and provide therapy for that individual. This article is a practical review of the clinically relevant literature that addresses robotic applications in the assessment and treatment of poststroke upper limb motor dysfunction. We discuss primarily haptic robots: systems that sense the movement of the user, use that information to make decisions and plan subsequent motion or output, and provide force feedback to the user via actuators (motors) in the system. Visual performance feedback is also frequently provided. A variety of neurorehabilitation strategies have

22

been used in the design of systems for robotic rehabilitation. For instance, repetitive movement training, or massed practice, and explicit learning paradigms are used by most current robotic systems for therapy. Robotic rehabilitation protocols incorporating neurorehabilitation strategies such as bilateral training, feedback distortion, implicit training, and functional task paradigms are in development at this time. In this review, we have grouped robotic systems by rehabilitation application: gross motor movements (e.g., reaching), bilateral training, fine moBambi R. Brewer, PhD, is an Assistant Professor, Department of Rehabilitation Science and Technology, University of Pittsburgh, and a member of the Human Engineering Research Laboratories, VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania. Sharon K. McDowell, MD, is the Director of Stroke Rehabilitation at Dodd Hall Rehabilitation Hospital, and an Assistant Professor, Department of Physical Medicine and Rehabilitation, The Ohio State University, Columbus, Ohio. Lise C. Worthen-Chaudhari, MS, is a Clinical Research Manager, Department of Physical Medicine and Rehabilitation, The Ohio State University, Columbus, Ohio. Top Stroke Rehabil 2007;14(6):22–44 © 2007 Thomas Land Publishers, Inc. www.thomasland.com doi: 10.1310/tsr1406-22

Robotic Systems and Clinical Results

tor movements (e.g., grasping), motor or visual feedback distortion to induce after-effects, telerehabilitation, and assessment. We hope this organization will be useful to both clinicians and researchers. As frequently as possible, we outline the underlying neurorehabilitation strategies of each system if such scientific principles are articulated in published literature. We begin by summarizing the literature on motor learning strategies recommended for poststroke neurorehabilitation and the research comparing the efficacy of robotic devices to conventional therapy in treating motor dysfunction after stroke. We then review the current robotic systems used for clinical research and for therapy and briefly summarize the results of published clinical research that has been performed with each system. We focus our discussion on systems with peer-reviewed clinical results, but we do include some systems that do not meet this criteria, particularly if the system is commercially available or if no mature systems exist for a given application area. Finally, benefits and concerns of incorporating robotics into standard rehabilitation practice are discussed, and challenges that we anticipate for robotic rehabilitation are summarized. Poststroke Motor Learning Principles It is important to incorporate motor learning principles into the design of any rehabilitation program. Although there is a body of literature regarding motor learning principles in individuals without neurological impairment, little research has been conducted to determine the motor learning paradigms that optimize poststroke motor recovery. As stated, most existing rehabilitation robots use a massed practice approach and an explicit learning paradigm, specifically a target endpoint or path shaping task; these assumptions are largely based on research conducted with the MIT-Manus robot demonstrating that repetitive, goal-directed, robot-assisted therapy can improve clinical outcomes.1–4 As pointed out in reviews by Krakauer,5,6 few rehabilitation robots designed for poststroke rehabilitation make use of other motor learning strategies, despite research suggesting that alternate paradigms induce better poststroke rehabilitation outcomes.

23

For example, it has been found that implicit learning paradigms may result in greater learning effects than explicit models.7,8 Implicit learning refers to skill acquisition without awareness or not directed on a conscious level; this is in contrast to explicit learning that refers to skill acquisition that is self or other directed.9 For instance, implicit learning may be incorporated into a robotic rehabilitation program by programming the target training movements into a game or other such virtual environment that the patient may explore. One example might be to train the arm to move in a specific trajectory by requiring the patient to move along the desired trajectory to navigate a virtual maze with a joystick. Other research has found that action observation and imitation10 and contextual interference11 may enhance retention of poststroke training. The literature on action observation and imitation indicates that robotic applications would be most successful if learning tasks were demonstrated to patients first. Contextual interference can be accomplished in a robotic environment through variations in the types of forces exerted by the robots, the types of movements required or facilitated by the system, and the type of visual feedback provided to users. In addition, it has been suggested that amplifying movement errors generated by patients while they train may result in greater rehabilitation gains than the more common strategy of assisting patients to move in the correct direction.12 For instance, one error magnification paradigm induces a patient’s limb to actively move in a particular trajectory by perturbing the limb in the direction opposite that of the desired trajectory. The subject counters the perturbing force in order to move in a straight line. Then, when the perturbing force is unexpectedly removed, the subject will move along the desired trajectory. This is known as a neuromuscular after-effect. In one study, 7 out of 10 individuals with poststroke hemiparesis who were tested demonstrated retention of such corrections for at least 2 hours after experiencing error magnified trials.13 For rehabilitation in a robotic environment to be useful, the skills learned in that environment must transfer to real-world tasks. This often means a focus by therapists on using real-world objects,

24

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

because movements toward a functional object differ from movements practiced with no target object.14,15 Sometimes robotic rehabilitation programs present virtual functional objects for manipulation within a virtual reality environment. This paradigm, however, has been shown to be effective in training functional tasks among individuals with and without poststroke deficits. Rose et al.16 found that, for a simple sensorimotor task, training in a virtual environment was as good as or better than training in the real-world environment when both groups were tested for retention in the real-world environment. In addition, Todorov, Shadmehr, and Bizzi17 used a virtual environment to train subjects to execute a difficult table tennis shot by following an expert’s trajectory. Subjects trained in the virtual environment performed the real-world task better than subjects trained with an equivalent amount of realworld practice. Holden et al.18 demonstrated that two stroke subjects were able to learn reaching trajectories in a virtual environment and transfer those skills to real-world reaching tasks. Because these studies show that motor learning can transfer from a virtual to a real environment, even for stroke patients, we can expect that skills learned in a robotic environment with a virtual reality interface will transfer to real-world tasks. In summary, the application of scientific principles within rehabilitation programs is essential if clinicians aim to maximize the functional recovery

Figure 1. InMotion2 (Interactive Motion Technologies, Inc., Boston, MA). This system is the commercially available version of the MITManus robotic system. Poststroke patients, both chronic and acute, who practiced reaching movements on this system improved their function.

of their patients. There is evidence that the motor learning strategy used by most existing rehabilitation robots, massed practice of an explicit learning paradigm, is effective. However, alternate motor learning strategies that have been found to be even more effective are rarely used in existing rehabilitation protocols. The scientific principles behind poststroke motor learning should be researched and considered when robotic and nonrobotic training programs are designed. Comparison of Robotic Therapy and Conventional Rehabilitation This article will discuss a wide range of systems that have been proposed to address a variety of issues in rehabilitation. Before going into the details of these systems, we want to provide a broad comparison of robotic and conventional therapy for the upper extremity, though trials directly comparing the two are rare and somewhat limited by methodological problems. This is a critical issue for the future of rehabilitation research, and one of the biggest concerns of physicians, therapists, and insurers. The earliest clinical trials involving rehabilitation robotics examined its usefulness in addition to conventional therapy. For example, the MIT-Manus (Figure 1; commercially available as the InMotion2; Interactive Motion Technologies, Inc., Boston, MA) provides therapy that uses re-

Robotic Systems and Clinical Results

petitive massed practice of reaching toward an endpoint. An evaluation of this robotic therapy compared with sham therapy was performed and reported in several studies for treatment of both acute and chronic poststroke hemiparesis.2–4,19,20 For individuals with acute stroke, robotic and sham therapies were considered in addition to conventional therapy. In a large study of acute stroke patients, the robotic therapy treatment group had significantly more improvement in motor function and strength at the shoulder and elbow (both initially and in 3-year follow-up) than the sham treatment group, although there were no other motor improvements at other joints and no functional improvements. Unfortunately, these results are not definitive because (a) the control group had significantly worse motor and cognitive function at baseline, and (b) the treatment group received 4–5 hours of additional therapy compared with 30 minutes for the control group. The additional time and work of therapy, not the robotic equipment, may account for the different results. Further studies evaluating the effect of robotic therapy with the MIT-Manus21,22 and other devices23 on chronic hemiparesis also show sustainable improvement in motor scores and function, although there was no comparison with conventional therapy. The Mirror Image Motion Enabler (MIME) system (Figure 2), a system designed for practice of upper extremity reaching, has been directly compared to conventional neurodevelopmental/ Bobath therapy for treatment of chronic stroke patients.24 This system incorporates an industrial robot that is coupled to the user’s hemiparetic arm. The subject moves in three dimensions while the force feedback of the robot assists or resists his or her movement. Relative to the group receiving neurodevelopmental therapy, the MIME treatment group had greater increases in reaching, strength, and Fugl-Meyer score for proximal movement, although there was little difference between the two groups at 6-month follow-up. Similar results were also found for individuals with subacute stroke.25 Thus, robotic therapy with MIME seems to speed recovery on some scales, but the long-term effect has not been found to be significantly different from that obtained with conventional therapy.

25

Figure 2. The Mirror Image Movement Enabler (MIME). This was the first system to explore bilateral training in robotic applications for stroke rehabilitation. Resulting data suggest that combined unilateral/bilateral therapy may reduce abnormal synergies more than unilateral therapy alone.

The Bi-Manu-Track (Reha-Stim, Berlin; Figure 3), another bilateral training device that focuses on forearm and wrist movement, has also been compared with nonrobotic therapy. In a relatively large trial (n = 44) of acute stroke patients, the patients in the treatment group who received robotic therapy improved more and maintained their advantage over 3 months after treatment compared with the patients in the control group who received movement practice aided by electrical stimulation.26 Finally, the Assisted Rehabilitation and Measurement (ARM) Guide (Figure 4), another system designed for practice of upper extremity reaching

26

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

Figure 3. Bi-Manu-Track (commercially available from Reha-Stim, Berlin). This system was designed for use in bilateral rehabilitation of wrist flexion/ extension and forearm pronation/supination. Individuals who received therapy with this system improved significantly more on the Fugl-Meyer scale than individuals who received electrical stimulation.

Figure 4. The Assisted Rehabilitation and Measurement (ARM) Guide. A motor can assist or resist in movement of the subject’s arm in the reaching direction along a linear track. For chronic stroke subjects, reaching practice with the ARM Guide led to motor improvements similar to those achieved through unassisted reaching. Reprinted with permission of the Journal of Rehabilitation Research and Development.

with robotic assistive or resistive forces, has been studied in a direct comparison with conventional therapy of free, unassisted reaching.27 No difference was found between the two methods. The authors are currently comparing unassisted reaching to reaching practice constrained, rather than assisted, by the robot. In fact, a study by Stein et al.28 compared efficacy of robotic assistive with resistive motor training and did not find any difference between these two very different modalities, leaving open the possibility that the movement practice itself may be more important for motor recovery than any robotic factors. The mixed results of these few direct comparative studies of robotic therapy may be related to a number of problems. As discussed earlier, many of the robotics designers are not applying the most accepted theories of motor recovery in their treatment design. Massed practice has been shown to provide significantly worse retention and generalizability of a skill compared with variable practice,29,30 yet much conventional and robotic

therapy continues to use a massed practice paradigm. Another potential difficulty with these trials is that many of the robots in these trials trained reaching, which impacts the shoulder and elbow; however, functional gains are more dependent on wrist and hand movement, issues not directly treated by the robotic devices described previously. Finally, robotic systems generally shape reaching movement according to the minimum jerk model, which is an appropriate model of reaching toward a point but less effective at describing reaching for a functional object.15 Many questions remain unanswered regarding the efficacy of robotic therapy versus conventional therapy and, more specifically, how various modes of robotic therapy compare to one another and to conventional therapy. Much more research in this area is necessary to determine how conventional therapy differs from robotic rehabilitation and how the two can best be combined for the optimal outcome of rehabilitation. A definitive study on this topic is currently in progress at Yale University

Robotic Systems and Clinical Results

(www.clinicaltrials.gov/ct/show/ NCT00372411?order=27). This study plans to recruit 158 individuals by 2009. The goal of the study is to compare robot-assisted rehabilitation with intensive conventional therapy and the standard of care. As described later, many researchers are investigating various types of robotic therapy to determine which are most effective for particular categories of individuals. In comparing robotic rehabilitation with conventional therapy, it is important to consider the goal of the robotic therapy. One goal is to increase the amount and intensity of rehabilitation available to individuals, to augment the conventional therapy received by individuals with acute stroke, and to enable stroke survivors to continue to receive therapy. In these cases, robotic rehabilitation protocols must be shown to have functional value, preferably equal to the functional value of an equivalent amount of conventional rehabilitation. However, robotic systems also enable modes of therapy that cannot be accomplished with conventional methods, for example, the use of precise, repeatable force patterns to address impairments in reaching.12 For these new techniques to be considered valuable, they must demonstrate advantages relative to conventional therapy. Rehabilitation of Gross Motor Movement As stated, most of the existing robotic therapy programs are based on repetitive movement training (massed practice) of explicit learning paradigms to combat learned nonuse. The principle underlying such programs is that repetitive practice with the hemiparetic arm over an extended period of time will encourage cortical reorganization and increased habitual use of the hemiparetic arm in activities of daily living (ADLs). This is also the idea underlying constraint-induced movement therapy. The majority of robotic rehabilitation applications for the upper extremity are designed to encourage gross motor improvements, particularly improvements in reaching with the shoulder and elbow. These systems use a shaping exercise where a target endpoint is defined and the patient is asked to move their arm to reach the endpoint. Variations between existing systems can be described by the

27

force-feedback paradigm used. These include passive movement of the impaired limb, movement assistance induced by presence or absence of force/ muscle activity, and movement resistance in response to motion or to directional errors. These paradigms require application of different types of force, and this feedback is performed easily and repeatably by robotic systems. Systems to facilitate passive movement move the patient’s limb without requiring muscular effort on the part of the patient. Active assistive systems provide extra force to help users complete the target movement. Resistive systems resist muscular force applied by the patient to the robot arm. Many of the existing robots can be operated in different force modes depending on which neurorehabilitation strategy the therapist wants to apply. To demonstrate the abilities of robotic systems for passive movement of a subject’s arm, we will consider the GENTLE/s system (Figure 5). In this system, a commercially available 3 degrees of freedom (dof) robot, the HapticMASTER from Moog FCS Robotics, Inc. (Nieuw-Yennep, The Netherlands), is attached to the wrist via a gimbel with 3 passive degrees of freedom.31 This system suspends the upper limb from a sling to eliminate gravity. When operated in the passive mode, the GENTLE/s system guides the user’s arm along a smooth predefined trajectory. When compared with task practice with the arm suspended in the sling, the GENTLE/s system seems to yield a greater rate of improvement for some physical variables such as shoulder range of motion and the ability to perform certain arm movements. This system can also operate in assistive and resistive modes.32 Chronic stroke patients were found to improve function, as measured by the Fugl-Meyer scale’s upper limb section, using the GENTLE/s therapy robot when trained in all three modes available.33 The MIME system described previously also has a passive mode that moves the subject’s arm along a specific, desired trajectory.24,25,34 The REHAROB system uses industrial robots to passively move the upper limb with the goal of decreasing spasticity.35 Eight patients who received therapy with REHAROB in addition to conventional physiotherapy decreased in spasticity, as measured by the Ashworth scale.

28

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

Figure 5. The GENTLE/s system. The user’s upper limb is suspended from a sling to eliminate gravity while the user interacts with a HapticMASTER robot. When compared with task practice with the arm suspended in the sling, the GENTLE/s system seems to yield a greater rate of improvement for some physical variables such as shoulder range of motion and the ability to perform certain arm movements.

A different neurorehabilitation technique uses force feedback to assist muscles by applying force in the direction appropriate for task completion. The MIT-Manus robot, a system commercially available as the InMotion2, has successfully used this active assistive strategy to achieve significant motor recovery gains in individuals with acute1 and chronic impairments.22,36,37 The InMotion2 system allows for smooth, 2 dof tabletop movement. The user is shown a target on a computer screen, and the system applies force to the user’s arm if he or she is unable to independently move the robot in the target direction. Particularly exciting, tests conducted with individuals after they underwent training on the InMotion2 robot demonstrated that they retained gains made during the initial robotic therapy both 3 months (measured by the Fugl-Meyer scale for Shoulder/Elbow and Coordination and Motor Power Scale for Shoulder/Elbow)22 and 3 years after training (measured by the Motor Status score for shoulder/elbow). 2 Some robotic assistance modes predefine the movement path, whereas other modes require the patient to plan the trajectory to the desired endpoint. MIME, GENTLE/s, and the

commercially available Reo Robot (Motorika, Ltd., Mount Laurel, NJ) engage this active assistance force paradigm while allowing for more range of motion at the shoulder and elbow joints than the InMotion2 planar system. Sometimes it is preferable to require motion from the patient before providing movement assistance. The ARM Guide linear robotic trainer27 requires patients to apply specific force patterns at threshold levels before movement is allowed or assisted (Figure 4). In this way, ARM Guide targets the training of coordination-specific movement initiation. Other robots with at least one mode that provides assistance after detecting a nominal force directed toward the desired trajectory include the GENTLE/s and MIME. Electrical stimulation systems also provide a novel way to deliver force to an individual’s muscles. These systems assist movement in a way similar to the haptic systems discussed earlier, but instead of motors they use the biological forceproducing unit—the individual’s muscle fibers. Electrical stimulation systems currently being marketed commercially include the H200TM system (Figure 6) by Bioness, Inc. (Santa Clarita, CA) and

Robotic Systems and Clinical Results

29

Figure 6. H200™ (Bioness, Inc., Santa Clarita, CA). This system is a commercially available system that electrically stimulates muscles to produce hand motion.

NeuroMoveTM by Zynex Medical (Littleton, CO). An electrical stimulation unit can be combined with an electromyographic recording unit so that the patient must demonstrate a threshold muscle activity before any stimulation is added. Resistive systems either impede the user’s movement in the target direction or constrain the direction of the user’s movements, preventing deviations from the target trajectory. The former is used in the active-constrained mode of the MIME system, whereas the latter is used in the GENTLE/s robot patient-active mode and in all modes of the MIME system. Another variation on force resistance provides force along the desired trajectory then unexpectedly removes resistance such that patients move along the desired trajectory8 (see the section titled, “Feedback Distortion for Rehabilitation”). A study by Stein et al.28 compared efficacy of movement assistance to motor training incorporating resistance to movement in the target direction. Both groups showed decreases in motor impairment and improved Fugl-Meyer scores, but this study found no difference between the two types of training. This may be because strength is only one component necessary for functional use of the upper limb. Further research is necessary to determine the optimal pattern of force feedback to be used in robotic therapy. The haptic robots discussed earlier allow individuals to practice reaching movements with a variety of assistive and resistive patterns of force

feedback, but because the individual typically grasps a handle on the endpoint of the robotic arm, the user cannot grasp everyday objects while using the robot. This is of concern because context and the use of real objects have been shown to affect movement14 and therapists try to incorporate daily activities and objects into therapy as much as possible. The Activities of Daily Living Exercise Robot (ADLER) provides one of the few robotic environments in which active assistance is provided to patients as they reach toward real, functional objects (e.g., coffee mug, food item).15 Still in development, this system aims to deliver poststroke motor rehabilitation paradigms involving ADL-specific context. As another approach to address this problem of incorporating daily activities and objects, exoskeletons are being developed to allow users to interact naturally with objects while receiving force feedback from the system. Examples of this approach include the ARMin38 and Pneu-WREX39 systems. Exoskeletons may also help to mitigate gravity during therapy. Research on the effects of assisting or enhancing gravity’s pull on paretic arm muscle synergies is currently being conducted using the Arm Coordination Training (ACT) 3D system.40,41 One commercially available exoskeletal system, the Myomo e100 (Myomo, Inc., Boston, MA) assists the elbow joint to move after activity is detected from selected muscles; six hemiparetic patients who practiced for 18 hours over 6 weeks on this device showed

30

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

significant improvement in Fugl-Meyer and summated modified Ashworth scales for the elbow flexors and extensors.42 Bilateral Training Symmetric bilateral movement of the paretic and nonparetic arms has been proposed as a potentially useful rehabilitation strategy. Bilateral movements may promote functional improvement by reinforcing corticospinal pathways from the undamaged cortical hemisphere to the affected limb.25 In addition, the undamaged hemisphere may influence the damaged hemisphere through the intercallosal fibers as an individual moves bilaterally.43 Repetitive bilateral movement using a mechanical (nonrobotic) device has been shown to increase cortical activation in the contralesional hemisphere that may be associated with improved function.44 The MIME (Figure 2) was the first robotic system to explore the use of bilateral training for upper extremity rehabilitation. MIME consists of an industrial robot that interacts with the affected arm of an individual with hemiparesis and a device that measures the position of the unaffected arm.45 In the bilateral mode, an individual using MIME moves the unaffected arm and the robot guides the paretic arm to mirror that movement. The MIME system can also operate unilaterally; the paretic arm can receive therapy in passive, active, and resistive modes. As described previously, therapy using the MIME has been compared to an equivalent amount of neurodevelopmental therapy involving the use of the paretic arm in functional tasks.24 To specifically test the efficacy of the bilateral mode, Lum et al.25 used four groups of individuals with subacute stroke: one group experienced only the three unilateral MIME modes, one group experienced only the bilateral mode, one group experienced both unilateral and bilateral modes, and the control group received neurodevelopmental therapy. Relative to the control group, the MIME groups that received unilateral and combined unilateral/bilateral therapy showed larger improvements in the proximal Fugl-Meyer score, but these differences were not maintained at a 6-month follow-up. There was no difference in the increase in Fugl-Meyer score be-

tween the unilateral and the combined unilateral/ bilateral groups. Similar improvements were not found for the group that used the MIME system only in the bilateral mode. However, the data did suggest that individuals who received combined unilateral/bilateral therapy had a greater reduction in abnormal synergy than individuals who received only unilateral MIME therapy. Hesse et al.46 developed the Bi-Manu-Track (Figure 3), a robot designed to rehabilitate wrist flexion/extension and forearm pronation/supination. This system allows passive movement of both arms, active movement of both arms against resistance, and passive movement of the paretic arm to mirror movement of the nonparetic arm.26 In a trial with 44 individuals, Hesse et al.26 compared therapy using the Bi-Manu-Track with an equivalent time period of electrical stimulation to facilitate wrist extension. All individuals had acute stroke, and all received standard physical and occupational therapy as well as the experimental intervention. Individuals who received robotic therapy improved significantly more on the FuglMeyer scale than individuals who received electrical stimulation. This difference was maintained at a 3-month follow-up. The authors did not assess the impact of the bilateral features of the system on subject outcome. Some bilateral haptic systems such as Driver SEAT47 and GameCycle48 (Three Rivers Holdings LLC, Mesa, AZ) use a gaming interface with potential for telerehabilitation use. These systems increase subject motivation by incorporating exercise into an entertaining video game. In addition to the robotic systems discussed earlier, cheaper mechanical (nonrobotic) systems have been developed for bilateral arm training, such as Nudelholz.43 Despite the range of systems that have been proposed for bilateral training, more research is needed to determine whether, and in what situations, bilateral training enhances unilateral training in stroke rehabilitation. Rehabilitation for Fine Motor Movement Although a variety of robotic systems are being used for the practice of gross reaching movements after stroke, the use of robotics for wrist and hand rehabilitation is much less developed. Robotic sys-

Robotic Systems and Clinical Results

Figure 7. InMotion3 (Interactive Motion Technologies, Inc., Boston, MA). This system incorporates the same rehabilitation science principles and movement characteristics as the InMotion2 (MITManus) system but targets rehabilitation of the wrist joint. Preliminary clinical results indicate that therapy using this robot leads to a decrease in wrist impairment.

tems for the hand and wrist can be categorized based on the strategy underlying the rehabilitation program. One group of robots focuses on the improvement of low-level motor variables such as range of motion, whereas the goal of the other systems is to increase coordination during functional tasks. Several systems have been proposed to increase the range of motion, velocity, and force capacity of the wrist or fingers through practice of simple movements. The InMotion2, described previously, is complemented by the InMotion3, a robot designed for wrist rehabilitation (Figure 7).49 This robot, also commercially available from Interactive Motion Technologies, Inc., can be used to train wrist flexion/extension, abduction/adduction, and pronation/supination. Preliminary clinical results indicate that therapy using this robot leads to a decrease in wrist impairment, as indicated by increased scores on the wrist and fingers subcomponents of the Fugl-Meyer scale.50 Practice of wrist flexion/extension using a simpler system designed

31

by Colombo et al.51 improved the Fugl-Meyer score and increased wrist extension for seven individuals with stroke. The Hand MentorTM system from Kinetic Muscles, Inc. (Tempe, AZ) is based on an active repetitive motion paradigm. The user is asked to actively flex and extend the wrist and fingers, and the robot assists when necessary. The system can also passively stretch the subject’s wrist and finger flexors and incorporates EMG recording to encourage increased muscle recruitment. Work with the Hand MentorTM resulted in functional improvement for a single individual with stroke.52 To move on to systems focusing on elemental features of hand movement, Boian et al.53 have created a robotic system for hand rehabilitation using a commercially available Cyberglove® (Immersion Corp., San Jose, CA) to measure the joint angles of the hand and a prototype, the Rutgers Master II-ND glove, to apply force to the fingertips (Figure 8). The current rehabilitation protocol for this system includes exercises designed to increase the range of motion, maximum velocity, and maximum force of the fingers. Eight subjects who have used the system demonstrated increased range of motion and velocity, as well as improved performance on the Jebsen Test of Hand Function.23 In addition to systems that address basic components of movement, several systems focus on promoting coordinated movements of the hand and the ability to move each finger independently of the others. Pernalete et al.54 have designed a system to improve eye–hand coordination in children with fine motor delays. Lambercy et al.55 have developed the Haptic Knob to allow users to promote the restoration of hand grip combined with pronation/supination of the forearm. Other systems target coordinated movements of the fingers. Taub et al.56 have used a variety of instrumented objects to automate constraint-induced movement therapy (AutoCITE), a paradigm consisting of repetitive task practice to overcome learned nonuse. These objects measure an individual’s completion of fine motor ADLs such as threading and objectflipping. When supervised by a therapist 25%, 50%, or 100% of the time, therapy with this system has been shown to be as effective as one-onone constraint-induced therapy. Wearable systems that provide assistance to the user in grasping and releasing real-world objects have also been pro-

32

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

Figure 8. System focusing on rehabilitation of independent finger movement.23,53 Research suggests that individuals who performed rehabilitation protocols with this system improved in fractionation of the fingers. Copyright © Rutgers University, CAIP Center. Reprinted with permission.

posed (Augmented Reality and HWARD).57,58 Brewer, Klatzky, and Matsuoka59 noted functional improvements in three patients who participated in a 6-week protocol that involved pinching and releasing the index finger and thumb in a robotic environment. Finally, the rehabilitation protocol developed by Boian et al.53 includes an exercise designed to promote independent movement of each finger, and individuals using this system improved in fractionation of the fingers.23 This system has recently been extended to include a bilateral virtual piano activity to encourage independent movement of the fingers.60 These applications demonstrate that robotic technology is beginning to be used for the restoration of normative coordination patterns. In addition, fine motor rehabilitation is receiving increased attention in rehabilitation robotics. A number of robots designed for hand rehabilitation are currently in development,61–64 including modules designed to work with the InMotion265 and the GENTLE/s systems.66 Feedback Distortion for Rehabilitation Robotic systems used in upper extremity rehabilitation generally provide at least two types of feedback to the individual—the force feedback of

the robot and visual feedback displayed on a computer screen. Several groups are currently investigating how distorting each type of feedback can be used to optimize the outcome of robotic rehabilitation. Patton and Mussa-Ivaldi8 used a 2 dof robot to explore a new method to produce desired movement patterns using force perturbations and an implicit learning paradigm. In an experiment with healthy subjects, participants learned to move the arm in a straight line in the presence of a force field applied by the robot. Then, when the training force was removed, the after-effect of the adaptation caused the subject to shift in the direction of the target movement (Figure 9). In this case, the distortion is the force perturbing the arm from a straight-line path. A similar experiment with 18 persons with chronic stroke found that subjects did have the ability to adapt to the distortion and that after-effects of the force training persisted longer in subjects with stroke.12 Brewer, Klatzky, and Matsuoka59 have investigated the use of visual distortion to encourage improved performance during rehabilitation of individuals with chronic stroke who may be reluctant to move beyond their habitual limits of motion. After measuring the limits of visual distortion67 and showing that vision dominates

Robotic Systems and Clinical Results

33

Telerehabilitation

Direction of movement

Direction of exerted force Direction of desired aftereffect

Figure 9. Schematic of training paradigm used by Patton et al.,12 the developers of the VRROOM system. This paradigm involves resisting patient movement when movement occurs out of the desired motion trajectory; it has been found that this intervention results in after-effects that ultimately benefit patient movement function.

kinesthesis in their robotic environment,68 they showed that individuals with chronic stroke follow visual distortion to levels of performance above that predicted by their initial performance in the robotic environment. Subjects also showed functional improvements in pincer grasp as a result of the rehabilitation protocol. Wei et al.69 investigated the effects of a different type of visual distortion, visual error augmentation. Each subject’s error was visually multiplied or increased by a constant offset while he or she made reaching movements with the robot. They found that visual error augmentation accelerated learning, whereas visual distortion via a constant offset both speeded and increased the magnitude of learning.

Many recent studies have highlighted the importance of intense training to promote cortical reorganization and overcome learned nonuse.70,71 Studies have documented functional improvements in individuals many years after stroke. At the same time, current mechanisms of reimbursement are curtailing the amount of therapy available to individuals. Telerehabilitation, meaning rehabilitation in the home supervised by a remote therapist, has been proposed to address these concerns. As an example, Reinkensmeyer et al.72 proposed Java Therapy, a system in which an individual exercises the wrist in the home by connecting to a website and using a commercially available force-feedback joystick to complete therapy games and assessments. A remote therapist can use the same website to monitor the individual’s progress and to design a program of exercises for the client. Java Therapy is currently being extended using T-WREX, an orthosis that provides gravity compensation while the user completes simulated functional activities.73 Five subjects who used the T-WREX for 8 weeks showed significant improvements on the FuglMeyer scale. Systems have also been proposed that extend the handle of the force-feedback joystick such that an increased range of motion is required for its use.74,75 Although systems for telerehabilitation such as Java Therapy use very simple robotic components, both the InMotion276 and the Rutgers Master II force-feedback glove77 have been suggested for telerehabilitation applications. However, the price of systems such as the InMotion2 will have to be greatly reduced for home-based therapy with these devices to become a real possibility. Cheaper, nonrobotic alternatives such as sensor gloves produced for video game use78 may provide a better alternative for current telerehabilitation work. Assessment for Rehabilitation The use of robotic technology for assessment has the potential to provide therapists with objective, precise measurements of an individual’s function. Robots can measure position and force at a high frequency with high accuracy and precision, gen-

34

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

erating data that can be used to quantify an individual’s functional performance. Such assessment can enable therapists to track an individual’s progress in therapy and to evaluate the efficacy of various interventions. Robotic assessment can also be used to choose an interface to assistive technology that is appropriate for a particular user.79 Simple, nonrobotic systems can be used to address these problems. For example, Memberg and Crago80 and Innman and Haugland81 have constructed instrumented objects, objects (e.g., a fork) that mimic everyday items but contain sensors to measure force or orientation as individuals use the objects in simulated ADL. Though these items are useful for evaluating specific ADLs, robotic systems have more flexibility to assess functional performance for multiple tasks. In addition, robotic systems provide the ability to easily modify a particular task, for example, by changing the characteristics of a virtual object. Motion capture, the reconstruction of human movement based on markers placed on the body and tracked by a camera system, is another nonrobotic method that has recently been proposed as an assessment tool for rehabilitation. Murphy et al.82 analyzed the motion of the upper extremity as subjects reached for a glass of water, took a drink, and returned the glass to the table. They measured the test–retest reliability of this analysis and proposed similar analyses of ADLs to quantify impairment of the upper extremity. Allin83 has proposed a motor statistic that can be measured for upper extremity functional tasks using a low-cost camera system. This motor statistic was validated relative to the Arm Motor Ability Test, a standardized functional test often used with individuals with stroke. In terms of robotic systems for assessment, Bardorfer et al.84 used a commercially available PHANTOM robot (Sensable Technologies, Inc., Woburn, MA) to create an assessment tool based on a virtual maze. As the user passed through the virtual maze, the robot measured the velocity of movement and the characteristics of the user’s collisions with the maze walls. Bardorfer et al.85 also used a PHANTOM robot to assess the ability of individuals to track circular or linear patterns. They analyzed the accuracy and velocity of track-

ing both with and without pseudorandom force disturbances exerted by the robot. Amirabdollahian et al.79 propose a task more closely related to a standard clinical measure; they have implemented a virtual peg-in-hole task using a PHANTOM robot. This use of a robot for an apparently simple task underscores some of the advantages of robotic assessment. By using the robot, Amirabdollahian et al. were able to measure the position of the virtual peg at a sampling rate of 1000 Hz throughout the task. This sampling rate was sufficient to enable the researchers to examine not only the total time required for the task but also the path taken by the peg and the spread of velocities used during the task. In addition, the robotic environment enabled them to easily vary hole separation, hole size, and the weight of the virtual peg. This sort of variation would require multiple sets of equipment for assessment systems using instrumented objects or motion capture, but the characteristics of virtual objects can be varied quickly and easily using a robotic system. Colombo et al.51 proposed several simple indices based on the amount of active movement observed during robotic rehabilitation; this group performed preliminary validation of these measures relative to the Fugl-Meyer scale. These characteristics indicate the potential of robotics for assessment. As yet, no robotic test has yet been evaluated for test–retest reliability or been thoroughly validated relative to clinical scales, although Colombo et al.51 performed a preliminary validation relative to the Fugl-Meyer scale of a measure of active movement during robotic rehabilitation. The aforementioned applications focused on the use of robotics for evaluating the functional performance of individuals and tracking this performance over time. Several research groups have instead concentrated on the use of robotic testing to elucidate the extracted submovements from reaching movements made by individuals with stroke and demonstrated that the submovements become less distinct as individuals recover.86 The authors argue that this blending of movements may reflect a recalibration of the individual’s internal model for predicting the physical effects of neural commands.87 Reinkensmeyer, Dewald, and Rymer88 examined the use of the ARM Guide robot

Robotic Systems and Clinical Results

to quantify abnormal synergies in gross reaching movements in individuals with brain injury. Reinkensmeyer, Schmit, and Rymer89 also used the ARM Guide to quantify and distinguish passive tissue restraint and active muscle restraint in individuals with brain injury. These examples demonstrate that research groups are using robotic assessment to investigate the underlying causes of functional impairment. The goal of such studies is to provide information that can be used to design effective rehabilitation protocols to address these underlying issues. Benefits of Robotic Rehabilitation A summary of the robotic systems discussed in this article is found in Table 1. In addition, commercially available systems designed for robotic rehabilitation and other commercially available systems that have been used in rehabilitation are summarized in Table 2. Robotic systems could help with rehabilitation in a number of ways. First, robotic therapy could provide more intense and longer duration therapy. A significant amount of rehabilitation and research for motor recovery currently uses the principle of massed practice and demonstrates motor recovery. If we can program a machine to perform appropriate motor tasks with a patient, it can automate repetitive tasks such as range of motion and reaching, providing those tasks at high repetitions and for long durations. In addition, this may increase availability of high intensity, extended duration therapy methods, such as constraint-induced movement therapy and repetitive task practice, which have been difficult to deliver in standard rehabilitation clinics. A second benefit of robotic therapy is that the device may be able to perform therapy that would be more difficult for a therapist to do, such as providing repeatable haptic or force feedback as well as visual feedback that amplifies movement errors and helps correct the movement pattern. Good therapists use their sense of touch to apply pressure to guide and correct movement patterns, but robotics potentially can simulate this type of environment while introducing even more complex feedback. Third, robotic devices can automate a significant

35

segment of therapy time. Using one of these devices, a therapist could set up a patient on the appropriate therapy protocol and then the remainder of a session could be run by a therapy aid. This would potentially allow workplace multiplication, if a therapist could treat more than one patient simultaneously with associated cost savings or increased therapy time provided to each patient. As robotic equipment becomes safer and more portable, a therapist may even be able to program a rehabilitation protocol that the patient performs at home for multiple sessions per day. This may facilitate the ability to continue delivering rehabilitation even in chronic populations, in contrast to the current environment of shorter periods of rehabilitation during only the acute poststroke stage. Indeed, the recent literature demonstrates continued motor recovery in chronic stroke patients, raising questions about the reality of a clinical plateau.90 A fourth benefit of robotics is the potential to deliver and even automate subject-specific therapy and progression of therapy. As the field has progressed, Krebs et al.91 have suggested that therapy should be adjusted based on the abilities of the individual. This is similar in concept to the traditional rehabilitation practice of customizing therapy goals to the needs and desires of the individual.92 Krebs et al. have focused on using subject performance to adjust the amount of assistance provided by the InMotion2. Based on subject performance within a rehabilitation session, Krebs and colleagues progressively adjust movement time and amount of variation allowed about the target direction. Robotics allows for measuring patient progression in one session then using the information to plot progression in future sessions. The aim of this adjustment is to enable the user to gradually relearn to make independent reaching movements. Similar gradual progression of movement goals has also been implemented by Stewart et al.93 The ability of robotic systems to precisely measure and recall force profiles and movement paths is an advantage when planning the course of rehabilitation over multiple sessions. Finally, these devices can provide more precise, objective, reliable measurement of motor function. In addition to measures of strength, sensors on robotic devices can more finely and objectively

Manufacturer/developer

Northwestern University

Hocoma Medical Engineering, Inc. & University Hospital Balgrist

Rehabilitation Institute of Chicago & University of California at Irvine

University of Alabama at Birmingham & VA Richmond

Department of NeuroRehab at Klinik-Berlin/Charité Hospital

Rehab Research & Development Center at VA, Palo Alto, CA/ Birmingham VA Medical Center

University of Reading

Kinetic Muscles, Inc.

University of California at Irvine

Interactive Motion Technologies, Inc.

System

Act 3D

ARMin

ARM-GUIDE

AutoCITE

Bi-Manu-Track

Driver SEAT

GENTLE/s

Hand Mentor™

HWARD

InMotion2 (MIT-MANUS)

- Massed practice with target endpoint - Addresses learned nonuse

- Functional tasks practiced

- Massed practice - EMG biofeedback - Gaming paradigms available

- Massed practice with target pathway - Addresses learned nonuse - Visual and haptic motor feedback

- Bilateral training - Functional ADL practiced

- Addresses learned nonuse - Mirror movement

- Constraint-induced movement therapy - Unilateral - Addresses learned nonuse - Functional tasks practiced

- Massed practice with target endpoint - Reaching in any plane - Addresses learned nonuse - Gravity mitigated

- Massed practice with target - Audio, visual and tactile feedback - Gravity mitigation (exoskeleton)

- Addresses inefficient synergies - Gravity mitigated or enhanced

Rehabilitation science principles

Table 1. Robotic rehabilitation systems in use for clinical research or therapy

Shoulder, elbow

Wrist, hand

Wrist, hand

Shoulder, elbow

Shoulder, elbow, wrist, hand

Forearm, wrist

Shoulder, elbow, wrist, hand

Shoulder,elbow

Shoulder, elbow

Shoulder, elbow

Joints

- Passive, active assistive, resistive - Unilateral - Planar movement

- Active assistive - Grasp and release

- Active assistive

- Passive, active assistive, resistive -Trajectory fork (decision making) - Unilateral

- Bilateral - ADL

- Passive, active assistive, resistive - Bilateral - Mirror movement

- Unilateral - ADL

- Passive, active assistive, resistive - Unilateral

- Passive, active assistive

- Passive, active assistive, resistive - Unilateral

Movement characteristics

36 TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

University of California at Irvine

Rehab Research & Development Center at VA, Palo Alto, CA

Myomo, Inc.

University of California at Irvine

Budapest University of Technology and Economics, Hungary (with 4 other collaborating institutions in Europe)

Motorika, Ltd.

New Jersey Institute of Technology /Rutgers University (Merians et al.,23 Boian et al.53)

Robotics Institute at Carnegie Mellon University (Brewer, Klatzky, and Matsuoka59)

Rehabilitation Institute of Chicago & Northwestern University

Java Therapy

MIME

Myomo e100

PneuWrex

REHAROB

Reo Robot

System in development

System in development

VRROOM

- Implicit learning -Trajectory error augmentation to induce after effects

- Visual feedback distortion - Addresses learned nonuse

- Massed practice - Gaming paradigm

- Variable training - Visual feedback

- Massed practice of single or multijoint coordination - Range of motion training to decrease spasticity - Visual feedback

- Functional task training - Gravity mitigation (exoskeleton)

- Massed practice - EMG biofeedback - Portable exoskeleton

- Massed practice with target endpoint - Mirror movements - Addresses learned nonuse

- Visual and haptic motor feedback - Functional tasks practiced - Telerehabilitation

- Massed practice with target endpoint - Progressive adjustment to patient - Addresses learned nonuse

All upper extremity joints

Hand, fingers

Fingers

Shoulder, elbow

Shoulder, elbow, forearm

Shoulder, elbow

Elbow

Shoulder, elbow

Wrist

Forearm, wrist

- Various

- Active

- Assistive

- Passive, active assistive, resistive

- Passive

- Passive, active assistive

- Active assistive

- Passive, active assistive, resistive - Unilateral/ bilateral

- ADL

- Passive, active assistive, resistive - Unilateral

Note: ACT 3D = Arm Coordination Training; ADL = activities of daily living; ARM Guide = Assisted Rehabilitation and Measurement Guide; AutoCITE = Automated ConstraintInduced Therapy Extension.

Interactive Motion Technologies, Inc.

InMotion3

Robotic Systems and Clinical Results 37

38

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

Table 2. Commercially manufactured systems and components of robotic clinical systems

System

Manufacturer / developer

Cost

Robotic systems using device

Characteristics

CyberGlove®

Immersion Corp.

$$$

- Boian et al.53 - VROOM

Measures joint angles, no force feedback

Force-feedback joystick

Various

$

Java Therapy

Force feedback at hand & wrist

GameCycle

3 Rivers, Inc.

$$

Independent system

Upper body ergometer

H200™

Bioness, Inc.

$$

Independent system

- Electrical muscle stimulation prosthesis - Wrist, fingers

Hand Mentor™

Kinetic Muscles, Inc.

$$

Independent system

See Table 1

HapticMASTER

Moog FCS Robotics, Inc.

$$$$

- ACT 3D - VRROOM - GENTLE/s

-3 dof -Force-controlled haptic interface - Shoulder, elbow, wrist

InMotion2

Interactive Motion Technologies, Inc.

$$$$

Independent system

See Table 1

InMotion3

Interactive Motion Technologies, Inc.

$$$$

Independent system

See Table 1

MyomoTM e100

Myomo, Inc

$$

Independent system

See Table 1

NeuroMove™

Zynex Medical

$$

Nonrobotic system

- Motor reorganization with electrical stimulation - Shoulder, elbow, wrist, fingers

PHANTOMTM

Sensable Technologies, Inc.

$$$

- VRROOM - Brewer, Klatzky, and Matsuoka59 - Amirabdollahian et al.79

- 3 dof - Force feedback in 3 dimensions

Reo Robot

Motorika, Ltd.

$$$$

Independent system

See Table 1

WAMTM

Barrett Technology Inc.

$$$$

VRROOM

- Large workspace - 3 dof - Force feedback in 3 dimensions

(MIT-MANUS)

Note: Cost: $ = Under $1000; $$ = $1000 to $10,000; $$$ = $10,000 to $50,000; $$$$ = over $50,000. dof = degrees of freedom.

measure aspects of spasticity, breaking it down into components parts and improper forces that are consistent with described synergy patterns.89,94 These devices can also measure other elements of motor function that have been more difficult to evaluate clinically, such as force, velocity, coordination, smoothness, the ability to make continuous movements, and the number of self-correc-

tions during a movement.19,86,95 These initial assessments can aid in diagnosing the pathophysiology underlying a patient’s impairments. The ongoing measures taken by the robotics devices during therapy can be very useful, because the measures are precise enough to capture even very small changes. Tracking those changes can help guide our therapy as we can evaluate efficacy of different

Robotic Systems and Clinical Results

physical or pharmacologic therapies. These changes may also help motivate patients during therapy and help document progress for patient insurers. Addition of Robotics to the Clinic: Practical Considerations Acceptance of robotics by patients and clinicians is a significant barrier to their use in the clinic at this point. Subjects in robotics research have been optimistic about having robotic therapists.19,96,97 These individuals did not express any concerns about their safety in using a robotic device. They also believed that they were benefiting from the robotic therapy. However, they are a self-selected group, because they agreed to participate in research with robotic devices. In addition, even these individuals, when given a choice, have stated that they prefer a human therapist.19 In general, people have accepted robotic assistance in many areas of life with greater ease through time. As robotics becomes more familiar, patients become more technologically savvy, and, ideally, robotic devices become more portable and user-friendly, we expect that patient acceptance will continue to improve. Many review articles also state that therapists have been reluctant to use robotic therapy.96,98,99 Therapists are trained to evaluate patient motor dysfunction by touch and to adjust treatment forces based on feedback. Although haptic robotics may be able to perform some of the same functions, the effectiveness of robotic systems has not been compared to a trained, experienced therapist in this area. Despite this, therapists involved in clinical research trials have been optimistic.97 A survey of therapists was published in 1991,96 which showed therapists felt that the robotic device was able to maintain a patient’s attention and provided good exercise, forcing patients to work harder and reach further. Therapists also liked some of the treatment programs available but wanted the ability to program custom movement patterns to use in the treatment of specific individuals. They were very interested in modules for cognitive training and auditory feedback. It is also clear that set-up time is crucial; it is not surprising

39

that therapists would prefer not to use a robotic device if time to set it up with the patient exceeded 5 minutes. Physical Medicine & Rehabilitation (PM&R) physicians will also need to become involved in robotic therapy if this treatment is to complete the transition to standard care. To improve our understanding of their concerns, one of the authors sent out an informal survey about the use of robotics to certified PM&R physicians who are members of the American Academy of PM&R and who had self-identified as specialists in treating stroke and other brain injuries. Response rate was about 16% on that survey (n = 24 respondents). Just fewer than 21% of respondents indicated that they were not interested in the use of robotic equipment in rehabilitation; other options were slightly interested, interested, and very interested. Reasons for citing no interest at all varied from expense and inadequate time to evaluate robotic options to concerns about ability to provide individualized therapy for patients and availability of devices appropriate for pediatric patients. The remaining discussion of the survey will review the responses of PM&R physicians who were interested in robotics. Responders were evenly split in their perceptions about efficacy of robotic therapy compared to conventional therapy for stroke patients, with many simply stating that they did not know. When asked about ways they would use robotics, 79% would use this equipment for rehabilitation and 58% would use it for research. Despite the qualities and promise of robotic equipment for use in diagnosis or outcome assessment, very few physicians stated any intent to use equipment in this way. The biggest concerns that these respondents listed were, in order of frequency, (a) not believing that robotics is better than conventional therapy, (b) expense of the equipment, (c) not having had appropriate time to evaluate the different options, and (d) not finding any device good enough or portable enough for easy use in the clinic. Despite the magnitude of concerns, these physicians report a relatively high level of interest in the area and also report purchase or plans to purchase a variety of robotic or related devices, including the Bioness, Robomedica, ro-

40

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

botic gait systems, Locomat, InMotion2, Rutgers ankle, and active joint brace. Cost is another concern that currently limits the number of rehabilitation clinics that use robotic devices. Many of the most effective machines are currently research prototypes and not commercially available. Indeed, the expense of the equipment was a concern for 25% of responding clinicians who said they were not interested in the use of robotic equipment and more than 50% of responding clinicians who were interested in and/or were using robotic equipment. Perhaps, with further development, the cost of this equipment and technology will decrease and this will ease adoption into the clinic. The cost and portability for most devices will need to improve by orders of magnitude before most patients can use these devices in their home to extend therapy duration. Adoption is also slowed by some additional financial considerations. Capital to purchase the equipment is not the only expense. In addition, staff need time to be trained to use the devices. Once the equipment is in use, it does not replace therapists; they will continue to be involved in designing the therapy, setting up the patients, and possibly monitoring the patients. It is unclear if they can successfully bill for this time at all, much less whether an additional bill for equipment use will be reimbursed by insurance. Insurance companies do not currently pay for robotics-enhanced services, perhaps because of too little evidence demonstrating efficacy. Challenges for Robotic Rehabilitation At this point, the biggest challenge for robotics in rehabilitation is the limited scientific evidence elucidating appropriate treatment of motor dysfunction following stroke. Clearly, if the most effective interventions to optimize neural plasticity are unknown, a robotic system cannot be designed to incorporate those interventions. It is important to point out that the need for more stroke-specific motor recovery research also limits the effectiveness of conventional therapy. Robotic applications have the potential to address the challenge of conducting clinically relevant research. The precise, quantified assessment of pathophysiology and re-

covery patterns that can be achieved with these devices has the potential to greatly facilitate research of poststroke motor function recovery and rehabilitation In addition, use of robotic equipment allows more consistent, standardized, controlled movement therapy with multiple patients, thus allowing for more rigorous testing of different therapy types. Current theoretical models of therapy suggest that therapy is more effective if it is functional and performed in a context-appropriate environment, for example, performing ADLs at the bathroom sink.100 One approach to meeting this goal is for clinicians to work with engineers to design lightweight devices that can be worn by a patient. Alternatively, systems such as virtual reality interfaces that enable multiple functional tasks to be practiced and varied are another approach to meeting this goal. An example of this approach is VRROOM,101 a system of multiple robotic devices and an immersive virtual reality environment that can be used to simulate a variety of functional tasks. In the future, we will need more collaborative studies to improve our understanding of the efficacy of robotic therapy. Currently, we can be certain that robotic therapy can be safely provided and that studies have demonstrated improved motor scores following certain robotic therapies. The relationship between robot-assisted and conventional therapy is still unclear; further controlled studies are critical. We also need to understand the combinational effects of robotic therapy with different lesion types and behavioral dysfunction. Both brain lesion size and specific lesion areas are related to recovery and response to therapy.4,102–105 In our informal survey, most physicians were interested in the use of robotic equipment to retrain movement, but a number were interested in other impairments such as spasticity, hemi-neglect, apraxia, and sensory loss. We need to determine optimal therapy strategies for different brain injuries and the subsequent impairments. More research is also needed to delineate proper treatment dosing/intensity, duration, and timing for robotic therapy and to consider interactions between specific lesions, specific neurobehavioral syndromes, and rehabilitation treatment (robotic or conventional).

Robotic Systems and Clinical Results

In summary, haptic robotic equipment continues to be developed at higher levels of technology. Different devices use various theories about motor recovery and neural plasticity, and reports about their efficacy are beginning to form a body of encouraging evidence. The challenge of applying robots in the clinic, however, is large. In addition

41

to practical considerations, we will need a significant body of research to define the appropriate populations for robotics and to develop the appropriate treatment protocols to maximize their use. Fortunately, the robotic equipment itself will improve our ability to accomplish this important work.

REFERENCES 1. Aisen ML, Krebs H I, Hogan N, McDowell F, Volpe BT. The effect of robot-assisted therapy and rehabilitative training on motor recovery following stroke. Arch Neurol. 1997;54:443–446. 2. Volpe BT, Krebs HI, Hogan N, Edelsteinn L, Diels CM, Aisen ML. Robot training enhanced motor outcome in patients with stroke maintained over 3 years. Neurology. 1999;53:1874–1876. 3. Volpe BT, Krebs HI, Hogan N, Edelstein OL, Diels C, Aisen M. A novel approach to stroke rehabilitation: robot-aided sensorimotor stimulation. Neurology. 2000;54:1938–1944. 4. Krebs HI, Volpe BT, Aisen ML, Hogan N. Increasing productivity and quality of care: robot-aided neuro-rehabilitation. J Rehabil Res Dev. 2000;37: 639–652. 5. Krakauer JW. Arm function after stroke: from physiology to recovery. Semin Neurol. 2005;25: 384–395. 6. Krakauer JW. Motor learning: its relevance to stroke recovery and neurorehabilitation. Curr Opin Neurol. 2006;19:84–90. 7. Pohl PS, McDowd JM, Filion D, Richards LG, Stiers W. Implicit learning of a motor skill after mild and moderate stroke. Clin Rehabil. 2006;20:246–253. 8. Patton JL, Mussa-Ivaldi FA. Robot-assisted adaptive training: custom force fields for teaching movement patterns. IEEE Trans Biomed Eng. 2004; 51:636–646. 9. Cleeremans A. Implicit learning. In: Nadel L, ed. Encyclopedia of Cognitive Science. London: Nature Publishing Group; 2003. 10. Buccino G, Solodkin A, Small SL. Functions of the mirror neuron system: implications for neurorehabilitation. Cogn Behav Neurol. 2006; 19:55–63. 11. Hanlon RE. Motor learning following unilateral stroke. Arch Phys Med Rehabil. 1996;77:811–815. 12. Patton JL, Stoykov ME, Kovic M, Mussa-Ivaldi FA. Evaluation of robotic training forces that either enhance or reduce error in chronic hemiparetic stroke survivors. Exp Brain Res. 2006;168:368–383. 13. Patton JL, Kovic M, Mussa-Ivaldi FA. Custom-designed haptic training for restoring reaching ability to individuals with poststroke hemiparesis. J Rehabil Res Dev. 2006;43:643–656. 14. Trombly CA, Wu C. Effect of rehabilitation tasks on organization of movement after stroke. Am J Occup Ther. 1999;53:333–344.

15. Wisneski KJ, Johnson MJ. Quantifying kinematics of purposeful movements to real, imagined, or absent functional objects: implications for modelling trajectories for robot-assisted ADL tasks. J Neuroengineering Rehabil. 2007;4:7. 16. Rose FD, Attree EA, Brooks BM, Parslow DM, Penn PR, Ambihaipahan N. Training in virtual environments: transfer to real world tasks and equivalence to real task training. Ergonomics. 2000;43:494–511. 17. Todorov E, Shadmehr R, Bizzi E. Augmented feedback presented in a virtual environment accelerates learning of a difficult motor task. J Mot Behav. 1997;29: 147–158. 18. Holden M, Todorov E, Callahan J, Bizzi E. Virtual environment training improves motor performance in two patients with stroke: case report. Neurology Rep. 1999; 23:57–67. 19. Krebs HI, Hogan N, Aisen ML, Volpe BT. Robotaided neurorehabilitation. IEEE Trans Rehabil Eng. 1998;6:75–87. 20. Volpe BT, Krebs HI, Hogan N. Is robot-aided sensorimotor training in stroke rehabilitation a realistic option? Curr Opin Neurol. 2001;14:745–752. 21. Fasoli SE, Krebs HI, Hogan N. Robotic technology and stroke rehabilitation: translating research into practice. Top Stroke Rehabil. 2004;11:11–19. 22. Ferraro M, Palazzolo JJ, Krol J, Krebs HI, Hogan N, Volpe BT. Robot-aided sensorimotor arm training improves outcome in patients with chronic stroke. Neurology. 2003;61:1604–1607. 23. Merians AS, Poizner H, Boian R, Burdea G, Adamovich S. Sensorimotor training in a virtual reality environment: Does it improve functional recovery poststroke? Neurorehabil Neural Repair. 2006;20:252–267. 24. Lum PS, Burgar CG, Shor PC, Majmundar M, Van der Loos M. Robot-assisted movement training compared with conventional therapy techniques for the rehabilitation of upper-limb motor function after stroke. Arch Phys Med Rehabil. 2002; 83:952–959. 25. Lum PS, Burgar CG, Van der Loos M, Shor PC, Majmundar M, Yap R. MIME robotic device for upper-limb neurorehabilitation in subacute stroke subjects: a follow-up study. J Rehabil Res Dev. 2006;43:631–642. 26. Hesse S, Werner C, Pohl M, Rueckriem S, Mehrholz J, Lingnau ML. Computerized arm training improves the motor control of the severely

42

27.

28.

29. 30. 31.

32.

33.

34.

35.

36.

37. 38.

39.

40.

41.

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

affected arm after stroke: a single-blinded randomized trial in two centers. Stroke. 2005;36: 1960–1966. Kahn LE, Lum PS, Rymer WZ, Reinkensmeyer DJ. Robot-assisted movement training for the strokeimpaired arm: Does it matter what the robot does? J Rehabil Res Dev. 2006;43:619–630. Stein J, Krebs HI, Frontera WR, Fasoli SE, Hughes R, Hogan N. Comparison of two techniques of robotaided upper limb exercise training after stroke. Am J Phys Med Rehabil. 2004;83:720–728. Schmidt RA, Lee TD. Motor Control and Learning. Champaign, IL: Human Kinetics Publishers; 1999. Shea CH, Kohl RM. Composition of practice: influence on the retention of motor skills. Res Q Exerc Sport. 1991;62:187–195. Coote S, Stokes E, Murphy B, Harwin W. The effect of GENTLE/s robot-mediated therapy on upper extremity dysfunction post stroke. Presented at: International Conference on Rehabilitation Robotics; 2003; Daejeon, South Korea. Amirabdollahian F, Loureiro R, Harwin W. Minimum jerk trajectory control for rehabilitation and haptic applications. In: ICRA 2002 - IEEE International Conference on Robotics and Automation. Washington, DC: IEEE; 2002:3380–3385. Amirabdollahian F, Loureiro R, Gradwell E, Collin C, Harwin W, Johnson G. Multivariate analysis of the Fugl-Meyer outcome measures assessing the effectiveness of GENTLE/S robot-mediated stroke therapy. J Neuroengineering Rehabil. 2007;4:4. Lum PS, Burgar CG, Shor PC. Evidence for improved muscle activation patterns after retraining of reaching movements with the MIME robotic system in subjects with post-stroke hemiparesis. IEEE Trans Neural Syst Rehabil Eng. 2004;12:186– 194. Fazekas G, Horvath M, Toth A. A novel robot training system designed to supplement upper limb physiotherapy of patients with spastic hemiparesis. Int J Rehabil Res. 2006;29:251–254. Fasoli SE, Krebs HI, Stein J, Frontera WR, Hughes R, Hogan N. Robotic therapy for chronic motor impairments after stroke: follow-up results. Arch Phys Med Rehabil. 2004;85:1106–1111. Krebs HI, Ferraro M, Buerger SP, et al. Rehabilitation robotics: pilot trial of a spatial extension for MIT-Manus. J Neuroengineering Rehabil. 2004;1:5. Nef T, Riener R. ARMin: Design of a novel arm rehabilitation robot. Presented at: International Conference on Rehabilitation Robotics; 2005; Chicago, IL. Wolbrecht E, Leavitt J, Reinkensmeyer DJ, Bobrow J. Control of a pneumatic orthosis for upper extremity stroke rehabilitation. In: EMBS ’06. 28th Annual Conference of the IEEE. New York: IEEE; 2006: 2687–2693. Sukal T, Dewald JP. Ellis M. Use of a novel robotic system for quantification of upper limb work area following stroke. Conf Proc IEEE Eng Med Biol Soc. 2005;5:5032–5035. Sukal T, Ellis M, Dewald JP. Source of work area reduction following hemiparetic stroke and pre-

42.

43.

44.

45.

46.

47. 48.

49.

50.

51.

52.

53. 54.

55.

56.

liminary intervention using the ACT 3D system. In: EMBS ’06. 28th Annual Conference of the IEEE. New York: IEEE; 2006:177–180. Stein J, Narendran K, McBean J, Krebs K, Hughes R. Electromyography-controlled exoskeletal upper-limb-powered orthosis for exercise training after stroke. Am J Phys Med Rehabil. 2007;86:255– 261. Hesse S, Schmidt H, Werner C. Machines to support motor rehabilitation after stroke: 10 years of experience in Berlin. J Rehabil Res Dev. 2006; 43:671–678. Luft AR, McCombe-Waller S, Whitall J, et al. Repetitive bilateral arm training and motor cortex activation in chronic stroke: a randomized controlled trial. JAMA. 2004;292:1853–1861. Burgar CG, Lum PS, Shor PC, Van der Loos HFM. Development of robots for rehabilitation therapy: the Palo Alto VA/Stanford experience. J Rehabil Res Dev. 2000;37:663–673. Hesse S, Schulte-Tigges G, Konrad M, Bardeleben A, Werner C. Robot-assisted arm trainer for the passive and active practice of bilateral forearm and wrist movements in hemiparetic subjects. Arch Phys Med Rehabil. 2003;84:915–920. Johnson MJ, Van der Loos HF, Burgar CG, Shor PC, Leifer L. Driver’s SEAT: a car steering upper limb therapy device. Robotica. 2003; 21:13–23. Widman LM, McDonald CM, Abresch RT. Effectiveness of an upper extremity exercise device integrated with computer gaming for aerobic training in adolescents with spinal cord dysfunction. J Spinal Cord Med. 2006;29:363–370. Celestino J, Krebs HI, Hogan N. A robot for wrist rehabilitation: characterization and intial results. Presented at: International Conference on Rehabilitation Robotics; 2003; Daejeon, South Korea. Charles SK, Krebs HI, Volpe B, Lynch D, Hogan N. Wrist rehabilitation following stroke: initial clinical results. In: IEEE International Conference on Rehabilitation Robotics. New York: IEEE; 2005:13–16. Colombo R, Pisano F, Micera S. et al. Robotic techniques for upper limb evaluation and rehabilitation of stroke patients. IEEE Trans Neural Syst Rehabil Eng. 2005;13: 311–324. Alberts JL, Tresilian JR, Stelmach GE. The co-ordination and phasing of a bilateral prehension task. The influence of Parkinson’s disease. Brain. 1998;121(pt 4):725–742. Boian R, Sharma A, Han C, et al. Virtual realitybased post-stroke hand rehabilitation. Stud Health Technol Inform. 2002;85:64–70. Pernalete N, Edwards S, Gottipati R, Tipple J, Kolipakam V, Dubey RV. Eye-hand coordination assessment/therapy using a robotic haptic device. Presented at: International Conference on Rehabilitation Robotics; 2002; Chicago, IL. Lambercy O, Dovat L, Johnson V, et al. Development of a robot-assisted rehabilitation therapy to train hand function for activities of daily living. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007:678–682. Taub E, Lum PS, Hardin P, Mark VW, Uswatte G.

Robotic Systems and Clinical Results

57.

58.

59.

60.

61. 62.

63.

64.

65.

66.

67.

68. 69.

70.

AutoCITE: automated delivery of CI therapy with reduced effort by therapists. Stroke. 2005;36: 1301–1304. Luo X, Kenyon RV, Kline T, Waldinger HC, Kamper DG. An augmented reality training environment for post-stroke finger extension rehabilitation. Presented at: International Conference on Rehabilitation Robotics; 2005; Chicago, IL. Takahashi CD, Der-Yeghiaian L, Le VH, Cramer SC. A robotic device for hand motor therapy after stroke. Presented at: International Conference on Rehabilitation Robotics; 2005; Chicago, IL. Brewer BR, Klatzky R, Matsuoka Y. Initial therapeutic results of visual feedback manipulation in robotic rehabilitation. Presented at: International Workshop on Virtual Rehabilitation; 2006; New York. Adamovich S, Qiu Q, Talati B, Fluet G, Merians A. Design of a virtual reality-based system for hand and arm rehabilitation. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007:958–963. Mali U, Goljar N, Munih M. Application of haptic interface for finger exercise. IEEE Trans Neural Syst Rehabil Eng. 2006;14:352–360. Worsnopp TT, Peshkin MA, Colgate JE, Kamper DG. An actuated finger exoskeleton for hand rehabilitation following stroke. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007:896–901. Kawasaki H, Ito S, Ishigure Y, et al. Development of a hand motion assist robot for rehabilitation therapy by patient self-motion control. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007: 234–240. Dovat L, Lambercy O, Johnson V, et al. A cable driven robotic system to train finger function after stroke. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007: 222–227. Masia L, Krebs HI, Cappa P, Hogan N. Design, characterization, and impedance limits of a hand robot. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007: 1085–1089. Loureiro R, Harwin W. Reach & grasp therapy: design and control of a 9-DOF robotic neurorehabilitation system. In: IEEE 10th International Conference on Rehabilitation Robotics. New York: IEEE; 2007:757–763. Brewer BR, Fagan M, Klatzky R, Matsuoka Y. Perceptual limits for a robotic rehabilitation environment using visual feedback distortion. IEEE Trans Neural Syst Rehabil Eng. 2005;13:1–11. Brewer BR, Klatzky R, Matsuoka Y. Visual-feedback distortion in a robotic rehabilitation environment. Proc IEEE. 2006;94:1739–1751. Wei Y, Bajaj P, Scheidt R, Patton JL. A real-time haptic/graphic demonstration of how error augmentation can enhance learning. Presented at: IEEE-International Conference on Robotics and Automation; 2005; Barcelona, Spain. Taub E, Uswatte G, Pidikiti R. Constraint-Induced

71.

72.

73.

74.

75.

76. 77.

78.

79.

80. 81. 82.

83.

84.

85.

43

Movement Therapy: a new family of techniques with broad application to physical rehabilitation— a clinical review. J Rehabil Res Dev. 1999;36:237– 251. Liepert J, Miltner WH, Bauder H, et al. Motor cortex plasticity during constraint-induced movement therapy in stroke patients. Neurosci Lett. 1998;250:5–8. Reinkensmeyer DJ, Pang CT, Nessler JA, Painter CC. Web-based telerehabilitation for the upper extremity after stroke. IEEE Trans Neural Syst Rehabil Eng. 2002;10:102–108. Sanchez RJ, Liu J, Rao S, et al. Automating arm movement training following severe stroke: functional exercises with quantitative feedback in a gravity-reduced environment. IEEE Trans Neural Syst Rehabil Eng. 2006;14: 378–389. Johnson LM, Winters JM. Evaluation of tracking performance using joystick manipulators that engage different arm workspaces. In: International Conference on Rehabilitation Robotics. New York: IEEE; 2005:86–90. Zipfel E, Verkaaik JV, Gaukrodger S, King M. Design and development of an upper limb exerciser and game controller. Presented at: Rehabilitation Engineering Society of North America Conference; 2006; Atlanta, GA. Carignan CR, Krebs HI. Telerehabilitation robotics: bright lights, big future? J Rehabil Res Dev. 2006;43:695–710. Popescu FC, Rymer WZ. End points of planar reaching movements are disrupted by small force pulses: an evaluation of the hypothesis of equifinality. J Neurophysiol. 2000;84:2670–2679. Morrow K, Docan C, Burdea G, Merians A. Lowcost virtual rehabilitation of the hand for patients post-stroke. In: International Workshop on Virtual Rehabilitation. New York: IEEE; 2006:6–10. Amirabdollahian F, Gomes GT, Johnson GR. The peg-in-hole: a VR-based haptic assessment for quantifying upper limb performance and skills. In: International Conference on Rehabilitation Robotics. New York: IEEE; 2005:422–425. Memberg WD, Crago PE. Instrumented objects for quantitative evaluation of hand grasp. J Rehabil Res Dev. 1007;34:82–90. Inmann A, Haugland M. An instrumented object for evaluation of lateral hand grasp during functional tasks. J Med Eng Technol. 2001;25:207–211. Murphy MA., Sunnerhagen KS, Johnels B, Willen C. Three-dimensional kinematic motion analysis of a daily activity drinking from a glass: a pilot study. J Neuroengineering Rehabil. 2006;3:18. Allin SJ. Machine perception for occupational therapy: toward prediction of post-stroke functional scores in the home. Presented at: Rehabilitation Engineering Society of North America (RESNA) Conference; 2006; Atlanta, GA. Bardorfer A, Munih M, Zupan A, Primozic A. Upper limb motion analysis using haptic interface. IEEE/ASME Transactions on Mechatronics. 2001;6:253–260. Bardorfer A, Munih M, Zupan A, Ceru B. Linear

44

86. 87.

88.

89.

90. 91.

92.

93.

94. 95.

TOPICS IN STROKE REHABILITATION/NOV-DEC 2007

and circular tracking exercises in haptic virtual environments for hand control assessment. In: International Conference on Rehabilitation Robotics. New York: IEEE; 2005:66–69. Rohrer B, Fasoli S, Krebs HI, et al. Movement smoothness changes during stroke recovery. J Neurosci. 2002;22:8297–8304. Hogan N, Krebs HI, Rohrer B, et al. Motions or muscles? Some behavioral factors underlying robotic assistance of motor recovery. J Rehabil Res Dev. 2006;43:605–618. Reinkensmeyer DJ, Dewald JP, Rymer WZ. Guidance-based quantification of arm impairment following brain injury: a pilot study. IEEE Trans Rehabil Eng. 1999;7:1–11. Reinkensmeyer DJ, Schmit BD, Rymer WZ. Assessment of active and passive restraint during guided reaching after chronic brain injury. Ann Biomed Eng. 1999;27:805–814. Page SJ, Gater DR, Bach YRP. Reconsidering the motor recovery plateau in stroke rehabilitation. Arch Phys Med Rehabil. 2004; 85:1377–1381. Krebs HI, Palazzolo JJ, Dipietro L, et al. Rehabilitation robotics: performance-based progressive robot-assisted therapy. Autonomous Robots. 2003;15:7–20. Radomski MV. Planning, guiding, and documenting therapy. In: Trombly CA, Radomski MV, eds. Occupational Therapy for Physical Dysfunction. Philadelphia: Lippincott Williams & Wilkins; 2002. Stewart JC, Yeh S, Jung Y, et al. Pilot trial results from a virtual reality system designed to enhance recovery of skilled arm and hand movements after stroke. In: International Workshop on Virtual Rehabilitation. New York: IEEE; 2006: 11–17. Dewald JP, Beer RF. Abnormal joint torque patterns in the paretic upper limb of subjects with hemiparesis. Muscle Nerve. 2001;24:273–283. Krebs HI, Aisen ML, Volpe BT, Hogan N. Quantization of continuous arm movements in humans

96.

97. 98. 99.

100. 101.

102.

103.

104. 105.

with brain injury. Proc Natl Acad Sci USA. 1999; 96:4645–4649. Dijkers MP, deBear PC, Erlandson RF, Kristy K, Geer DM, Nichols A. Patient and staff acceptance of robotic technology in occupational therapy: a pilot study. J Rehabil Res Dev. 1991;28:33–44. Volpe BT, Krebs HI, Hogan N. Robot-aided sensorimotor training in stroke rehabilitation. Adv Neurol. 2003;92:429–433. Glass K, Hall K. Occupational therapists’ views about the use of robotic aids for people with disabilities. Am Occup Ther. 1987;41:745–747. Hidler J, Nichols D, Pelliccio M, Brady K. Advances in the understanding and treatment of stroke impairment using robotic devices. Top Stroke Rehabil. 2005;12:22–35. Carr J, Shepard R. A Motor Relearning Programme for Stroke, 2nd ed. Gaithersburg, MD: Aspen Publishers; 1998. Patton J, Dawe G, Scharver C, Mussa-Ivaldi F, Kenyon R. Robotics and virtual reality: a perfect marriage for motor control research and rehabilitation. Assist Technol. 2006;18:181–195. Pantano P, Formisano R, Ricci M, et al. Prolonged muscular flaccidity after stroke. Morphological and functional brain alterations. Brain. 1995;118(pt 5): 1329–1338. Chollet F, DiPiero V, Wise RJ, Brooks DJ, Dolan RJ, Frackowiak RS. The functional anatomy of motor recovery after stroke in humans: a study with positron emission tomography. Ann Neurol. 1991;29:63–71. Frackowiak RS, Weiller C, Chollet F. The functional anatomy of recovery from brain injury. Ciba Found Symp. 1991;163:235–244; discussion 244–249. Ward NS, Brown MM, Thompson AJ, Frackowiak RS. Neural correlates of motor recovery after stroke: a longitudinal fMRI study. Brain. 2003;126:2476–2496.