Combined Tendon Vibration and Virtual Realityfor Post-Stroke Hand Rehabilitation

Mike D. Rinderknecht1,2*, Yeongmi Kim1†, Laura Santos-Carreras1,2‡, Hannes Bleuler2§, and Roger Gassert1¶

1 Rehabilitation Engineering Lab, ETH Zurich, Zurich, Switzerland2 Laboratoire de Systemes Robotiques, Ecole Polytechnique Federale de Lausanne (EPFL), Lausanne, Switzerland

ABSTRACT

Sensory function is essential for functional post-stroke recovery andcontrol of basic hand movements like grasping. Despite this fact,therapy focuses strongly on motor aspects of rehabilitation, requir-ing active participation and thus excluding stroke patients with se-vere paresis. The aim of our novel therapeutic approach combiningvirtual reality, based on clinically proven mirror therapy, and ten-don vibration of hand and wrist muscles is to induce neuroplasticchanges leading to improved hand function. This paper presents thefurther development and evaluation of a robotic device, which canapply vibrations at precise locations on the finger flexor tendonsto create illusions of extension movements and visualize the move-ments with a virtual hand. A preliminary study including 16 healthysubjects investigated the influence of the virtual reality on the per-ception of proprioceptive illusory movements. The experimentalresults provided evidence that the addition of the virtual reality en-hanced the perception of the illusory movement generated by ten-don vibration, by inducing movements with significantly higher ex-tension (+4.5%, p< 0.05). Furthermore, the virtual reality alloweda better controlled temporal elicitation of the illusion. These find-ings indicate the potential of this novel strategy for a more effectivetherapy, especially for severely impaired patients.

Keywords: Haptic device design, tendon vibration, virtual reality,kinesthesia, proprioception, upper limb, rehabilitation, stroke.

Index Terms: Human-centered computing [Human computerinteraction (HCI)]: Interaction devices—Haptic devices; Human-centered computing [Human computer interaction (HCI)]: Interac-tion paradigms—Virtual reality

1 INTRODUCTION

Ten million people survive a stroke each year worldwide and re-quire rehabilitation [1]. With a prevalence of 75%, upper limbparesis is very common after stroke [2]. Of these patients, two outof three do not recover fully and have persisting upper limb pare-sis [3, 4]. Additionally to motor deficits, sensory dysfunction hasa strong impact on the patient’s autonomy in daily life, since es-pecially the hand function relies on multiple sensory inputs. As aconsequence, activities of daily living, such as grasping a cup ofcoffee and buttoning a shirt, become extremely difficult or impossi-ble. Furthermore, several studies have revealed that learning of newmotor skills is impeded by post-stroke sensory dysfunction, leadingto poor functional recovery, as this relies on feedback from sensoryinputs [5–8]. Diverse rehabilitation methods devote effort to motor

*e-mail: mike.rinderknecht@hest.ethz.ch†e-mail: ykim@ethz.ch‡e-mail: laura.santoscarreras@epfl.ch§e-mail: hannes.bleuler@epfl.ch¶e-mail: gassertr@ethz.ch

rehabilitation to help stroke patients regain function of their mildlyparetic limb. This process can be supported by robotic devices [9].As the tools to assess and treat sensory deficits are rather crude,robotic devices have also been specifically proposed for the synthe-sis of sensory stimuli in a controlled and repeatable manner, bothfor assessment and therapy of hand sensory function [10].

Motor imagery offers an alternative to active motor rehabilita-tion. Evidence exists that even mental practice of a movement haspositive effects on motor function recovery in chronic stroke pa-tients [11] and improves motor performance in healthy subjects al-most as successfully as physical practice [12]. Research findings in-dicate that motor imagery recruits sensorimotor networks but withreduced activation compared to motor execution [13].

Mirror therapy is one of the current sensorimotor rehabilitationmethods building upon the previous concept, additionally providingvisual feedback of the movement [14,15]. To create the illusion thatthe impaired hand is moving, the stroke patient moves the healthyhand in front of a vertical parasagittal mirror, while the impairedhand is hidden behind the mirror. Through activation of various re-gions in the affected brain hemisphere, cortical reorganization takesplace [16]. A study by Dohle et al. [17] showed that, in a groupof patients with distal plegia at the beginning of the therapy, thosereceiving mirror therapy regained more distal function than the pa-tients of the control group receiving equivalent therapy but withoutmirror, thus seeing the affected limb. Furthermore, mirror therapyimproved recovery of surface sensibility across all patients. A ran-domized controlled trial reported an improvement of hand function-ing after 4 weeks and at 6-month follow-up in subacute stroke pa-tients after mirror therapy in addition to conventional rehabilitationcompared with a control treatment [18]. These studies have shownsome effectiveness of mirror therapy and that it might be a promis-ing method to reduce sensory deficits to support motor recovery.

Another way to induce such illusory movements is offered bytendon vibration, discovered in the early 1970s [19–21]. Thesestudies conclude that, by vibrating a muscle tendon of a staticlimb in the range of 50 to 100 Hz, a proprioceptive illusion ofa movement corresponding to the movement generated by a con-traction of the antagonist muscle can be elicited [22–25]. It wasshown that the strongest illusions can be created by using a vibra-tion frequency of 80 Hz, approximately [23], and that the illusionsinduced by tendon vibration correlated with antagonist motor re-sponses of a given movement [26, 27]. Thus, tendon vibration of-fers the possibility of inducing multi-dimensional illusory move-ments, as already done e.g. on the ankle or wrist [28]. Researchgroups started to investigate the possible use of tendon vibration fortherapy of sensory-motor deficits for proprioceptive and movementre-education [29–31]. Most studies applied vibration to the wrist,elbow or ankle tendons due to the bulky design of the devices (e.g.VB115 from TechnoConcept, Mane, France or the low-cost deviceproposed by Celik et al. [32]). Only one group applied vibrationsto the hand and finger tendons using a customized single-site vi-brator [30]. The latter study conducted in 11 patients with complexregional pain syndrome I showed that the intervention group (addi-tional tendon vibration) had a 30% range of motion gain and about

277

IEEE World Haptics Conference 201314-18 April, Daejeon, Korea978-1-4799-0088-6/13/$31.00 ©2013 IEEE

Figure 1: The Picidae Robot, a novel robotic device combining ten-don vibration and virtual reality for neurorehabilitation of hand func-tion. The robot’s name originates from the Latin for “woodpecker”,known for their pecking behavior similar to the vibrations producedby the device.

50% lower pain severity than the control group. However, to thepresent day, this method is not applied in clinical neurorehabilita-tion. Thyrion et al. [33] found that, like motor imagery, tendon vi-bration activates cortical sensory and motor areas as well as parietalregions. They further postulate the existence of strong analogiesbetween motor imagery and proprioceptive illusions and proposeto combine the two processes for sensorimotor rehabilitation. Theconcept of assisting mirror therapy by tendon vibration and viceversa for therapy of stroke patients is relatively new and there areno published results yet from the ongoing clinical trial [34].

The study described in this paper aimed to go one step further bycreating a novel neurorehabilitation strategy for severely impairedstroke patients suffering from upper limb paresis by combining ten-don vibration with virtual reality (VR) to gain more control overand thus augment the mirror therapy to support the positive effectsof motor imagery reported in the literature. This novel method us-ing a virtual visualization of the hand and tendon vibration providesmultisensory feedback (i.e. visual and proprioceptive) and createsa complete illusion in contrast to mirror therapy where the patientneeds to consciously move the unimpaired hand to see the impairedhand moving. In the absence of the proprioceptive cues comingfrom the healthy limb, we hypothesise that a stronger illusion willbe created, leading to a more effective therapy inducing and sup-porting plastic processes in affected brain regions. Furthermore, theVR offers a wide range of applications and interactions, e.g. frombasic gestures like grasping to motivational rehabilitation games.This may help stroke patients with severe motor deficits, who can-not move their hand and for whom many therapies are therefore notyet an option, to initiate motor rehabilitation.

This paper presents a the further development and initial evalu-ation of the novel robotic device, the Picidae Robot (Fig. 1). Thepresented device can apply precisely localized and independent vi-bration on the flexor tendons of the five fingers of the left hand [35]and visualize the perceived illusory movement using a virtual handon a touchscreen located directly above the hand receiving stimula-tion. Results of a preliminary study on a group of healthy subjectsto assess the influence of virtual reality on the perception of propri-oceptive illusory movements induced by tendon vibration are alsopresented.

F2 F3

F4

F1

F5

(a)

vibrator head F1vibrator head F1

vibrator head F2vibrator head F2

illusory movementillusory movement

(b)

Figure 2: (a) Palmar view of the left hand and indication of the five vi-brator locations on the finger flexor tendons. (b) Left hand resting onthe hemispheric hand support with visible vibrator heads for locationsF1 and F2.

2 METHODS

2.1 Apparatus

2.1.1 Mechanical Design

In order to induce illusory finger extension movements, the optimallocations for tendon vibration were previously identified and vibra-tion modules were developed [35]. Each vibration module consistsof a reduced slider-crank mechanism (i.e. no slider, connecting rodis the output) actuated by a cost-effective DC motor RK-370CA-15370 (Mabuchi Technologies Inc, Matsudo, Japan) and a vibratorhead. This type of mechanical design assures a constant vibrationamplitude of 1 mm. The current version of the Picidae Robot in-cludes five vibrator modules (F1–5), each of them applying vibra-tion to one of the five finger flexor tendons (Fig. 2a) of the lefthand. The hemispheric hand support was designed considering er-gonomic requirements of stroke patients defined by their hand spas-ticity. Thus, the subject’s hand is naturally and properly alignedwith respect to the vibrator heads (Fig. 2b).

2.1.2 Electronics and Control

To control the DC motors in a closed loop and thus assure theappropriate vibration frequency, reflective infrared (IR) sensorsTCRT1000 (Vishay, Malvern, PA, USA) measure the position ofthe vibrators in the axial direction. The DC motors, as well as thesensors, are driven by customized modular electronic circuit boardsand powered by a switching power supply TXL 150-24S (TRACOELECTRONIC AG, Zurich, Switzerland). To ensure safety for thesubject, an emergency stop button has been included (see Fig. 1).The communication to an all-in-one touchscreen PC (Shuttle Bare-bone, Intel Atom D510 1.66 GHz, 4 GB RAM, Windows 7 Enter-prise) running LabVIEW (National Instruments, Austin, TX, USA)is ensured by two USB data acquisition (DAQ) cards: (i) for themotors the USB-4301 (Measurement Computing Corp., Norton,MA, USA) and (ii) for the sensors the USB-6008 (National Instru-ments). The USB-4301 can generate independent pulse width mod-ulations (PWMs) to control the speed of the motors (i.e. vibrationfrequency). However, the PWM is interrupted for 0.1 s, approx-imately, whenever the duty cycle is changed, due to the internaldesign of the DAQ card. To avoid too many interruptions and un-smooth behavior, an acceptable compromise for this application isto control the duty cycle with an interval of 1 s, as the precisionof the tendon vibration frequency is not very delicate. Further-more, a range of ±2 Hz is accepted for accuracy, also reducingthe number of interruptions. The controller is composed of an inte-gral frequency control (KI = 0.005) plus a feed-forward commandfollowing the simple law:

U =RMkM

+2π f kM D =U

24V+KI ∑e f (1)

278

where U is the voltage, R = 26 Ω the motor resistance, M =0.003 Nm the estimated torque, kM = 0.01426 Nm/A the motorconstant and f the set point frequency in Hz. The voltage U isthen coerced to [0,24] V. The PWM duty cycle D (coerced to [0,1])includes the rescaled voltage and the integral term with e f , the fre-quency error defined by the difference between the measured anddesired vibration frequency. For every change of the set point fre-quency the integral term is disabled and reset during two periods,resulting in the motor immediately reaching the approximate setpoint, using the feed-forward model. Afterwards, the integral con-trol is switched on again to eliminate residual steady-state error.The addition of a proportional control (KP) resulted in increasedoscillations after reaching the set point and was thus omitted. Thefrequency of the vibrator modules can be controlled up to a the-oretical limit of 200 Hz, however, only 0 to 120 Hz are requiredfor tendon vibration. The output signal of the IR sensors is readcontinuously at 1.25 kHz by the DAQ card USB-6008. In orderto transform the sensor signal to the instantaneous vibration fre-quency, every 200 ms a new data packet of 250 samples is appendedto the previous one before extracting the single tone (i.e. main fre-quency with highest amplitude). This technique provides a morerobust frequency extraction at a reasonable iteration period.

2.1.3 Graphical User Interface and Virtual Reality

The software and interface are separated into two programs, bothrunning on the all-in-one touchscreen PC. The graphical user inter-face (GUI) implemented in LabVIEW offers the possibility to col-lect various feedback of the subject on the right half of the screen(Fig. 3). This includes in a first place real-time feedback: a but-ton to acknowledge the start of the illusion, the time difference be-tween vibration start and illusion start is referred to as latency. Thesecond part consists of the following consecutive post-trial feed-back: (i) numeric fields to report the vividness of the perceivedillusion on a free scale, (ii) a questionnaire asking how well theVR corresponded to the perceived illusion (five choices from “verybadly” to “very well”) and (iii) sliders to manipulate a small 3Dhand model to report the perceived maximal extension/flexion ofthe vibrated fingers. The left half of the screen shows the VR ap-plication (Fig. 3), which is programmed in C++ using the IrrlichtEngine [36], displays a realistic 3D model of a hand mimickingthe illusory movement. The finger angles are defined in percent-ages (maximum flexion: −100%, and maximum hyperextension:+100%), based on the range of motion values reported in [37].Since it is even possible to perceive limb positions beyond the natu-ral range of motion when applying tendon vibration [38], hyperex-tension was set as the upper limit instead of extension. The move-ment resembles a minimum jerk trajectory with a smooth start andstop. For instance, a movement from a natural resting position (0%)to an extension of 75% lasts 6 s, approximately. The VR hand isunidirectionally controlled by LabVIEW through a user datagramprotocol (UDP) connection. The touchscreen is mounted on top ofthe robotic device, so that the VR hand is exactly superimposed onthe left hand of the subject creating the illusion that it is the sub-ject’s own hand. Furthermore, the VR hand is shown grasping aball simulating the hemispheric hand support of the device in or-der to match the tactile cues. The VR can be switched on and offaccording to the experimental protocol. During the experiment thesubjects can give feedback on the touchscreen, which is facilitatedby large buttons, by using their the right hand.

2.2 Subjects

Sixteen healthy subjects (S1–16, 2 female and 14 male, mean ±SD: 26.4± 2.7 years, 14 right-handed and 2 left-handed) partici-pated in this preliminary study (see demographic characteristics inTable 1). The study was approved by the institutional ethics com-mittee of the ETH Zurich (EK 2012-N-17). All the subjects were

Figure 3: Screenshot of the touchscreen showing the VR on theleft and the feedback GUI for the perceived maximal finger exten-sion/flexion on the right.

familiarization test

experimentdecision

• no VR• 1x / finger

• no VR• 3x / finger

• no VR• 5x / finger

• VR• 5x / finger

• VR• 5x / finger

• no VR• 5x / finger

A

B2nd trial

1s15s

3s unrestrained

latency vividness corresp. extensionreal-time post-trial

exclusion

inclusion

Figure 4: Schematic representation of the experimental protocol.The upper part shows the overall experimental protocol. Within eachblock, the details are shown (i.e. if the VR is shown or not and thenumber of trials per finger). According to the outcome of the testtrials, the subject is included or excluded from the experiment (lasttwo blocks) based on a threshold. The lower part (gray box) illus-trates the details of one specific trial n (top: feedback type, bottom:vibration pattern). The saw tooth line represents the vibration period.

recruited among students and staff of the university. Exclusion cri-teria were sensory deficits in the upper limbs, any history of neuro-logical or hand injury, discomfort or pain during finger movementor applied tendon vibration. Furthermore, subjects with too largehands to fit the device appropriately were also excluded. Informedconsent was given by all the subjects before participating in thestudy.

2.3 Experimental ProtocolSubjects sat comfortably on an adjustable chair in front of the Pi-cidae Robot. To allow a good view on the touchscreen and showthe VR hand superimposed on the subject’s hand, the height of thechair was adjusted. The left hand of the subject was placed on thehemisphere of the device below the touchscreen and adjusted in or-der to align the optimal vibration locations with the vibrator headsas properly as possible.

The experiment consisted of three successive blocks: familiar-ization, test and actual experiment (see Fig. 4). Each block in-cluded a total of N trials (familiarization: 5, test: 15, experiment:50) where the vibration stimuli of 80 Hz during 15 s was presentedseparately at the five fingers (F1–5) in a pseudorandom order. Thelower part of Fig. 4 shows the details of one trial with the variousfeedback and timing.

The familiarization period allowed the subjects to get used to the

279

stimulus, device and the onscreen evaluation. The second blockaimed at excluding subjects having problems with the perceptionof the illusions. The inclusion criteria for the third block, the ac-tual experiment, were (i) a latency ≤ 10 s to allow the illusion to beelicited early enough before the vibration ends and (ii) a perceivedextension ≥ 20%, both in at least 75% of the test trials. Theseempirical values were based on the outcomes of preliminary ex-periments. In case of failure, the subject received a second chancefor the test. After a second failure, the subject was definitively ex-cluded. During familiarization and test no VR was shown and theVR correspondence feedback was thus not evaluated. The actualexperiment consisted of two parts: 25 trials without VR and thesame number of trials with VR. To avoid bias, the order of no VRand VR conditions was randomly assigned to each subject (A andB in Fig. 4). The parameters of the VR, i.e. extension of 75% andVR movement latency of 3 s, were established experimentally in apreliminary study.

To prevent subjects from hearing the vibration noise, they worehearing protection earmuffs during the complete session. The totalduration of the session ranged from one to one and a half hours.

2.4 Data AnalysisStatistical analysis was performed only on data from the actual ex-periment (block three). In order to determine the influence of theVR on the latency, the absolute difference of the illusion latencyprovided by the user’s real-time feedback L and the VR latency LV Rwas computed: ∆L = |L−LV R|, referred to as latency deviation andmeasured in seconds. As recommended for psychophysical experi-ments of magnitude estimation [39] and as previously done in [35],the vividness of the illusory movement was rated on a free (self-determined) scale that was then linearly normalized (Vrescaled) foreach subject to [0,1], whereby 0 corresponds to “no illusion per-ceived”. Furthermore, the geometric mean Vgeom and standard er-ror SEV was calculated for each subject according to the followingequation:

Vgeom = 10

1nV 6=0

∑V 6=0

log10 Vrescaled

SEV =

sV√nV 6=0(2)

where sV is the sample standard deviation and nV 6=0 is the numberof trials where an illusion was perceived. Since all data followeda normal distribution, paired t-tests on the subject means were per-formed to test for significant differences of the latency deviation ∆L,vividness Vgeom and perceived extension E% between the two con-ditions no VR versus VR. Significance levels were set to α = 0.05.The VR correspondence feedback was attributed a numerical valueranging from 0–4 as follows: “very badly” (0), “badly” (1), “moder-ately” (2), “well” (3) and “very well” (4). To assess the correlationbetween latency deviation, vividness, perceived extension and VRcorrespondence, Pearson’s correlation tests were run. The VR cor-respondence for the no VR condition is not applicable and this itwas excluded.

3 RESULTS

Only 2 subjects from the test group of 16 failed the test block twiceand were excluded from the actual experiment. Subject S4 failedonce but fulfilled the inclusion criteria on the second trial. Table 1summarizes the demographic subject characteristics, the test resultsand the inclusion/exclusion decision. In total, 87.5% (14/16) of thesubjects participated in the actual experiment with the two condi-tions no VR and VR.

In 99.1% of all the trials of the actual experiment (no VR:347/350, VR: 347/350) the illusions were perceived in accordanceto theory, i.e. finger extension. Two trials failed in the no VR con-dition, because the feedback was not provided successfully by the

Table 1: Demographic characteristics of the subjects, results of thetest block (passed: 3, failed: 5, not applicable: –) as well as inclusiondecision and group attribution.

subject gender age dominant hand test 1 test 2 inclusion / group

S1 M 26 right 3 – yes / AS2 M 29 right 3 – yes / BS3 M 25 right 3 – yes / AS4 M 30 right 5 3 yes / BS5 M 28 right 3 – yes / AS6 M 31 right 3 – yes / BS7 M 25 right 3 – yes / AS8 M 22 right 5 5 no / –S9 F 25 right 3 – yes / AS10 F 24 right 3 – yes / BS11 M 25 right 3 – yes / AS12 M 22 left 5 5 no / –S13 M 27 left 3 – yes / BS14 M 30 right 3 – yes / BS15 M 26 right 3 – yes / AS16 M 27 right 3 – yes / B

Table 2: Pearson’s correlation matrix summarizing the coefficient rfor all the combinations. ∗∗ p < 0.001 (2-tailed).

VR no VR

latency deviation - extension -0.350∗∗ -0.408∗∗

latency deviation - vividness -0.312∗∗ -0.408∗∗

extension - vividness 0.415∗∗ 0.495∗∗

latency deviation - VR correspondence -0.288∗∗ –extension - VR correspondence 0.328∗∗ –vividness - VR correspondence 0.429∗∗ –

subject (i.e. total of 345/350 valid trials for no VR). The failed trialswere excluded from further statistical analysis.

The paired t-test on the group means of the latency deviationshowed no significant difference (t(13) = 1.44, p = 0.17). How-ever, the total sample mean ± standard error in the VR condi-tion (2.188± 0.254 s) is smaller than in the no VR condition,where they are 2.435± 0.278 s. The difference in perceived ex-tension was found to be significant (t(13) = 2.44, p < 0.05) with50.542±3.747% and 55.049±4.192% for no VR and VR. Thus, theperceived extension is 4.507% higher when tendon vibration is sup-ported by VR. No significant difference was found for the vividness(t(13) = 1.06, p = 0.31, no VR: 0.683±0.038, VR: 0.702±0.039).The overall mean ± standard error of the VR correspondence re-sulted in 2.209±0.129, corresponding to a rating between “moder-ately” and “well”. Subject specific and group means and standarderrors of all the parameters are shown in Fig. 5 for the two condi-tions (except for VR correspondence, which applies only for VR).Differences between groups A and B were not significant.

The Pearson correlation showed negative correlations betweenlatency deviation and perceived extension, as well as latency devia-tion and vividness, which were both found to be moderate in the VRcondition and strong in the no VR condition. Additionally, a weak,negative correlation was assessed between the latency deviation andthe VR correspondence. A strong, positive correlation exists be-tween the pairs extension - vividness (VR and no VR conditions)and vividness - VR correspondence (VR condition). Furthermore,a moderate, positive correlation was found between extension andVR correspondence. The correlation coefficients r are summarizedin Table 2 for n = 347 in VR and n = 345 in no VR. All correlationswere statistically significant (p < 0.001).

280

subjects

vivi

dnes

s [−

]

group0

0.2

0.4

0.6

0.8

1

no VRVR

subjects

VR

cor

resp

onde

nce

[−]

group0

1

2

3

4

very badly

badly

moderately

well

very well

subjects

exte

nsio

n [%

]

group0

20

40

60

80

100

no VRVR

*

subjects

late

ncy

devi

atio

n [s

]

1 2 3 4 5 6 7 9 10 11 13 14 15 16 group0

1

2

3

4

5

no VRVR

1 2 3 4 5 6 7 9 10 11 13 14 15 16

1 2 3 4 5 6 7 9 10 11 13 14 15 16

1 2 3 4 5 6 7 9 10 11 13 14 15 16

Figure 5: Top to bottom: latency deviation, perceived extension,vividness and VR correspondence for each subject and the group(mean ± standard error). The perceived extension shows a signifi-cant difference (∗ p < 0.05) between the two conditions.

4 DISCUSSION AND CONCLUSION

This paper presented the further development and a preliminarystudy using the Picidae Robot, the first device for a novel neurore-habilitation strategy based on illusory finger movements elicited bytendon vibration and reinforced by VR. This device is able to ap-ply independent vibration stimuli on precise locations of the fingerflexor tendons, creating the illusion of finger extension. Further-more, the illusory movement is visualized by a virtual hand on atouchscreen with the aim of reinforcing the illusion. The results ofthe preliminary study including 16 healthy subjects provided evi-dence that the addition of the VR enhanced the perception of theillusory movement generated by tendon vibration.

As expected, significantly larger illusory movements (i.e. higherperceived extension) could be achieved with this combined ap-proach when compared to those induced with proprioceptive stim-uli alone. This is well in accordance with earlier findings, wherea visual movement feedback through a mirror biased the perceivedposition induced by tendon vibration [40]. This study showed that,when the mirror image and proprioceptive illusions were in thesame direction, they reinforced each other. A previous study withhealthy subjects and stroke patients showed that the rates of ampli-tude increment and latency decrement of motor evoked potentials(MEPs) were higher when using a virtual mirror compared to a realmirror [41]. It is suggested that the reason for this MEP enhance-ment could be the more interactive and task oriented characteristicsof the virtual mirror, leading to increased attention. This may alsosupport the use of the virtual mirror paradigm (i.e. VR) for a more

effective therapy.Given that vision of a static hand attenuates proprioceptive illu-

sions [42], we hypothesized that a VR hand starting to move 3 safter the vibration onset would help to suppress an eventual early,imprecise onset of the proprioceptive illusion and trigger the illu-sion at this specific moment in time, thus allowing a more controlledelicitation of the illusion. Indeed, the results indicated that the la-tency deviation of the illusion onset was decreased when addingVR. However, the difference was not found to be significant, whichmight arise from a too large temporal mismatch in some trials.

The correlation analysis suggested that a better synchronizationof the proprioceptive illusion and the VR lead to a perception of ahigher extension and a stronger vividness. Furthermore, the vivid-ness appeared to be strongly correlated to the extension. One rea-son could be that the psychophysical evaluation of the vividnessmight also partially include the perceived extension. Especially thestrong correlation between the VR correspondence and the vivid-ness showed that the match between the proprioceptive illusionand the visual feedback is very important to obtain a vivid illusionwhich could result in body ownership. Slater et al. [43] showed thatillusory ownership of a virtual body can be produced if synchronousvisual, motor and proprioceptive stimulation is applied.

Since especially in terms of latency the inter- and intra-subjectvariability remains, more research is needed to assess the cause.Future work will also include the investigation of new ways to syn-chronize the virtual reality with the proprioceptive illusory move-ments in order to decrease temporal mismatch and elicit an evenmore vivid illusion. Leonardis et al. [44] proposed to use a braincomputer interface in order to control the tendon vibration and thevisual feedback through brain activity. Additional experiments witha larger, age-matched healthy group will allow us to collect base-line data for future clinical studies with stroke patients. Stroke is theleading cause of care dependency and the most expensive diseasein industrial countries [45] which often leads to complete paralysisof the hand, as well as sensory deficits. Thus, the potential of thePicidae Robot for a cost-effective rehabilitation system providingan effective therapy is even more vital. This project might lead toa novel strategy for early neurorehabilitation widely employed inclinics, addressing the recovery of sensorimotor deficits in severelyimpaired stroke patients.

ACKNOWLEDGEMENTS

Grateful thanks go to W. Popp, J. Duenas, J. C. Metzger,A. Moser and R. Safaı-Naeeni for interesting and profitable dis-cussions as well as technical advice, P. Wespe for manufacturing,J. D. Rinderknecht for medical advice and the subjects for partic-ipating in the study. M. D. Rinderknecht gratefully acknowledgesthe support of the IEEE Life Member Fund through the 1st prizeof the IEEE Region 8 Student Paper Contest [35] and the Region8 Student Activities Fund through the “Dick Poorvliet Award”.R. Gassert is supported by the National Center of Competence inResearch on Neural Plasticity and Repair of the Swiss National Sci-ence Foundation.

REFERENCES

[1] World Health Organization, “The world health report 2002 - ReducingRisks, Promoting Healthy Life,” 2002.

[2] S. S. Rathore, A. R. Hinn, L. S. Cooper, H. A. Tyroler, and W. D.Rosamond, “Characterization of incident stroke signs and symptoms:findings from the atherosclerosis risk in communities study,” Stroke,vol. 33, pp. 2718–21, 2002.

[3] H. Nakayama, H. S. Jørgensen, H. O. Raaschou, and T. S. Olsen,“Compensation in recovery of upper extremity function after stroke:the Copenhagen Stroke Study,” Arch Phys Med Rehabil, vol. 75, pp.852–857, Aug. 1994.

281

[4] H. T. Hendricks, J. van Limbeek, A. C. Geurts, and M. J. Zwarts,“Motor recovery after stroke: a systematic review of the literature,”Arch Phys Med Rehabil, vol. 83, no. 11, pp. 1629–1637, Nov. 2002.

[5] S. B. O’Sullivan and T. J. Schmitz, Physical Rehabilitation: Assess-ment and Treatment. F A Davis Co., 1988.

[6] A. Kusoffsky, I. Wadell, and B. Y. Nilsson, “The relationship betweensensory impairment and motor recovery in patients with hemiplegia,”Scand J Rehab Med, vol. 14, pp. 27–32, 1982.

[7] R. J. Nudo, K. M. Friel, and S. W. Delia, “Role of sensory deficits inmotor impairments after injury to primary motor cortex,” Neurophar-macology, vol. 35, no. 5, pp. 733–742, 2000.

[8] M. J. Reding and E. Potes, “Rehabilitation outcome following initialunilateral hemispheric stroke. Life table analysis approach,” Stroke,vol. 19, pp. 1354–1358, 1988.

[9] A. C. Lo et al., “Robot-assisted therapy for long-term upper-limb im-pairment after stroke.” N Engl J Med, vol. 362, no. 19, pp. 1772–1783,May 2010.

[10] O. Lambercy, A. J. Robles, Y. Kim, and R. Gassert, “Design of arobotic device for assessment and rehabilitation of hand sensory func-tion.” in IEEE Int Conf Rehabil Robot, vol. 2011, 2011, p. 5975436.

[11] S. J. Page, P. Levine, and A. Leonard, “Mental practice in chronicstroke: results of a randomized, placebo-controlled trial,” Stroke,vol. 38, pp. 1293–1297, Mar. 2007.

[12] R. Gentili, C. E. Han, N. Schweighofer, and C. Papaxanthis, “Motorlearning without doing: trial-by-trial improvement in motor perfor-mance during mental training.” J Neurophysiol, vol. 104, no. 2, pp.774–783, Aug. 2010.

[13] M. Roth, J. Decety, M. Raybaudi, R. Massarelli, C. Delon-Martin,C. Segebarth, S. Morand, A. Gemignani, M. Decorps, and M. Jean-nerod, “Possible involvement of primary motor cortex in men-tally simulated movement: a functional magnetic resonance imagingstudy,” Neuroreport, vol. 17, no. 7, pp. 1280–4, May 1996.

[14] E. L. Altschuler, S. B. Wisdom, L. Stone, C. Foster, D. Galasko, D. M.Llewellyn, and V. S. Ramachandran, “Rehabilitation of hemiparesisafter stroke with a mirror,” Lancet, vol. 353, no. 9169, pp. 2035–6,Jun. 1999, 2035-6.

[15] V. S. Ramachandran and E. L. Altschuler, “The use of visual feed-back, in particular mirror visual feedback, in restoring brain function.”Brain, vol. 132, no. Pt 7, pp. 1693–1710, Jul. 2009.

[16] M. E. Michielsen, R. W. Selles, J. N. van der Geest, M. Eckhardt,G. Yavuzer, H. J. Stam, M. Smits, G. M. Ribbers, and J. B. J. Buss-mann, “Motor recovery and cortical reorganization after mirror ther-apy in chronic stroke patients: a phase II randomized controlled trial,”Neurorehabil Neural Repair, vol. 25, pp. 223–233, Mar. 2011.

[17] C. Dohle, J. Pullen, A. Nakaten, J. Kust, C. Rietz, and H. Karbe, “Mir-ror therapy promotes recovery from severe hemiparesis: a randomizedcontrolled trial,” Neurorehabil Neural Repair, vol. 23, no. 3, pp. 209–17, Mar. 2008.

[18] G. Yavuzer, R. Selles, N. Sezer, S. Sutbeyaz, J. B. Bussmann,F. Koseoglu, M. B. Atay, and H. J. Stam, “Mirror therapy improveshand function in subacute stroke: a randomized controlled trial,” ArchPhys Med Rehabil, vol. 89, pp. 393–398, Mar. 2008.

[19] G. M. Goodwin, D. I. McCloskey, and P. B. Matthews, “A systematicdistortion of position sense produced by muscle vibration,” in Journalof Physiology, vol. 221, Feb. 1972, pp. 8P–9P.

[20] G. M. Goodwin, D. I. McCloskey, and P. B. Matthews, “The contri-bution of muscle afferents to kinaesthesia shown by vibration inducedillusions of movement and by the effects of paralysing joint afferents,”Brain, vol. 95, no. 4, pp. 705–48, 1972.

[21] F. J. Clark, P. B. Matthews, and R. B. Muir, “Effect of the amplitudeof muscle vibration on the subjectively experienced illusion of move-ment,” in Journal of Physiology, vol. 296, Nov. 1979, pp. 14P–15P.

[22] J. P. Roll and J. P. Vedel, “Kinaesthetic role of muscle afferents inman, studied by tendon vibration and microneurography,” Exp BrainRes, vol. 47, pp. 177–190, 1982.

[23] J. P. Roll, J. P. Vedel, and E. Ribot, “Alteration of proprioceptivemessages induced by tendon vibration in man: a microneurographicstudy,” Exp Brain Res, vol. 76, no. 1, pp. 213–222, 1989.

[24] P. Cordo, V. S. Gurfinkel, L. Bevan, and G. K. Kerr, “Proprioceptiveconsequences of tendon vibration during movement,” J Neurophysiol,

vol. 74, no. 4, pp. 1675–1688, Oct. 1995.[25] J. T. Inglis and J. S. Frank, “The effect of agonist/antagonist muscle

vibration on human position sense,” Exp Brain Res, vol. 81, pp. 573–580, Feb. 1990.

[26] S. Calvin-Figuiere, P. Romaiguere, J. C. Gilhodes, and J. P. Roll,“Antagonist motor responses correspond with kinesthetic illusions in-duced by tendon vibration,” Exp Brain Res, vol. 124, pp. 342–350,Aug. 1999.

[27] F. Albert, M. Bergenheim, E. Ribot-Ciscar, and J. P. Roll, “The Iaafferent feedback of a given movement evokes the illusion of the samemovement when returned to the subject via muscle tendon vibration,”Exp Brain Res, vol. 172, pp. 163–174, Jan. 2006.

[28] J. P. Roll, F. Albert, C. Thyrion, E. Ribot-Ciscar, and M. Bergenheim,“Inducing any virtual two-dimensional movement in humans by ap-plying muscle tendon vibration,” J Neurophysiol, vol. 101, pp. 816–823, Dec. 2009.

[29] H. Neiger, J. C. Gilhodes, M. F. Tardy-Gervet, and J. P. Roll,“Reeducation sensori-motrice par assistance proprioceptive vibra-toire,” Kinesitherapie Scientifique, no. 252, pp. 6–21, 1986.

[30] A. Gay, S. Parratte, B. Salazard, D. Guinard, T. Pham, R. Legre, andJ. P. Roll, “Proprioceptive feedback enhancement induced by vibra-tory stimulation in complex regional pain syndrome type I: an opencomparative pilot study in 11 patients,” Joint Bone Spline, vol. 74, pp.461–466, Aug. 2007.

[31] J. P. Roll, “Reeducation proprioceptive par vibration tendineuse,” Pro-fession Kinesitherapeute, no. 23, pp. 11–16, Jun. 2009.

[32] O. Celik, M. K. O’Malley, R. B. Gillespie, P. A. Shewokis, and J. L.Contreras-Vidal, “Compact and low-cost tendon vibrator for inducingproprioceptive illusions,” in Third Joint Eurohaptics Conference andSymposium on Haptic Interfaces for Virtual Environment and Teleop-erator Systems, Salt Lake City, UT, USA, Mar. 2009, pp. 623–624.

[33] C. Thyrion and J. P. Roll, “Perceptual Integration of Illusory and Imag-ined Kinesthetic Images,” J Neurosci, vol. 29, pp. 8483–8492, Jul.2009.

[34] Hadassah Medical Organization, “Use of Tendon Vibration and Mir-ror for the Improvement of Upper Limb Function and Pain Reduction(VibMirror),” http://clinicaltrials.gov/ct2/show/study/NCT01010607,May 2011.

[35] M. D. Rinderknecht, “Device for a novel hand and wrist rehabilitationstrategy for stroke patients based on illusory movements induced bytendon vibration,” in Electrotechnical Conference (MELECON), 201216th IEEE Mediterranean, Mar. 2012, pp. 926 –931.

[36] N. Gebhardt et al., “Irrlicht Engine,” http://irrlicht.sourceforge.net/,Oct. 2012.

[37] S. Cuccurullo, Physical Medicine and Rehabilitation Board Review,S. Cuccurullo, Ed. Demos Medical Publishing, 2004.

[38] B. Craske, “Perception of impossible limb positions induced by ten-don vibration.” Science, vol. 196, no. 4285, pp. 71–73, Apr. 1977.

[39] G. Gescheider, Psychophysics: The Fundamentals. Lawrence Erl-baum Associates, 1997.

[40] M. Tsuge, M. Izumizaki, K. Kigawa, T. Atsumi, and I. Homma, “Inter-action between vibration-evoked proprioceptive illusions and mirror-evoked visual illusions in an arm-matching task,” Experimental BrainResearch, pp. 1–11, 2012.

[41] Y. J. Kang, H. K. Park, H. J. Kim, S. J. Im, J. Ku, S. Cho, S. I. Kim, andE. S. Park, “Upper extremity rehabilitation of stroke: Facilitation ofcorticospinal excitability using virtual mirror paradigm.” J NeuroengRehabil, vol. 9, no. 1, p. 71, Oct. 2012.

[42] J. R. Lackner and A. B. Taublieb, “Influence of vision on vibration-induced illusions of limb movement.” Exp Neurol, vol. 85, no. 1, pp.97–106, Jul. 1984.

[43] M. Slater, D. Perez-Marcos, H. H. Ehrsson, and M. V. Sanchez-Vives, “Inducing illusory ownership of a virtual body,” Front Neu-rosci, vol. 3, no. 2, pp. 214–220, Sep. 2009.

[44] D. Leonardis, A. Frisoli, M. Solazzi, and M. Bergamasco, “Illusoryperception of arm movement induced by visuo-proprioceptive sensorystimulation and controlled by motor imagery,” in Haptics Symposium(HAPTICS), 2012 IEEE, Mar. 2012, pp. 421 –424.

[45] P. Berlit, Basiswissen Neurologie, 5th ed. Springer, 2007.

282