Evaluation of Position Based Cueing Strategies for Bilateral Robotic Assessment and Therapy after Stroke

Shantanu Karnik, BS1,2 , Michelle J. Johnson, PhD1,2,3 Robert Scheidt, PhD2
1 Rehabilitation Robotics Research and Design Lab, Zablocki VA Medical Center, Milwaukee, WI.
2 Department of Biomedical Engineering, Marquette University, Milwaukee, WI.
3 Department of Physical Medicine and Rehabilitation, Medical College of Wisconsin, Milwaukee, WI


Stroke is the leading cause of disability in the United States. Hemiparesis, affecting the majority of the subjects, leads to other related process like learned non-use, which compromises the affected individual's functionality. Many current therapy techniques do not incorporate new research which point towards involvement of both arms in a synchronous, bimanual but decoupled environment to achieve maximum recovery. A novel method of robotic therapy, using position based cues, was tested on able-bodied individuals with strong handed preferences as a model for stroke survivors with severe arm bias due to learned non-use, to verify the effectiveness in assessing and changing in the short term their hand preference. The reduction in non-dominant arm error in a significant number of subjects shows this method can be used as a useful model for future therapy techniques using robotic systems.


Robot Therapy, Bimanual co-ordination, position based cueing, hemiparesis


Stroke is the leading cause of disability among adults in the United States[(1)].  According to current statistics, between 500,000 to 750,000 people have a stroke each year and there are currently about 5.4 million persons living with post-stroke aftereffects [(2)]. As the average age of the US population increases, incidences of stroke will increase because those over 55 years are at higher risk.  Hemiparesis is the most common deficit after stroke, affecting greater than 80% of subjects acutely and greater than 40% chronically. Even though there is a wide range in the degree of recovery only 5% of adults regain full functionality while 20% show no recovery at all.[(3)] The disabilities caused by hemiparesis and other associated processes like learned non-use, often disrupt the individual’s ability to function independently during the everyday activities in his or her chosen living environment.

Learned non-use affects a large number of stroke survivors [(4)].  There are two aspects to learned non-use of the impaired arm after a stroke.  Physical inability, which is actual loss of ability to do activities which involve the stroke affected side and psychological, i.e., an increased dependence on the non-affected side to do activities to avoid appearing clumsy and doubting the ability of the affected side to carry out simple tasks.

Many current therapy and rehabilitation strategies for stroke survivors advocate the use of the impaired arm in a unilateral exercise environment [4]. Depending on the therapy method prescribed, the unimpaired arm is either constrained from contributing to the therapy or has minimal involvement in the task being performed. Recent evidence shows that functional recovery might improve more and endure for a longer time period if the therapy were to be performed in a bilateral environment [(5),(6)].

We assumed able-bodied handedness as a model for hemiparesis after stroke and looked at strategies to maximize motor learning to increase affected, upper limb function. Active force-feedback cues[(7)] have been demonstrated as effective therapy techniques to help hemiplegic subjects become aware of their overuse of the less-affected limb. The research assesses position-based cues as a way of disrupting bias for the non-dominant arm. Using these cues, the dominant (D) was perceptually made unreliable or less effective in an experimental setup, in a bilateral target acquisition. Our aim was to verify if the subject started relying on the non-dominant(ND) equally even when the bias against the D is removed.  This paper reports on our findings.


Equipment and Setup:

The passive robots are made of nylon plastic.  The subject grips the manipulandum (Hand Grip) and moves the wrist back and forth. The total angle covered is about 120 degrees. The change in angle will be converted to linear movement of a position cursor. Fig.1: Passive Robots (Click image for larger view)

The experimental was conducted using two 'Passive Robots' as shown in Fig.1. The passive robots are made of nylon plastic [(8)].  The subject grips the manipulandum (Hand Grip) and moves the wrist back and forth. The total angle covered is about 120 degrees. The change in angle will be converted to linear movement of a position cursor. There is minimal resistance on the manipulandum and the complete movement is due to the subject’s effort. Fig. 2 shows the system setup with the subject present. Joint angle is measured with an Agilent HEDM-6540, 3-channel, Mylar film optical encoder (Agilent Technologies, Inc., Palo Alto, CA), paired with a Measurement Computing PCI-QUAD-04 incremental encoder driver (Measurement Computing Co., Middleboro, MA). A custom written code in MATLAB (Simulink, Stateflow and XPC Target modules) was used to process the data.

The subjects grip the 2 manipulandums of the passive robots and start the movement with their wrists extended. Moving the manipulandum towards the midline of the body will move the position cursor towards the target on the visual display. Fig.2 Experimental Setup (Click image for larger view)


All subjects tested were right-hand dominant as determined from a Handedness Survey[(9)] administered before the experiment. The subjects were asked to move the position cursor onto a target on the visual display screen using the flexion-extension movement of the wrist. Fig. 3 shows the visual display seen by the subject. Depending on the experiment, the tracking task required either unilateral or bilateral wrist movements.  The experiment was a two session protocol. In the 1st session (Day 1), target parameters such as distance were varied and the scaling factor associated with the movement of each wrist was varied with respect to the dominant(D) and non-dominant(ND) arm. The scaling factor was included to equalize the effect of the tracking movements of the D and ND or favor the ND over D. We recorded wrist angles and derive relevant dependent variables such as the final position of the position cursor, accuracy of arm movements for different target distances and scaling factors. There was a uniform time limit to complete the movement to maintain uniformity and the subjects were not be informed of the bias (scaling factor) used during the movements. In the 2nd session (Day 2), the subject performed experiments that determined the target distance and scaling factor at which the error was the least. This ensured that the bias against the D as only due to the perturbation introduced.  

The white circle shows the position cursor on the visual display. The movement of the wrist robots moves the cursor towards the target as indicated by the black ring towards the top of the figure. Fig.3: Visual Display (Click image for larger view)

The perturbations introduced were divided into 3 types. The perturbations were only introduced into the movement of the UI when the difference between the angle contributions of both arms exceeded a pre-determined limit in favor of the UI. This indirectly encouraged symmetrical movement.

  1. Dynamically Varying Gain: The angle contributions of either arm were dynamically compared and the scaling factor was biased against the UI according to the condition described above.
  2. Predictable Perturbation: The angle contributions were continuously monitored and a fixed amplitude Gaussian noise perturbation in terms of position was introduced when the UI started dominating the movement.
  3. Unpredictable Perturbation: A random, Gaussian noise perturbation was introduced when the angle contribution by the UI was greater than the pre-determined limit.

The subjects were tested before and after the 'Perturbation Experiments' to assess the change in error and arm bias. The pre and post tests were conducted with static target distances and scaling factors. . RMS error and standard deviations were calculated for each trial by normalizing target distances and scaling factors for all subjects.


The figure shows the difference in the movement for a sample subject trial for each cueing strategy. Graph 1: Difference in Movement due to Cueing Strategy (Click image for larger view)

Graph 1 shows the difference in the movement as performed by a sample subject categorized by the cueing method. The unpredictable perturbation group typically took the longest to complete the movement while the predictable perturbation group took the shortest time for the same distance.  Graph 2 shows the RMS error values for each set of subjects before and after the perturbation experiments. The errors for both arms decreased after perturbation experiments in the Predictable and Unpredictable perturbation subject group while the error for the left (non-dominant) arm increased for the Dynamically Varying Gain subject group.

The graph shows the change in RMS error and standard deviation for each arm before and after the training session as categorized by cueing strategy Graph 2: Change in RMS Error (Click image for larger view)

This task was designed to include several factors that have proved to increase the efficacy of rehabilitation therapy. Due to the use of robotic techniques to perform the task, it is highly repeatable. As both the arms are involved, the synchronous, bilateral nature of the task has a facilitation effect from the I to the UI i.e. the quality of movement (measured here in terms of accuracy and rate of learning) improved in bilateral movement as opposed to unilateral movement thus indication that both arms are strongly linked as a coordinated unit in the brain[(10)]. There is also evidence that learning as task in the unilateral condition will result transfer of similar learning skill to the other arm after a period of time[(11)].

The repetitive, rhythmic nature of the task has been shown to promote motor learning[(12)]. Repetition, or “time on task,” is a well-known motor learning principle, and recent animal studies have demonstrated that forced use involving a repetitive motor task rather than forced use alone may best promote central neural plasticity. The goal setting inherent to the task has also been shown to promote motor learning[(13)]. The sensory feedback, from visual and auditory cues, provides information about the goal of the movement which increases motor learning behavior[(14)]. The reduction in non-dominant arm error in a significant number of subjects shows this method can be used as a useful model for future therapy techniques using robotic systems.

Table 1: Error and Std. Dev



Std. Dev










Predictable Perturbation
Unpredictable Perterbation
Dynamically Varying Gain


This work was supported by NSF BES0238442. We would like to thank Aaron Suminski,PhD and Timothy Haswell,BS for their technical guidance and also Ruta Paranjape, BS, Dominic Nathan,BS for their assistance in running the sessions and data acquisition. We'd also like to thank the Physical Medicine and Rehabilitation Department, Medical College of Wisconsin for their support.


 Michelle J. Johnson, PhD (mjjohnso@mcw.edu)
Rehabilitation Robotics Research and Design Lab, Zablocki VA Medical Center, 5000 W. National Ave.
Milwaukee, WI, 53295.


  1. G. Gresham, P.W. Duncan.Post-stroke Rehabilitation. Clinical Practice Guideline. Washington DC: US Department of Health Services, AHCPR Publication vol. 16, no. 95-0662, 1995
  2. Stineman MG, Maislin G, Fiedler RC, Granger CV (1997).A prediction model for functional recovery in stroke. Stroke ;28:550-6.
  3. Gowland C, deBruin H, Basmajian J, Plews N, Nurcea I(1992).Agonist and antagonist activity during voluntary upper-limb movement in patients with stroke. Phys Ther.; 72:624–633.
  4. Taub E,Uswatte G,Pidikiti R.(1999).Constraint-Induced Movement Therapy:A new family of techniques with broad application to physical rehabilitation-a clinical review.J Rehabil Res Dev 36:237-51
  5. Kelso JAS, Southard DL, Goodman D(1979).On the coordination of two handed movements. J Exp Psychol Hum Percept Perform 5: 229–238.
  6. Kelso JAS, Putnam CA, Goodman D(1983).On the space-time structure of human interlimb coordination. Q J Exp Psychol. ;35A:347–375.
  7. Johnson, M.J., Van der Loos, H.F., Burgar, C.G., Shor, P., Leifer, L.J.(2005).Experimental results using force-feedback cueing in robot-assisted stroke therapy. IEEE Transactions on Neural Systems and Rehabilitation Engineering 13, 335–348.
  8. Aaron J. Suminski, Stephen M. Rao, Kristine M Mosier, and Robert A. Scheidt(2006).Neural and Electromyographic Correlates of Wrist Posture Control. J Neurophysiol .
  9. Oldfield RC(1971).The assessment and analysis of handedness:The Edinburgh inventory. Neuropsychologia 9:97-113.
  10. Geffen GM, Jones DL, Geffen BL(1994).Interhemispheric control of manual motor activity. Behav Brain Res. ;64:131–140.
  11. Lazarus JC, Whitall J, Franks CA(1995).Age difference in isometric force regulation. J Exp Child Psychol. ;60:245–260.
  12. Schmidt RA, Lee TD(1998).Motor Control and Learning. 3rd ed. Champaign, Ill: Human Kinetics; .
  13. Locke EA, Bryan JF(1966).Cognitive aspects of psychomotor performance: the effects of performance goals on level of performance. J Appl Psychol. ;50:286 –291.
  14. Salmoni AW, Schmidt RA, Walter CB(1984). Knowledge of results and motor learning: a review and reappraisal. Psychol Bull;95:355–386.


  • Source Ordered
  • No Tables
  • Very Compatible


Disney produced a television show in the mid 1990s called Gargoyles. It's a great show and I'm a big fan. A few years ago Disney started to release the show on DVD. The last release was of season 2, volume 1. That was two years ago. Volume 2 has not been released. Why? Poor sales. So if you should find yourself wanting to support my work, instead I ask you pick up a copy of season 2, volume 1. It's a great show and you might find yourself enjoying it.