A Paradigm Shift forRehabilitation Robotics

Therapeutic Robots Enhance ClinicianProductivity in Facilitating Patient Recovery

BY HERMANO I. KREBS, LAURA DIPIETRO,SHELLY LEVY-lZEOEK, SUSAN E. FASOU,AVRIELLE RYKMAN-BERlAND, JOHANNA ZlPSE,JENNIFER A. FAWCETT, JOEL STEIN,HOWARD POIZNER, ALBERT C. LO,BRUCE T. VOLPE, AND NEVILLE HOGAN

he demand for rehabilitation services is growing apacewith the graying of the populatioo. According to theWorld Health Organization (WHO), senior citizens atleast 65 years of age will increase in number by 88% in

the coming years. By 2050, the United States' contingent ofseniors is expected to double from approximately 40 to 80 mil-lion (Figme I). With this increase comes increased incidenceof age-related disorders. The following is an example.

Cerebral vascular accident (S/7oke): For every decadeafter age 55. the relative incidence of stroke doubles (1].At present,. more than 700,000 Americans suffer strokeseach year, more than half survive. In the United Statesalone. close to 5 million stroke victims are alive today[lJ. Even higher incidence is observed in other devel-oped countries with older populations (e.g., Japan).There can be some respite from stroke if phannacologi-cal agents are successfully developed to preserve vesselpatency, to protect neurons, and to stimulate neosynapto-genesis. However, if that should happen, an increase instroke survival rates may well increase the number ofstroke victims in need of rehabilitation services.

The need for rehabilitation services is even more pressing ifwe consider neurological diagnoses ocher than stroke. The fol-lowing are some examples.

Cerebral palsy (CP); This is a term used to describe agroup of chronic conditions affecting body movementand muscle coordi.o.ation. which affects 2.8 in 1,000 chil-dren born in the United States each year [2}. It is causedby the damage to one or more areas of the brain, usuallyoccwring during fetal development; before, during, orshortly after hirth; or during infancy (3). In addition,studies have shown that at least 5,()(X) infants and tod~d1ers and 1,200-1,500 preschoolers are diagoosed withCP each year as developmental and motor delays be-come more apparent.Multiple sclerosis (MS): It is the third leading cause ofdisability in young adults in the United States, with aprevalence of approximately one in every 1,000, withtwo thirds of these cases occurring in females for anestimated total of 350,000 in the United States. MS is a

chronic disease of the central nervous system, and cur-rently there is no cure for it. None of the current immu-nomodulatory therapies convincingly alters long-tennprogressive disability. Patients with nontraumatic spinalcord injUI)' (SCI) may well equal this incidence.SCI: It is the leading cause of disability in YOWlg adults inthe Uoited States. The incidence of SO has beeo estimatedto be between 30 and 60 new cases per million of the U.S.population per year, with an estimated prevalence of250,000 (700-900 per million of population). Almost80% are males younger than 40 years (4J, [5J.Parkinson's disease (PDJ: It is a neurodegenerative disor-der characterized by bradykinesia., resting tremor. rigidity,and postural reflex impairment. It is one of the most com-moo neurological disorders, with estimates ranging from500,000 to 1.5 million affected in the United States, andthis number will increase over the next 50 years as theaverage life expectancy increases (6), {7]. TIle time ofonset is typically between 40 and 70 years of age, withpeaks in the sixth decade of life. Most conunonly, theclinical status of the patient with PD progresses from arelatively modest limitation at the time of diagnosis to anever-increasing disability over a period of 10--20 years.

There is both a need and an opportunity to deploy technolo-gies such as robotics to assist recovery. This, in essence, consti-tutes a paradigm shift moving the field of rehabilitation roboticsbeyond assistive technology that helps an individual cope withthe environment to a new class of physically interactive, user·friendly robots that facilitate recovery. 1bera.peutic robotsfurther the clinicians' goal of facilitating recovery not only bydelivering measured therapy but also by affording new ways toevaluate patients' progress. Here, we will focus on this para·digm shift: robots that support and enhance the productivity ofclinicians in their efforts to facilitate an individual's recovery.

A recent IEEE puhlication described the rematkahle groW1hofactivities in the rehabilitation roOOtics in the past few years andourdevices deployed in clinical trials (see (9J fordetails). In addi-tion, the lead anthor recently edited two special issues presentinga broad spectrum of leading research efforts in rehabilitationrobotics: from the clinical perspective in the Veterans Adminis~trarion's Journal of Rehabilitation Research and Development(JRRD), September/Octoher 2006, and from the eogineering

perspective in the IEEE Transactions on Neural Sysrems andRehabilitation Engineering, September 2007 (IEEE lNSRE).Therefore, rather than attempting to replicate the content of thesepublications, in this article, we restrict ourselves to a sununary ofthe growing activities with our robots. Most researchers employ~ing rehabilitation robotics have focused on stroke, as this is thesingle largest cause of permanent disabiliry. We briefly reviewthe pohlished clinical literature in this emerging field and ourinitial clinical results in stroke. However, we also report ourinitial effOI1S that go beyond stroke, broadening the potentialpopulation that might benefit from this class of technology hydiscussing case studies of applications to other neurologicaldiseases. We will also highlight the underexploited potential ofthis technology as an evaluation tool.

RobolicGymFigure 2(a) shows the pioneer of its class, the M1T-MANUSfor planar shoulder-and.-elbow therapy, whose developmentstarted in late 1989 (IOJ. MIT-MANUS. from the Massachu-sens Institute of Technology's (Ml'I) mono "Mens et Manus"(mind and hand), was desigoed for clinical oeurological appli-cations. Unlike most indusoial robots, MIT-MANUS wasconfigured for safe, stable. and highly compliant operation inclose physical COlltact with humans. This was achieved usingimpedaoce control. a key feamre of the robot control system.Here, we opted for a fixed-based design robot (actuators arefixed with respect to a stationary coordinate system) versus anexoskeleton design. {Fixed-based or end-cffector designs likeMlT-MANUS are simpler, afford sigoificantly faster "don"and "doff" (setup time much smaller) than exoskeletondesigns but typically occupy a larger volume. We employ arule of thumb to guide us in the selection of configurationbased on the target range of motion. For limb segmentmovements requiring joint angles to change by 45° or less, end-effector designs appear to offer better compromises. Conversely,exoskeletal designs appear to offer better choices for largerranges of motion.) Its computer control system modulates the

way the robot reacts to mechanical penurbation from a patien:or clinician and ensures a gentle compliant behavior. Toemachine was designed to have a low i~trinsic end-ooim im-pedance (i.e,. be back-drivable) to allow weak patie~ts to ex~press movements without constraint and offer minimal resistanceat speeds up 2 m/s (the approximate upper limit of unimpairedhuman performa.'1ce, hence the target of therapy, and the maxi-mwn speed observed in some pathologies, e,g., the shock-likemovements of myoclonus).

Following the successful clinical trials with WT-MANUS,a one-degree-or-freedom (DoF) module [Figure 2(b)J wasconceived to extend the benefits of planar robotic therapy tospatial arm·movements, including movements against gravity.Therapists' suggestions that functional reaching movementsoften occur in a range of motion close to shoulder scaption areincorporated in the design {I]].

Fig. 2. A gym of robots. (0) The shoulder-and-elbow MIT~MANUS delivering therapy to a child with CP (SpauldingRehabAttatfon Hospital). (b) The antigravity module (Balti-more VAMC). (c) A person with PO and deep brain stimula-tlon COBS) practicing with the wrist robot (University ofCalifornia Son Diego). (d) The .Integrated system affordingwhol~rm training (transport of the arm and manipulationof objects, VA CooP Study Randomized Clinical Triol CSP558). (e) The hond module (without cover). (f) A person withMS during therapy with the anklebot 0Nest Hoven VA Medi-cal Center).

To extend the Dearment envelope beyond the shoulder andelbow, we designed and built a wrist module for robotic therapy[12]. It features three active DoF, namely, flexion or extension,abduction or adduction, and forearm pronation or supination,and can be operated as a stand-alone unit (Figure 2(c)J ormounted to the end of a shoulder-and~elbow robot (Figure2(d»). The most common fonn of unimpaired upper-extremieymovement appears to be a combination of translating the hand(with the shoulder and elbow) to a location in space and orient-ing the hand (with the wrist) to facilitate object manipulation.We are employing the wrist robot in both configurations todetermine what works best and for what rype of patient «(l3);VA Coop Study, Randomized Clinical Trial, CSP 558).

The next module required for whole-arm therapy is a handrobot [Figure 2(e)]. Moving a patient's hand is not a simpletask because the human hand has 15 joints with a total of22 DoF. Thus, it was prudent to determine how many DoF arenecessary for a patient to perform the majority of everydayfunctional tasks. Here, our clinical experience with more than300 stroke patients was invaluable as it allowed us to identifywhat was most likely to work in the clinic (and what probablywould not). Although individual digit opposition (e.g., thumbto pinkie) may be important for the unimpaired human hand, itis clearly beyond the realistic expectations of most of ourpatients whose impairment level falls between severe andmoderate; a device to manipulate 22 DoF is unnecessary (or atleast premature). The hand therapy module is a novel designthat converts rotary into linear movement and may be used totrain grasp and release, with its impedance determined by thetorque between the TOtorand a free-floating stator [14).

Of note, it remains an empirical question whether we shoulddeliver therapy to the whole arm or train individual limb seg.ments. A potential approach to increase the effectivenessbeyond Ourpast studies is to develop new whole·arm function.ally based therapy approaches that integrate robotic therapywith clinical practice and enhance the carry over of robot-trained movementS into functional tasks. Two potentialapproaches to deliver such a functional training are 1) to trainfunctional tasks with the robot or, alternatively, 2) to train byaiming for impairment reduction at the capacity level, with dif-ferent robotic modules breaking these functional tasks intocomponents, and leave the carry over of the observed impair-ment gains from robotic training into functional gains to thetherapist. These alternatives are not mutually exclusive, but wemust understand the potential and benefit of both approachesto maximize recovery and to begin building a scientific basisfor the best rehabilitation practice. For example, it appears thatthe first approach might lead to benerresults with mild strokes,whereas the second approach appears to lead to bener resultsfor patients with severe to moderate strokes [15}, [l6}, [17].The modularity of our suite of therapeutic robots, which is acornerstone of our design philosophy, is uniquely suited to chisinvestigation. It allows us to employ them in a standalonemode to deliver therapy for particular impainnents or to inte-grate the modules to deliver whole-arm functional robotic ther-apy {Figure 2(d»). To the best of our knowledge, no otherresearch group can presently pursue this objective becausethere are no otherreconfigurabJe rehabilitation robots.

The devices described earlier were intended to treat theupper extremity, and as we completed the enabling tee hnologyfor reconfigurable whole-arm therapy, we initiated th~ devel.opment of lower-extremity devices. Again. we are aiming for

Robot Therapy: StrokeVolpe et a1. reported the results of robotic training with 96consecutive inpatients admitted to Burke Rehabilitation Hos~pital who met inclusion criteria and consented to participate[19J. Inclusion criteria were diagnosis of a single unilateralstroke within four weeks of admission to the study; the abilityto understand and follow simple directions: and upper limbweakness in the hemiparetic arm (i.e., a strength grade of 3/5or less in muscle groups of the proximal arm) as assessed withthe standardized Medical Research Council banery. Patientswere randomly assigned to either an experimental or controlgroup. The sensorimotor training for the experimental groupconsisted of a set of video games in which patients wererequired to move the robot end-effector according lOthe game'sgoals. If the patient could not perform the task., the robotassisted and guided the patient's hand. Although the patientgroups were comparable on all initial clinical evaluationmeasures, the robot-trained group demonstrated significantly&'teater motor improvement (higher mean interval change±' SEM) than the control group on the trained limb segment(shoulder and elbow). In fact, patients in the robot-trained groupimproved twice as much as patients in the control group for thetrained limb segments.

In addition to the inpatient studies (see Table 1), werecruited 117 community.<fwelling volunteers in the chronicstage of stroke recovery at the Burke Medical Research insti-tute, Spaulding Rehabilitation Hospital, and Baltimore VAMedical Center. Prior to engaging in robotic therapy, thesepatients were assessed on three separate occasions to determinebaseline function and to establish a within-subject contrOl. Theprimary outcome measures were the Fugl-Meyer and theMotor Power score. Our baseline analyses revealed no statisti-cally significant differences among any of the pretreatmentclinical evaluations, indicating the stabiliry of chronic motorimpairments in this subject group. However, after robotic train-ing, we found significant reductions in motor impairment ofthe hemiparetic upper limb [20), [21].

In fact. results from many other research groups have shownthe same kind of impact. Figure 3 shows the result of a mera-analysis on upper-extrernjty robotic training trials published up toOctober 2006 with the first. generation of thenpeutic robots. Acomputerized literature search was conducted in MEDLil'JE,CINAHL, EMBASE, Cochrane Controlled Trial Register,DARE, SciS=h, Doconline, and PEDro, and it returned 173hits. Only those articles that compared robot training against acontrol group were included. Studies that compared differentforms of robotic ilierapy and stUdies on chronic stroke that com-pared discharge values with admission values were excluded. fInaddition, a significant number of hits were eliminated because of

poor design. It emphasizes again the need fora truly multidisciplinary effort. Of note,Dr. Bruce Dobkin (University of California,Los Angeles), who is the editor·in-chief ofNeurorehabilitation and Neural Repair, iden-tified this weakness and established the basicprinciples to design successful clinical trialsthat can be accessed at that journal.) Theresults demonstrated small but statisticallysignificant improvements because of therobot~assisted therapy even when comparedhead to head with conventional therapy instroke (22]. Another meta-analysis study.

maximum reconfigurability llIld have taken me same modularapproach employed for the upper extremiry. The first lower-extremity robot module deployed to the clinic has beendubbed the anklebot [9J, (18). The ankle joint is of panicularimportance because of the prevalence of drop foot, which is asimple name for a complex problem. The foot needs to clearthe ground during the swing phase of gait, and it needs to havea controlled landing during heel strike. Lack of proper controlduring these two phases increases the likelihood of Dips andfalls. At present, drop foot is typically addressed in the clinicvia an ankle-foot orthosis (AFO) that restricts the ankle'srange of motion. However, this approach has limitations andoffers linle hope of reducing the impairment. To address theimpairment reduction, we recently introduced to the clinic (atthe Center on Task-Oriented Exercise and Robotics in Neuro-logical Disease at the Baltimore Veterans AdministrationMedical Center and at the West Haven VA Medical Center) anovel ankle robot that allows normal range of motion in all3 OaF of the foot relative to the shank while walking over-ground or on a treadmill. Only 2 of these 3 OaF are actuated:plantar or dorsiflexion and inversion or eversion. As with allour devices. we purposely underactuated the ankle robot withfewer DoF than are anatomically present. Not only does thissimplify the mechanical design, it allows the device to beinstalled quickly without problems of misalignment with thepatient's joint axes.

wi!! be able to segregate science from fiction. A very commonassumption is that movement therapy works by helpingpatients relearn motor control (part of the lext in this section\\'as extracted from [181). Although intuitively sensible, thisnotion may need to be refined. First, normal mo:or learningdoes not have to contend with the neuromuscular abnormal-ities that are common sequelae of neurological injury, includping spasticity. abnormal tone, disrupted or unbalanced sensorypathways, and muscular weakness. These deficits appear toinvolve the peripheral nervous system and might suggest thatmuscles should be the focus of therapy. Nevenheless, centralnervous system plasticity appears to underlie recovery. Thus,recovery may resemble motor learning in some respects, but itis likely to be a more complex process.

Second, nonnal motor learning is far from fUlly understood.Topics of ongoing vigorous debate include questions such as:What variables or parameters of action does the brain com·mand and control? How are these encoded and represented inthe brain? How are these encodings or representations acquiredand retained? What training schedule optimizes acquisition? Isa period of consolidation between training sessions (e.g .• sleep)required for long-term retention? All these questions havepractical relevance for therapy. For example, if the brain repre-sents action as a sequence of muscle activations, focusingsensorimotor therapy on muscles would seem profitable. How-ever. a large and growing body of evidence indicates that undermany circumstances the brain does not directly control mus-cles; instead, it controls the upper limb primarily to meet kine-matic specifications (such as a simple motion of the hand in avisually relevant coordinate frame) that adjust muscle forces tocompensate for movement-by-movement variation of mechan-icalloads, which suggests that focusing on motions rather thanmuscles may be more profitable. Of course, these are only twOof a large number of possible therapy variations. In our re-search on robotic stroke rehabilitation, we have attempted toassess some of these possibilities.

In the following sections, we will describe the robust resultsin stroke and present case srodies of the effect of robotic ther-apy on other populations, in panicular, CP, MS. and SCI. Asthese studies are ongoing, we include data from representativesamples of convenience (twO subjects per condition). Finalresults including all the cohort of enroHed patients will be pub-lished in the near future in clinical journals by some of thecoauthors, Fasoli (CPl, Lo (MS), and Volpe (SCI), at the com-pletion of these ongoing studies. All protocols reported in thisarticle were approved by the Committee on the Use of HumanExperimental Subjects of MET and by the institutional reviewboard of the respective testing site, and the informed consentwas obtained from all participants.

Robotic Theropy: A Paradigm ShinThe infancy of therapeutic robotics is easily demonstrated: aMEDUNE search prior to 1990 will return no articles ontherapeutic robotics. Of course, the application of robotics torehabilitation has a longer history, but as mentioned earlier,the strong and sustained growth of activity in recent years isdue to a significant shift away from assistive technology forpeople with disabilities toward robotic therapies, which usethe technology to support and enhance clinicians' productivityand effectiveness as they try to facilitate the individual'srecovery. The magnirode of this change goes far beyond theusual ebb-and-flow of activiry in technology~related fields.Tracking the approximate number of articles submined to theInternational Conference on Rehabilitation Robotics from1997 to 2007 demonstrates a sharp increase in articles on ther-apeutic robotics, which rose from 33% to almost 80% of thesubmined articles. This is not a minor shift: robotics promisesto transform the way physical medicine is practiced.

Although the sharp increase in activities is encouraging, wemust highlight the importance of the multidisciplinary effortneeded to segregate science from fiction and push the bounda.ries of the state of the art (please note the large number ofcoauthors). Sustainable growth of activities in the area of reha-bilitation robotics will only be achieved if we engage in seriousclinical Dials to understand what works and what does not.That said, engineers must recognize the need to bring thetechnology out of the lab and into the clinic. Successful transla-tional will occur only if we engage clinicians (physicians andtherapists) and patients (and their families). Otherwise, muchof this flurry of excitement and growth will be transitory innature (regrenably, not an unprecedented pattern in robotics).

There are a multitude of variables that may influence out~come, and we must determine the interaction or independenceamong these variables and their actual impact on outcomes. Ifwe can achieve significant inroads in this investigation, we

Between~G·r~~¢~~pOris·ons: Fi~~: Robot TrainedEvaluation Minus Initial Evaluation (0 = 55)

Impairment measures (±SEM)Fvgl·Meyer shoulder/elbowMotor powerMotor stotus shoulder/elbowMoror status wrist/hand

6.7 ± 1.04.1 ±0.48.6 ± 0.84.1.± 1.1

4,5::!: 0.7

2.2 ± 0.33.8 ± 0.52.6 ± 0.8

which included studies on chronic5'l'oke that compared discharge val-ues with admission values, indicatesa similar positive impact of thetechnology [23].

Although stroke is the single larg-est cause of permanent disability inthe United States, our interests gobeyond stroke: we believe that ro-botics may prove to be applicable tomost neurological disorders and willfundamentaJly change the processof rehabilitation. To achieve this, wemust understand whether deliveringmovement therapy via robotic assis-tants will have a genuine impact onother afflictions, and we must under~stand the distinctive features of eachdisease. Here, we present our initialresults with samples of conveniencein CP, MS, and SO.

movements in a temperanll'e-conrrolled environment. Thisprotocol included twice-a-week training sessions for a total of12 sessions of approximately 45 minutes duration per session.Outcomes were measured a[ baseline and at discharge andincluded the timed 25-ft walk (T25FW), the 6-min walk(6rvfW), isometric strength for ankle flexion or extension andinversion or eversion, muscular fatigue as tested with sustainedhold, and accuracy. Here, accuracy was defined as the scorepatients achieved in the driving simulator training game.Patients were asked to move their ankle unassisted in plantar-dorsiflexion or inversion or eversion and pass the vehiclethrough incoming randomly positioned gates. For every success-ful maneuver through a gate, the patient scored +10 points. Forevery collision with the walls, patient lost -10 points. Maxi-mum score was 800. In addition, kinematic data were collectedby the Ankleoot (see Figure 4).

Of particular interest in this pilot study is the impressivechange in torque production at the ankle and the improvementin movement accuracy (tested in the S,une temperature-controlled environment) for these persons with MS after the 12training sessions. The training protocol does not include agait-specific task. Nevertheless. we observed a cany over tocharacteristics of gait with a general improvement in the dis-tance covered in a 6WN and the time for a 25FW, even thoughwalking had not actually been trained.

Robot Therapy: SCIEmergency medical and surgical services have decreased mortal-ity rates after traumatic scr, which results in almost 1O,<XX:lnewsurvivors of traumatic SeI per year in the United States. Newtreaonents are desperately needed for the subacute stage. Cur-rently, methylprednisolone is used as an effective treaonent,which has been delivered in high doses within eight hours ofinjury (as a bolus and then continued for the first 24 h) {33] withother potential candidates [32J. Improved gains in function and.importantly, gains in independence have been made primarilyvia assistive devices. Nevenheless, it is possible that therapeuticrobotics may have a place in reducing impairment On the basisof preclinical information that plasticity occurring in the spinalcord is possible, we anempted to influence the functional abil-ities of patients with chronic traumatic SCI by using roboticprotocols that have been effective in patients with stroke. Wescreened 39 persons with SCI who were admined to BurkeRehabilitation Hospital during the period of 2002 to 2006. Ofthose, nine fell within the incomplete C4-C6 lesion category.We trained these SUbjects with the shoulder-and-elbow robotfor 18 sessions over six weeks with one ann, followed byanother 18 sessions of training with lhe other arm. We eval-uated these subjects using the Fugl-Meyer and Motor Powerscales, and we present here a case study of a representativesample of convenience of two ofthese subjects (Table 4).

For these patients, we recordedchanges greater than 10% in the Fugl-Meyer Assessment and 20% in the MotorPower scales. Note especially that whilewe trained one ann at a time, both limbssimUltaneously improved by comparableamounts.

Fugl-Meyer (Mean and 95% el)+=Kahn 2000 N= 10 -0.58 [-1.88-Q.71} ~ IL-

I,Hesse 2005 N = 39 3.74 (2.68-4.80J

Daly 2005 N= 12 0.11 [-1.02-1.24) -!<> _

Kahn 2006 N= 19 0.46 (-o.51-1.31}

Lumo 2006 N= 15 0.48 [~.S7-1.S3]

Robot Therapy: CPWe engaged children 4-12 yearsold having hemiplegic CP in 16robotic sessions of task-specifictraining with the MIT·MANUS.Hemiplegic CP is the most com-mon syndrome in children born atterm and is second in frequency Fig. 3. Meto-ona!ysis of robot-ossisted therapy trials on motor recovery following stroke.

only to spastic diplegia among pre-term infants [24J. Children with hemiplegia typically haveimpaired sensory mechanisms [25), decreased motor controland muscle weakness {26], and spasticity. Typical therapy forchildren with hemiplegia is often based on motor learning, inwhich varied experiences that promote task-speCific training areused. Motor learning strategies that incorporate practice, repeti-tion, and context are commonly used in the current therapypractices. Emolled children received [\1.'0 sessions of robotictherapy per week at Spaulding Rehabilitation Hospital, with anemphasis on repetitive reaching exercises for the pareticshoulder and elbow. These 16 one-hour therapy sessions, andclinical evaluations that measured changes in motor impair-ments, were administered over a 12-14-week period. Clinicalassessments included the Modified Ashwonh Scale, FugJ-Meyer Assessment (with specific scoring established for thispediatric population), and the Quality of Upper Extremity SkillsTest (QUES1).

The initial results for the first two children are shownhere (Table 2). Findings suggest that 16 h of robotic therapywith the MIT -MANUS can have a positive impact on reduc·ing tone and motor impairment in the paretic arm to theorderof7-9%.

Post-Anklebot

Pre-Anklebot

9.62563.740.4560

Average of Two Subjects

T2SFW(s).6MW(min)Ankle p1ontar1l~xion (N x m)Ankle dorsiflexion (N x m)Movement accuracy (dorsi-plantorflex)

11.50%11.30%645%

3475%39.30%

Robof·Based EvaluafionThe results discussed earlier were limitedto standard clinical scales and did nottake full advantage of the robot's capa-bilities [0 quantify patient performance.

ChangesAdmission to Modified Fugl·MeyerDischarge Ashworth Assessment" Upper Quest(ChildID) Scale (/35) .Extremity (/66) (/100)

PII~02 -2.75 4.5 7.1PII_03 -2.5 6.0 7.7

Here. we are showing the results with two of the 12children enrolled in this pilot study. Trace 01 Ankle Plantar or Dorsiflexion (MS Subject 302)

10

8

6

Trace of Ankle Plantar or DorsifleXion(MS SUbject 302)

10

8

6usual care for the chronic phases of MS. Studies employing inpa-tient rehabilitation have shown that rehabilitation in this settingcan help to restore function in relapsing-remitting patientS withnonremitting deficits [30]. However, the p:>tential role and effi-cacy of outpatient rehabilitation have not been fully explored.There are no current interventions directed at correcting ankleimpairment The current solution is to fit an AFO, which helps toensure that the foot clears the ground and also fixes the ankle andthus induces compensatory movements at the knee and hip toaccommodate this loss of articulation.

To address this problem, we are employing the AnkJebot in apilot study at the West Haven VA Medical Center. Here, wereport a case study of a sample of convenience of twO subjectswho had progressive disease and were ambulatory with caneassistance (Table 3). Both subjects had prominent drop foot andwere engaged in a protocol similar to a driving simUlator. Seatedsubjects practiced a block of 320 plantar-dorsiflexion move-ments, followed by another block of 320 inversion-eversion

coo.~

~a -2o -4

~ -6a: -8

-10o 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Tlme(s)

(b)

c00

~'1!i'i -2E -4

j -00- -8

-100 0.5 1

Robot Therapy: MSSpecific gait abnormalities have been observed in ambulatoryMS patients [27]-[29], including those with even minimalimpairment as measured by the Expanded Disability StatusScale (EDSS < 2, [27]). These gait abnormalities have gener-ally included reduced velocity. reduced stride length, increaseddouble suppolt time, and gait asymmetry. Rehabilitation for thepurpose of improving or sustaining motor function is not a pan of

1.5 2 2.5 3 3.5 4 4.5 5

Time(s)

(a)

Rehabilitation robotics can not only introduce new efficienciesinto cenain routine therapy activities but also provide a richstream of objective data to assist in patient diagnosis, custom-ization of therapy , adaptation of the way the robot is controlledduring therapy [33], assurance of patient compliance withtreatment regimens. and maintenance of patient records.Robotics can also ease the transition to fully electronic medicalrecords. Here, we will present two examples. The first exampledemonsrrates how we can characterize interlimb joint coordi-nation in stroke patients. The ability to reach appropriately foran object or 10 move objects requires proper interjoint coordi-nation. The second example demonstrates that even a simplemetric (the deviation from a straight line) can detect differen-ces between the effectiveness of surgical (implantation of deepbrain stimulators) and medical (pharmacological) therapies inPD. Thus, such robot-based evaluations can potentially add anew and additional tool to the process of deciding when to con-sider the option of surgical implantation.

Robol-Bosed Evaluation:Stroke and InferJoinl CoordlnaMonThe evaluation games employing the shoulder-and-elbowMIT ~MANUS included drawing circles, stars (point-to-pointmovements). squares. diamonds, and navigating through win-dows [34J. Some games reqUired predominantly shouldermotion, whereas others required predominantly elbow motion.Additional games required the coordination of both shoulderand elbow. For example, the axis ratio of the ellipse fined to asubject's attempt to draw a circle provides a metric of the abil-ity of subjects to coordinate interlimb joint movement. Wefound that SUbjects were able to draw better circles over thecourse of 18 sessions of the robotic therapy program. Figure 5shows the changes in the axis ratio for 16 subacute inpatientsattempting to draw circles (see [35) for OlOTedetails of thismetric and results with 117 chronic stroke volunteers). Thisresult demonstrates that thesensorimotor point-to-pointtraining appears to facil-itate coordination andgeneralize to tasks notex;plicitly trained.

Robot-BasedEvaluaMon:PO and DeepBroin StimulationDeep brain stimulation(DBS) is the most com-mon surgical procedurefor patients with PD.Although DBS has beenshown to have a positiveeffect on PO symptoms,the specific nature of itseffects on motor control isnot yet understood. Wepreviously introduced theuse of a wrist robot tostudy the effects of stimu~lation on motor perform-ance and learning [36J.Here. we present some

.SO!0.6

••c: 0.5."~ 0.4

results from PO patients with implanted stimulators, testedwith and without stimulation, and compare them to resultsfrom PD patients without stimulators. tested on and off medi-cation. The subjects perlormed point-to-point reaching move~ments with lheir wrist. We scored the movements based ontheir accuracy and compared the scores from the four condi-tions: stimulation on, stimulation off, on medication, and offmedication.

Tlle results from 12 subjects diagnosed with PD arereported here. Five subjects had implanted stimulators (meanage 74.2 ± 4.9 y) and seven subjects did nOt (mean age 69.7 ±7.6 y). The subjects with stimulators were tested both on andoff stimulation while continuing to follow their normal medi~cation regimen. The subjects with no stimulators were testedboth on and off medication. The United Parkinson's DiseaseRaring Scale (UPDRS) was used to provide a measure of theclinical severity of each subject at the time of testing. TIle clinicalstate of all subjects improved on the UPDRS when given theirtherapy, be it surgical or medical. We note that the severity ofsubjects in the DBS group was similar to that of subjects in themedication group, both on and offthcrapy (upDRS DBS-ON =28.1: DBS..QFF = 38.6; Medication-ON = 27.1; Medication-OFF = 36.4; total possible score = 108; higher scores reflectmore impaired performance). Subjects were presented with

••••.g 0.025..~~ 0.02

!3

On ON Stimulation StimulationMedication Medication On Off

Fig. 6. Mean values or lateral deviotion across experimentalgroups and conditions. Whisker bars ~epresent standarderrors. A significant difference exists bei"Ween the mean atthe subjects tested off stimulation and the other three groupmeans ('P < 0.05: •• p < 0.01).

one center target and eight peripheral ones [Figure 2(c»). Afteran initial practice set. subjects performed a set of 80 reachingmovements to a randomly selected peripheral target within atime window of 1.6 s and then moved back to the center posi-tion. We found that the subjects with stimulators turned off hadthe highest mean deviation from the straight line connectingthe center and the peripheral target (Figure 6). Vlhen stimula·tion was turned on. these subjects deviated significantly less(P < 0.05). The performance of subjects \\ithout stimulatorswas not significantly affected by medication. Subjects withstimulators turned off deviated significantly more than subjectswithout stimulators. whether on or off medication (P < 0.01).

Chonges fromAdmisS~on to DIscharge

Umb;Umb2

FUg,7r~';e:y!'tr Motor PowerAsse'sSment"U66) u70)

6.66 13:8'9.33 15.5

Here. we ore showing the results with· two of the ninesubjects enrolled in this pilot.

ConclusionsIn this anic1e, we presented an overview of the remarkablegrowth in the activities in the area of therapeutic robotics andof ex;periences with our devices. We presented mounting evi-dence that robotics can be used as a general tool to harnessbrain plasticity and promote recovery, and this improvement,at least for stroke, is on average sustainable in the long ron forboth subacute and chronic cases (see (37) for subacute and [20]for chronic stroke). We emphasized that the same kind oftechnology may have a broader impact beyond stroke therapy,with the preliminary evidence of its value for CP, MS, and SCIbeing quite promising, even though these diseases impair verydifferent parts of the central nervous system. Nonetheless •significant challenges face us in the nex;t five to ten years. Erst,we must determine, among the multitude of variables that mayinfluence outcome, the level of interaction or independencebetween these variables and their actual impact on outcomes.Second. movement therapy in general might only do so much,and we need to investigate the potential impact of the combina-tion of different modalities of therapy (e.g .• robotics, pharma-cological, electrical stimulation). Only then we will besuccessful in the quest to optimize therapy to meet particularpatient needs. We must pursue this challe:lge in a truly multi·disciplinary fashion armed with tools that nOt only facilitateand augment therapy delivery but also augment present clinical

I evaluation scales with robot-based evaluation tools.

Fig. 5. Axis rotlos tOt odmission ond discharge for eoch subject. Subjects were sorted accordingto the value of oxis rotio at admission (subject labels have been omitted for clarlty). Bigger posi·tive changes from admission to diSCharge correspond to subJecll with lower axis ratios at admis·sian. Note that an axis ratio equal to one indicates a perfect circle .• indicates stotlsticaltysignitlcant change from admission to discharge (P < 0.05).

AcknOWledgmentsThis work is supported by the National Institute of ChildHealth and Human DevelopmentlNational Center for MedicalRehabilitation Research (N1CHDINCMRR) grant I ROI-HD045343; the VA Veterans Affairs grants B3688R andB3607R; and the NYSCORE. S. Levy- Tzcdek is a HowardHughes Medical Instirute predocloral fellow. A.c. La andJ.A. Fawcett are supported by grants from the Depanment ofVeterans Affairs Rehabilitation Research and DevelopmentService (B4145K and B54031). S.E. Fasoli was supported bythe Charles H. Hood Foundation. H. Poizner was supportedby the National Institutes of Health (NIH) grants 2 R56NS036449-~AI and 7 ROI NS036449 and National Sciencefoundation (NSF) grant SBE-0542013. H.I. Krebs andN. Hogan are coinventors in the MIT-held patent for therobotic device used to treat patients in this work. They holdequity positions in Interactive Motion Technologies. thecompany that manufactures this type of technology underlicense to MIT.

Hermano J, Krebs has been a principalresearch scientist and lecturer at MIT'sMechanical Engineering Departmentsince 1997. He also holds affiliate posi-tions as an adjunct research professor ofneuroscience at Weill Medical College ofCornell University and adjunct professorof neurology at the University of Mary~

land School of Medicine. He is one of the founders of Inter~active Motion Technologies. a Cambridge-based start-upcompany commercializ.ing robot technology for rehabilita-tion. He received his electrician degree in 1976 from EscolaTecnica Federal de Sao Paulo, Brazil, and the B.S. and M.S.degrees in naval engineering (option electrical) from theUniversity of Sao Paulo, Brazil. in 1980 and 1987, respec~tively. He received another M.S. degree in ocean engineer-ing from Yokohama National University, Japan, in 1989,and the Ph.D. degree from MIT in 1997. His overarchinggoal is to pioneer life-improving and cost~saving rehabilita~tion technologies beneficial to patients, clinicians. and thehealth care system.

••••......._~.;:.. Laura Dipietro received the Laurea degree. (summa cum laude) in electrical engineer~

. ).'. : ing (biomedical curriculum) from the Uni-.j i~ versity of Florence. Florence, Italy, in 1998.. '-ie:~·-" and the Ph.D. de~e in biomedica~ roboticsr '.'~, from Scuola Supenore S. Anna. Pisa. Italy.

, , in 2003. She is currently a postdoctoralassociate at the Newman Laboratory. MIT.

Cambridge. Her current research interests include robot.assisted neurorehabilitation. motor control and learning, andbiomedical signal processing.

Shelly Levy· Tzedek is currently a Ph.D.candidate at the Newman Laboratory forBiomechanics and Human Rehabilitationat MIT. She is a Howard Hughes MedicalInstitute predoctoral fellow. She receivedher B.S. degree summa cum laude in bio-engineering from the University of Cali~fornia in Berkeley in 2002, where she was

j ajep:ntaltior. rea Meg:xelSci6ouonncut Statein benpn~ M1OOl:heUni in ~ShlS a h a:ant/exer-j thMTIideiello cisologthe.erclCltorom 2002

to 21e frings, sled research

~

anaSO~letIOCtllsscnd :a1 1inaYaTniversityo lawn tlvmaorySchMe( anA. Cth ( System,

~>mecs lJm~litaWen.'MJ'DOSlor~ass.: thctivof uorOlehe ediofficer of:rappenith lis paabilHos.l and theke,emi result tet! thrtmof Physi-abiln ponm p al indilita at Har-

andanfe platicrl idwa ard Scle lined hislot-sistcrapimpunc of . .A: C< Urrsity andarnShentl~lucburppo , .Qlm en }tein Col-h p(ectexaJ thtts ahe egc1e. \Jleresidencydrewitbiplcue tlr stcaJtraiJphjmercruon Dlumbia-rofeionS. d in ctian anPreln Ml Cholopoient as anlege inapestud Sdegcassaraf(in merysiMedicineandehatiomcemoity.andiIiuat :ediool. medicaln inuc;>pont ovardSclprac de' to nd :tatiof stroke'ideher beflreStndoraclSuT"HislI'cl'td (be aereise asl mobeue kan Ass Sna trc. aftokes a 'aT iest in thelitatn ~mn in n airs use lOV. tee to ~e ceise and!itatn T:>m,:.Fa thlO CrecC'tere.: Rabilil Mle.

- .,;.;-Ioveri1rchfessor in- e )Nempl,on at the

, niWihn go, and afeusccient Rutgm

nhr to Rus Univer-. iry,rt a! diPr of the

ogiOscabory at theerewashaixrad ;alkor hi ~ies in La.ss dectedilkevollfonie i?: re of Rutgers

Shhas a s-Ievf ocl thUniBofTr:llcte~ Award.Bur: Rbtaios, 'Ws, lHisch :stsloticysi! patients;inc2OQth e.l S'jZatirolewiti' din ain cs I underlied cdio/:>nalctice 2as OnOfro!.1insomed Gltio:rapons

fovanagt palOS. bellbiS a:ant fessor inrolticsarclrdirt th\1e e It of.og)the War-

:h Ittitute Jy 2( , n edifloof Brown, - ni'S reinte focuses

'~'. nlpsluat Bo '. ,; n al h I clinicale:witlA. log. sects d aeventing.; cwwacli isat hto ore func·

ar60ci thlY1e iOOirolinj\ such asealStitJdyfecttho:to SCInd lic In injury.s o>tic y oItrerHe ed ).An bistirom the

:~ovelowoke Uni offonc:clcthelceived a.-t c.Psreen Hiveld sequently

ean Mnd reeUOSlee from

Inil.It jl Athe FoIl'ni'ooldiciin Nonhset' 1~ wz) acCaIHe Iletms~ne1.Dgy resi-y i12ar.ed lergden{alc is :rtitlrolit by the. tCO.arg patAmBof Pmd og)C is cur~

inon vacltiorrenl.stmairA ~tivcudy 558. Kl Anred, CrVJtic,mu, cliial etermine,ha,lait.ret1R.esl$afeeffJ of f lITe r( devicestuS003seN 4<deso aiumne\J)i!it,n in per-

t Slpcmpital 2he Ihe son:hruro}