Background
Stroke often results in impairments of upper extremity, including hand and finger function, with 75% of stroke survivors facing difficulties performing activities of daily living [
1,
2]. Critically, impairments after stroke not only include muscle- and joint-specific deficits such as weakness, and changes in the kinetic and kinematic workspace of the fingers [
3,
4], but also coordination deficits such as reduced independent joint control [
5] and impairments in finger individuation and enslaving [
6‐
9]. Therefore, understanding how to address these coordination deficits is critical for improving hand rehabilitation.
Typical approaches to hand rehabilitation emphasize repetition [
10] and functional practice based on evidence that such experience can cause reorganization in the brain [
11]. Although this has proven to be reasonably successful, functional practice (such as repetitive grasping of objects) does not specify the coordination pattern to be used when performing the tasks. As a result, because of the redundancy in the human body, there is a risk that stroke survivors may adopt atypical compensatory movements to perform tasks [
12]. These compensatory movements have been mainly identified during reaching [
13,
14], but there is evidence that they are also present in finger coordination patterns during grasping [
15]. Although there is still debate over the role of compensatory movements in rehabilitation [
16], there is at least some evidence both in animal and humans that continued use of these compensatory patterns may be detrimental to true recovery [
17‐
19].
To address this issue, there has been a greater focus on directly facilitating the learning of new coordination patterns. Specifically, in hand rehabilitation, virtual tasks (such as playing a virtual piano) have been examined as a way to train finger individuation [
20,
21]. In these protocols, individuation is encouraged by asking participants to press a particular key with a finger, while keeping other fingers stationary. A similar approach to improve hand dexterity was also adopted by developing a glove that could be used as a controller for a popular guitar-playing video game [
22]. However, directly instructing desired coordination patterns to be produced becomes challenging as the number of degrees of freedom involved in the coordination pattern increase. For example, the hand has approximately 20 kinematic degrees of freedom, and providing verbal, visual or auditory feedback for simultaneously controlling all these degrees of freedom would be a major challenge. A potential solution that has been suggested is not to directly instruct the coordination pattern itself, but rather let participants explore different coordination patterns [
23]. This idea of motor exploration is based on dynamical systems theory that suggests that variability and exploration may help participants escape sub-optimal pre-existing coordination patterns and potentially settle in more optimal coordination patterns [
24‐
27]. Such exploration has been shown to be important in adapting existing movement repertoire [
28], and has also been shown to be associated with faster rates of learning [
29].
In order to test the hypothesis that exploration of novel coordination patterns can improve overall movement repertoire, I used a body-machine interface [
30,
31] to examine how stroke survivors explore and reorganize finger coordination patterns with practice. A body-machine interface maps body movements (in this case finger movements) to the control of a real or virtual object (in this case a screen cursor), which can provide a way to elicit different coordination patterns in the context of an intuitive task. Specifically I examined: (i) how stroke survivors reorganize their finger coordination patterns, (ii) how training to explore novel coordination patterns affects their ability to reorganize their coordination pattern, and (iii) if training to explore novel coordination patterns has an effect on their overall movement repertoire. In this context, I use the term “novel” to indicate coordination patterns that require finger individuation. This assumption is motivated by the finding that stroke survivors have difficulty producing finger individuation even under explicit instruction [
6,
9], and therefore it is highly likely that they would not use coordination patterns requiring finger individuation frequently in activities of daily living.
Discussion
The aim of this study was to examine how stroke survivors explore and reorganize finger coordination patterns. Participants used a body-machine interface that mapped their finger movements in different ways to the control of the cursor. The map between finger and cursor motion was changed without the knowledge of the participants to get participants to explore different coordination patterns to perform the task. I found the following results: (i) stroke survivors have difficulty when the mapping requires finger individuation, and the amount of exploration observed was associated with clinical tests of hand function, (ii) with 4 sessions of learning, participants were better able to reorganize their coordination patterns to perform the task requiring individuation, and (iii) this training was also associated with a modest improvement in overall movement repertoire outside of the task.
Our first finding that stroke survivors had difficulty with finger individuation is consistent with the literature [
6,
9]. Movement times were larger in the IM-RL map, which required some degree of individuation between the IM and the RL fingers. Even though this result was true in both the paretic and the non-paretic hands indicating that some of this increase in movement time could be due to the non-intuitive mapping in the IM-RL map, the magnitude of increase in movement time for the IM-RL map was much larger in the paretic hand. I also found that hand function at baseline (as measured by the Box and Blocks test) correlated with the degree of exploration in the IM-RL task – i.e., the lesser the function, the less exploration there was during the task (i.e. higher the PC1 VAF). This is in line with recent evidence that that the degree of individuation is correlated to damage to the hand area in motor cortex and the corticospinal tract [
37]. One important difference from earlier studies is that in the current paradigm, there was no instruction to the participants regarding the coordination pattern to be produced (because the task was redundant, there were multiple coordination solutions that could be used to reach the targets). The fact that I still found similar results suggests that reduced finger individuation may be indicative of an overall reduced movement repertoire.
The second finding was that with training, participants showed improvements in reorganizing their coordination pattern when performing the task. Participants who were unable to complete the full set of targets at the first session increased the number of targets that they were able to reach by the last session, and there was also an overall drop in the movement time. Reorganization times also decreased with practice, indicating participants were more efficient in their exploration after practice. The reorganization was also correlated with their initial hand function – this was indicated by a correlation between the change in PC1 angle and the initial Box and Blocks score. These results are consistent with evidence that stroke results in altered muscle synergies, indicating reduced movement repertoire [
38] and a recent study that found that the degree of task-specific modulation in muscle activity was correlated with the level of impairment [
39].
Finally, I also found changes in movement repertoire during the free exploration task after training (as quantified by the PC1 VAF). In the free exploration task, there was no cursor to control, but instead participants were simply asked to explore the repertoire of finger movements. Here, there were two observations – (i) the initial movement repertoire (the PC1 VAF) was correlated to the Box and Blocks test, indicating that initial hand function was associated with movement repertoire, and (ii) I found a decrease in VAF after 4 sessions of training, suggesting that movement repertoire improves when participants learn to explore and reorganize their finger coordination when learning the cursor control task (as seen by the changes in movement time as a function of trial block in Fig.
6).
These results point to an important, but often underappreciated role of motor exploration in rehabilitation. First, exploration can serve as an assessment tool to quantify the existing movement repertoire. For example, a recent study quantified the movement repertoire in stroke survivors using free motor exploration during a reaching task, and the results showed characteristic differences in exploration in stroke survivors that were not detected by more common measures such as range of motion [
40]. Second, exploration can also serve as a rehabilitation tool by expanding the movement repertoire. Conventional task-specific training for the hand (such as precision and power grasping) only span a limited movement repertoire [
41]. Repetitions of functional tasks may therefore not help participants recover the full dexterity of the hand, leading to the use of compensatory strategies. In contrast, the use of motor exploration to facilitating practice of lesser used coordination patterns may make it possible to widen the movement repertoire. However, one downside to using exploration during training is that there is no guarantee that the coordination patterns being trained are functional in any way. Therefore, one way to structure rehabilitation to get the benefit of both exploration and task-specific training may be to combine exploration as a “priming” tool to improve the repertoire and then follow up with task-specific functional training so that appropriate coordination patterns are learned.
The results also show the potential of body-machine interfaces as an important tool to reorganize coordination [
30,
42]. The presence of a large number of degrees of freedom (especially in the hand) means that direct feedback on each individual degree of freedom may be impractical. By mapping this high dimensional space to the control of a low dimensional control object, body machine interfaces provide an intuitive way of eliciting different coordination patterns. This property of body-machine interfaces has typically been used for the control of assistive devices for persons with limited mobility [
43‐
45]. However, the current results support recent evidence that these interfaces can also be used as a rehabilitation tool to improve the movement kinematics [
46,
47] as well alter abnormal muscle activity [
48].
There were some limitations to the study: first, this was a single group study with no control group, which makes it difficult to attribute improvements to the training protocol, and if the improvements observed in the cursor control task were simply effects of repeated exposure to the task. However, there are two reasons why I think that improvements observed are likely due to training: (i) this study was done in chronic stroke survivors who were generally several years past the stroke, making the change in the course of a few weeks less likely due to other factors, and (ii) unlike functional tests where there is clearly repeated effects of exposure, I found improvements in movement repertoire during the free exploration task, which was unrelated to the cursor control task. A second limitation was that although I was able to recruit participants with a wide range of impairment levels, all participants had to be able to wear the data glove, and have at least some volitional movement of the fingers, which resulted in a rather small sample size that potentially limits the generalizability of the findings. However, the same approach could be used in principle with more severe motor impairments using assistive robots that can aid finger movements [
49].
Conclusion
In conclusion, stroke survivors show deficits in movement repertoire that can be assessed through motor exploration. With training, participants were able to better explore novel coordination patterns outside the movement repertoire, and that also resulted in an increase in the movement repertoire. These results point out that exploration may be a critical tool to improve hand dexterity.
Acknowledgments
The author also wishes to thank Meghan Morrow, OT for assisting with participant recruitment and clinical tests.