In this study changes in unsupported arm movements, induced by gravity compensation training, and the impact of abnormal coupling on arm movements in chronic stroke patients were studied by means of a circle drawing task . After 6 weeks of moderately intense gravity compensated reach training, a sample of 7 chronic stroke patients showed improved arm function, as indicated by a median increase of 3 points on the FM scale. Subjects also significantly increased the active work area of the hand as indicated by the normalized circle area, whereas circle roundness remained almost constant. Statistically significant changes were observed in excursions of EP and EA, accompanied by increasing trends in excursions of AR and EF. The occurrence of synergistic movement patterns was similar before and after training.
The effect of abnormal coupling between shoulder and elbow movements on circle roundness was previously studied by Dipietro et al. . Roundness of the circles drawn in the present study before training (mean ± SD 0.32 ± 0.14) was lower compared to the roundness of circles (mean 0.51) drawn in the study by Dipietro et al., despite a less severe patient group in the present study, as indicated by a higher FM score (FM = 27.1 and 20.5 respectively). This discrepancy is most likely related to the level of arm support during evaluation measurements. Dipietro et al. evaluated circle drawing while the subject’s arm was supported against gravity, while in the present study unsupported circle drawing was evaluated. Application of gravity compensation has been shown to reduce the influence of abnormal coupling between the shoulder and elbow [15, 29] which is likely to result in rounder circles.
After robot-assisted, gravity compensated point-to-point reach training Dipietro et al. found an average increase in roundness of 0.10 [21, 26]. The increase in roundness was the result of increasing minor axis of the fitted ellipse, while the major axis remained constant. In the present study, roundness remained similar before and after training, while the normalized circle area increased, i.e. both minor and major axes increased. A possible explanation for this discrepancy is a difference in training method. In both studies the arm was supported against gravity. However, in Dipietro et al. subjects who were not able to reach a target were assisted by the robot to complete the movement task, as well. Although recent reviews [30, 31] that addressed technology supported arm training could not discern whether or not certain training modalities are more effective than others, it may be that differences in training modalities influenced roundness of the circles. A second explanation is related to the nature of the movement task that was assessed during evaluation measurements. In the study by Dipietro et al. subjects were asked to draw a copy of a template circle with a fixed radius of 14 cm. In the present study, subjects were asked to draw circles as big and as round as possible, during both evaluation measurements before and after the training period. As confirmed by the circle metrics, the focus of most subjects in the present study was in increasing circle area at the expense of increasing roundness.
Horizontal circle drawing can be seen as a continuous reaching task in the medio/lateral and forward/backward direction. Previous studies showed that gravity compensation training led to increased range of motion of the impaired arm as represented by increased maximal unsupported reach distance [13, 18, 19]. Ellis et al.  observed increased work area at various limb loadings after point-to-point reach training during which the level of arm support decreased progressively. Increased range of motion during unsupported arm movements was also found in the present study, as indicated by an increased normalized circle area after training.
Dipietro et al.  observed that the elliptical shapes drawn by stroke patients became rounder throughout the robot-assisted training period because the minor axis of the ellipse increased while the major axis of the ellipse remained almost constant. It was concluded that existing coupling between the shoulder and elbow joint remained after robot-aided reach training, but that the strength of the coupling decreased, which led to more selective control of the shoulder and elbow joint, as indicated by a lower correlation between shoulder horizontal ab-/adduction and elbow flexion/extension.
Although roundness and the occurrence of synergistic movement patterns were comparable before and after training in the present study, the changes in work area present some indications that a reduced impact of abnormal coupling may play a role in improved arm function. The improved ability to move the shoulder, in combination with a slight increase in elbow flexion and extension resulted in an increased circle area. However, when elbow extension is increased, i.e. the hand is moved away from the torso, higher muscle activations in the shoulder and elbow joint are needed to hold the arm against gravity, and stabilize the joints . Consequently, these higher abduction torques will induce an increased amount of involuntary elbow flexion [9, 10, 13]. In other words, it is possible that stroke patients increased their work area because of a decreased impact of abnormal coupling between shoulder and elbow, but that the higher shoulder abduction torques needed to perform the movement task, again provoke an abnormal coupling at the elbow, resulting in similar amounts of movement within/out of synergistic patterns and consequently a proportional increase in both the major and minor axes of the fitted ellipse before and after training. A possibility to increase insight into the mechanisms involved in improving post stroke arm function, is to combine the circle drawing task used in the present study, in which patients maximize circle area, with a circle tracking task in which the size of the circles remains constant. With this second circle drawing task, the effect of synergistic movement patterns on circle shape can be studied, without the effects of increases in shoulder abduction torques that are needed to draw bigger circles. Differences in occurrence of synergistic movement patterns are likely to result in changes of circle area and roundness [21, 22].
Besides involuntary coupling between the shoulder and elbow joint, many stroke patients also have to deal with muscle weakness  and/or strength imbalances across joints . It is possible that patients strengthen their muscles during training, improve their temporal muscle activation or muscle coordination in general. To illustrate this, in the same sample of stroke patients, increased activity of agonist muscles during a maximal forward reaching task was observed after training . This increased agonist muscle activity can be a result of increased ability to selectively activate agonist muscles, or improved control of agonist muscles. More research regarding changes in muscle activation patterns or changes in maximal voluntary torques (MVT) is needed to study the working mechanisms involved in changes of arm function after gravity compensated reach training.
Limitations and recommendations
The present findings show that reach training with a low-cost arm support system and a low-tech computer game is able to improve hemiparetic arm function in a sample of chronic stroke patients. The reported increases in FM scores are comparable with interventions using more advanced training systems . Nevertheless, results of the present study should be interpreted carefully, because of the small sample size of the study and the absence of a control group. Because of the small number of participants, it was not possible to subdivide the subjects into subgroups with different levels of stroke severity, to study potential differences in effects of gravity compensated reach training on hemiparetic arm function.
Since all subjects who participated were in the chronic phase after stroke, it may me be that subjects learned to avoid using the impaired arm . It is not known whether improvement in hemiparetic arm function is due to improved neuromuscular control induced by the training, or by overcoming possible learned nonuse of the impaired arm. Inclusion of a control group in future research can yield information to what extent both processes occur.
Further research with larger and more homogeneous samples of stroke patients is needed to increase insight in the physiological mechanisms involved in the training induced changes in arm function, for example by studying training induced changes in muscle activation patterns.
The present study indicates that a moderately intense training program consisting of gravity-compensated point-to-point reach training within a VR augmented training environment can lead to increased work area of the hemiparetic arm in a sample of mildly to severely affected chronic stroke patients. Results concerning the underlying mechanisms causing these changes point towards a less pronounced influence of synergistic movement patterns, although more research is needed for further elucidation. The used training setup is simple and affordable and is therefore suitable to be deployed in clinical settings.