Interaction of post-stroke voluntary effort and functional neuromuscular electrical stimulation

Abstract

Functional Electrical Stimulation (FES) may be able to augment functional arm and hand movement after stroke. Post-stroke neuroprostheses that incorporate voluntary effort and FES to produce the desired movement need to consider how the forces generated by voluntary effort and FES combine together, even in the same muscle, in order to provide an appropriate level of stimulation to elicit the desired assistive force. The goal of this study was to determine if the force produced by voluntary effort and FES add together independently of effort, or if the increment in force is dependent on the level of voluntary effort. Isometric force matching tasks were performed under different combinations of voluntary effort and electrical stimulation. Participants reached a steady level of force and while attempting to maintain a constant effort level, FES was applied to augment the force. Results indicate that the increment in force produced by FES decreases as the level of initial voluntary effort increases. Potential mechanisms causing the change in force output are proposed, but the relative contribution of each mechanism is unknown.

Introduction

Stroke is a leading cause of disability in the US. Six months after their stroke, 50% of ischemic stroke survivors over the age of 64 still have a degree of upper limb hemiparesis (1) that limits arm and hand function, making bimanual tasks difficult if not impossible. Hemiparesis is worsened by disuse and co-contraction patterns across multiple joints (i.e. synergy patterns) (2). These synergy patterns have been well quantified (3, 4) and appear to be expressed in proportion to effort (5, 6). Greater effort to abduct the arm increases the involuntary flexor contractions that oppose the desired movement.

Functional Electrical Stimulation (FES) of paretic muscles has the potential to elicit functional limb movements (7), such as reaching and hand opening (7–9). For example, electrical stimulation of finger extensors in stroke (10–14) can produce hand opening while the participant is relaxed. However, when the user exerts effort to open the hand during stimulation, the hand does not open as much as when the person remains relaxed, presumably because their effort to open the hand produces involuntary finger flexor contractions (10, 15, 16). Therefore in order to receive maximum hand opening from the stimulation, the user must remain relaxed, which is unnatural and runs counter to motor rehabilitation principles, which encourage active attempts to produce functional movement. Our long-term objective is to develop an upper limb FES system that the stroke survivor controls with residual EMG signals recorded from their affected upper limb. Thus, the control strategy will require the user to exert some effort to produce desired arm and hand movements.

In order to develop an effective neuroprosthesis that integrates stimulation and voluntary effort in an efficient and natural way, we need an understanding of their interaction. Previous studies have reported that exerting effort reduces the effectiveness of stimulation both across joints (9), and within the same joint (17). As a result of changes to central commands and reflexes after stroke, there could be differences between how stroke and able bodied participants’ volitional and electrically stimulated forces interact. The quantitative relationship between the level of voluntary effort and the effect of electrical stimulation in the same direction in stroke has not been quantified. It may be possible that small or moderate amounts of effort do not completely limit the effects of stimulation.

The goal of this study was to determine how voluntary effort and electrical stimulation of elbow extensors add together to produce force after stroke (i.e. do the two forces add linearly?). To quantify the relationship between voluntary effort and the effect of electrical stimulation, we stimulated the elbow extensor muscle (triceps) of stroke patients and measured the isometric force that was produced while the participants were asked to exert various levels of simultaneous voluntary effort to push forward, extending the elbow. There are two hypotheses for this study. The first hypothesis is that electrical stimulation of elbow extensors produces greater forward force than voluntary effort alone even in the presence of simultaneous voluntary effort. The second hypothesis is that as the level of voluntary effort increases, the amount of force added by stimulation decreases.

Methods

Participants

Six people who suffered a single stroke were enrolled in this study (Table I). Participants were recruited from an outpatient stroke clinic. The primary inclusion criteria included 1) being at least 6 months post-stroke, 2) the ability to reach at least 20% of their full passive reach starting from the closest their hand can passively sit in front of the shoulder while an investigator manually supported the elbow and wrist at 90 degrees of shoulder abduction, and 3) the ability to follow 3-stage commands. Exclusion criteria included 1) shoulder pain, 2) uncompensated spatial neglect, 3) apraxia, 4) insensate chest, arm, or forearm, and 5) impaired cognition or communication. One participant was excluded from data analysis due to great difficulty completing the tasks involved in this study. All participants provided informed consent in accordance with the Declaration of Helsinki prior to participation in this study, which was approved by an Institutional Review Board.

Table 1.

Participant Demographics.

	Age	Gender	Affected Side	Dominant Side	Time Since Stroke	FMA (arm)
S1	61	M	R	R	1 yr 10 mos	17 (16)
S2	53	M	L	R	11 mos	48 (26)
S3	43	F	R	R	10 mos	26 (21)
S4	57	M	R	R	2 yrs	37 (23)
S5	79	M	R	R	2 yrs 1 mo	25 (20)
S6*	48	M	R	R	4 yrs 8 mos	18 (16)

Abbreviations – FMA: upper limb portion of Fugl-Meyer Motor Assessment, 66 point maximum, shoulder/elbow/forearm in parentheses, 34 point maximum; Gender: M-Male, F-Female; Side: R-Right, L-Left;

^*S6 was excluded from data analysis.

Setup

Participants performed a series of isometric upper extremity forward force generation tasks (described below) while seated with their arm in two different standardized positions (near or far). Each participant sat in front of a computer monitor with his or her trunk restrained as shown in Figure 1. The reference frame used for force directions is shown in Figure 1a.

Figure 1.

Figure 1(a,b). 1a – Example of general setup of how the participant’s arm is oriented. The x, y, and z axes show the coordinate frame for measurements and y and z indicate the directions of rotations for a hypothetical target placement, as indicated by the darker arrows. 1b - Setup showing the near arm location and relative equipment locations.

The arm was positioned at 90° shoulder abduction with the shoulder flexed and elbow extended to bring the wrist directly anterior to the shoulder at a distance from the shoulder of either 50% (near) or 75% (far) of the maximum passive reach. Different positions were compared to determine if arm posture affected the interaction. A fiberglass cast over the wrist and forearm connected the wrist to a 2-degree-of-freedom (DOF) gimbal that was attached to a 6-DOF force transducer (JR3 Inc. Model 30E15A-U560A). The forearm was supported by an elevating mobile arm support (Jaeco JME) in addition to the gimbal in order to reduce pressure at the proximal edge of the cast due to the weight of the arm. Less than 0.059 N of horizontal force was required to move the arm support in the horizontal plane. For the participant with the smallest stimulated force, the force required to move the arm support was less than 0.9% of their stimulated force. The elevating mobile arm support used rubber bands to apply a vertical passive force. The rubber bands providing the support are highly compliant, and the stiffness of the device is less than 9 N/m. We measured the maximum vertical elbow movement during all of the trials to determine the maximum force transmitted through the support. The participant whose movement could have the largest effect on their normalized force had a maximum vertical elbow translation of 0.024 m during a single trial. The maximum active vertical force that could have been transmitted through the support during an entire trial was 3.1% of the stimulated force magnitude. The force transmitted through the support would have been less in the rest of the trials. While less than 3.1% of the stimulated force magnitude was transmitted through the support, the majority of the active vertical forces were recorded by the transducer. Isometric endpoint forces generated by pushing forces were low pass filtered at 5 Hz, sampled at 60 Hz, and then down sampled to 12 Hz.

Surface stimulation electrodes (0.038m × 0.089m) were placed over the triceps to generate elbow extension torques. The stimulation electrodes were approximately placed over the long and lateral heads of the triceps. It is difficult to estimate the proportional contributions of the different heads of the triceps to the stimulated force because the primary target was simply the radial nerve. Current pulses were delivered through a custom computer controlled stimulator with a pulse frequency of 35 Hz, pulse amplitude of 40–60mA, and pulse width that could be modulated from 0 μs to 255 μs. The pulse amplitude and pulse width were set for each participant to levels that produced the desired magnitude (see below) of isometric force without pain when the participant was relaxed with his or her upper arm fully supported. The stimulator ramped the pulse width from 0 μs to the preset level in 0.5 seconds corresponding to cues during the tasks, and remained on for 3 seconds.

Position tracking markers (LED clusters) were placed on the trunk, upper arm, and forearm to record limb movement relative to the trunk with an optical tracking system (Northern Digital, Inc.) in order to confirm that the participant did not move during the trials.

Experimental Procedures

Prior to any isometric force task sessions, an occupational therapist performed a Fugl-Meyer Motor Assessment (18) to characterize the degree of upper limb motor impairment. Participants then returned to the lab for 2 to 4 sessions to learn the force generation tasks and become accustomed to the sensation of surface electrical stimulation. During two final sessions, force and kinematic data were collected for analysis.

Participants performed upper extremity isometric forward force generation tasks that included various combinations of two factors: voluntary effort and electrical stimulation. Before beginning the force generation tasks, participants were shown how to interpret force magnitude and direction feedback on a computer monitor; this was repeated at the beginning of every session. For each force generation task, the participants were presented a target force direction and magnitude. The target direction was the direction closest to a line extending directly anterior from the shoulder in which the participant could consistently generate voluntary force. The target magnitude was a percentage of the maximum voluntary force that participants could maintain in the target direction. Participants were given verbal encouragement to maintain the target force magnitude in a direction as close to anterior as possible. During these tasks, the percentage of the maximum voluntary endpoint force served as an estimate of effort level. Zero effort was treated as the participant being relaxed and not actively exerting any force. Maximum effort (100%) was considered to be when the maximum force was exerted. Measuring endpoint force provided a quantifiable estimate of the participant’s effort during the tasks.

Due to difficulty generating a completely anterior force, a target direction for voluntary force was chosen for each participant in a direction they could consistently generate force. We measured the maximum endpoint force that each participant could generate in the target direction. The participant was instructed to reach a target force and try to maintain that force level. Both the force generated and target force were displayed on the monitor. The magnitude of the maximum force was determined by starting with a level that was initially well within the participants’ achievable range. The magnitude of the target was then incrementally raised in successive trials until participants could no longer generate sufficient force to reach the target. The investigator verbally encouraged participants to generate as much force as possible while determining their maximum force. The maximum magnitude force that participants could maintain in the target direction for 1.5 seconds was defined as their maximum endpoint force, which was used to determine the target force magnitude for each participant.

During the tasks, tolerances on the target force allowed participants to stay within the target without maintaining an exact force. The vertical force tolerance in either direction was 5% of the force being generated in the target direction. The horizontal force tolerance from the target was 25% of the magnitude of the target force. Participants who could generate force more consistently had smaller tolerances on their targets. The reason for a 3-dimensional force matching task was to elicit consistent voluntary force levels by allowing proportional co-contraction levels. While it has been shown that during maximum voluntary contractions stroke victims generate secondary forces in constrained patterns (19); at sub-maximal effort levels people can generate movement/force outside of these patterns (5, 20).

The following force matching tasks were repeated during both the practice and data collection sessions. Each trial type was repeated four to eight times per session depending on the participant. Table II shows each participant’s average rotations for the isometric force target placement around the y and z axes shown in Figure 1a, range of repetitions of the number of trials, and the total average error that participants had during the 1.5 seconds prior to removing feedback and adding stimulation.

Table 2.

Participants average preferred directions for voluntary force generation, the range of repetitions for various tasks and each participants average error during the hold prior to removing feedback and adding stimulation

	Preferred direction rotations (θz/θy)	Repetitions	Error
S1	3°/17°	4–5	17%
S2	14°/25°	5–7	19%
S3	48°/4°	6–7	16%
S4	8°/14°	4–7	19%
S5	24°/12°	5–8	16%

Voluntary Effort Maintenance Task

Participants exerted effort to match a force magnitude that was 20% or 50% of their maximum voluntary force (vFmax) in the target force direction. Once they reached the target magnitude and maintained a steady force for 1.5s, visual force feedback was removed while the participant was instructed to continue exerting the same level of effort.

Stimulation Force Task

Triceps was stimulated to generate a force magnitude equal to 20% of the anterior component of the participant’s vFmax in the target direction. The participant was instructed to remain relaxed and visual force feedback was not provided. A 0.5s linear ramp increase in pulse width began when the stimulation was turned on. Stimulation stayed on for at least three seconds.

Combined Voluntary Effort and Stimulation Force Task

These tasks started in a similar manner to the voluntary effort maintenance task, with the participant generating either 20% or 50% of vFmax. When the participant had maintained a steady force for 1.5s, visual force feedback was removed and they were instructed to “keep pushing in the same way you’re pushing, even when the stimulation comes on” in order to have them maintain the same level of effort during stimulation, similar to the ‘do not intervene’ instructions in (21–23). The same stimulation parameters used in the Stimulation Force Task were applied. An example of this combined effort and stimulation force trial is shown in Figure 2.

Figure 2.

Example of a combined low voluntary effort and stimulation trial performed by subject 1.

Data Analysis

For each participant in each position, the change in force in the stimulated direction after feedback removal and/or stimulation onset was calculated. Separately, for each participant and position, the stimulation alone trials where the participant remained relaxed were all ensemble averaged from the time when the stimulation ramp was started to two seconds after that point, i.e. 1.5 s after the stimulation reached its plateau in order to determine the change in force in the direction of the stimulation. The maximum force magnitude generated during that two second window was considered the maximum stimulated force. Then, using that stimulation alone ensemble averaged dataset, two time points were found. The first time point was when the stimulation alone force reached 5% and the second was when the stimulation alone force reached 90% of the maximum stimulated force. Using these two time points, the change in force in the direction of the triceps stimulation alone was calculated with respect to when feedback was removed and stimulation was turned on. The change in force was calculated for all three conditions as the difference between the forces at the points where the stimulated force was at 5% and 90% of its maximum stimulated force. To reduce variance in each trial from changes in voluntary effort while feedback was removed and the arm was being stimulated, the force at 90% was averaged together with the forces from 85%–95%. To compare force changes caused by stimulation across participants and arm positions, the changes in force were normalized by the maximum voluntary force that each participant volitionally generated in the direction of stimulation at each arm position.

The normalized force increments generated by the stimulation were compared across different levels of initial voluntary effort to assess whether the dependence on initial voluntary effort was statistically significant (p<0.05). An ANOVA was used to compare the change in force using initial effort level (none, low, and high) and position (near and far) as factors while blocking for participant. If values were statistically significantly different, the analysis was repeated while including participants as a fixed factor. Then the Tukey-Kramer comparison of means was used to determine statistical difference between force increments for separate factors. To evaluate the effects of the change in force during the voluntary force maintenance task’s effect on the combined task, the average change in force during the voluntary force maintenance task using the time points described above was subtracted from each of the combined voluntary effort and stimulation force tasks. The analysis described above was repeated comparing the stimulation alone values to the combined effort and stimulation values minus the average of the change during the voluntary force maintenance task.

Results

At all three levels of voluntary effort, the addition of stimulation increased the force, as illustrated by the superimposed individual trials from one participant shown in Figure 3a. However, the force increment caused by stimulation decreased as the voluntary force (effort) increased, as shown by the superimposed incremental force responses in Figure 3b.

Figure 3.

Figure 3(a,b). Examples from participant 5 of the stimulation response while the participant is maintaining different levels of voluntary force (effort). 3a shows the total force in the stimulated direction before and after the onset of stimulation. 3b shows the change in force from the time point when feedback was removed and the stimulation ramp was started. Solid lines are trials where the participant is relaxed and exerting no effort. Dotted lines represent low (20%) voluntary effort during the onset of stimulation. Dashed lines represent moderate (50%) voluntary effort during the same stimulation.

To compare force changes caused by stimulation across participants and arm positions, the changes in force from stimulation were normalized by the component of the maximum voluntary force that is in the direction of stimulation for each participant and arm position. For all five subjects at both arm positions, the combined stimulation and voluntary force increased with increasing voluntary force (effort), as shown in Figures 4a and 4c. However, the size of the force increment after the onset of stimulation decreases with increasing voluntary force (effort) in all cases, except for one: subject 3 at the near position, as shown in Figures 4b and 4d.

Figure 4.

Figure 4(a,b,c,d). 4a/c – The total force after the onset of stimulation normalized by the magnitude of the maximum voluntary force in the near/far position. 4b/d – The normalized force increment after the onset of stimulation in the near/far position. The horizontal axis represents the normalized voluntary force prior to the onset of stimulation. The vertical axis shows the normalized force response. Shapes represent participants. The lines are fitted to the force responses showing the change in force as a function of the normalized voluntary force.

Using the previously described statistical model that blocks for participant and includes initial effort level and position as factors, the response to stimulation was evaluated. Main effects were effort level, position, and participant while interaction effects included two way combinations of all three main effects. All of the interaction effects were statistically significant (participants and effort p<0.001, participant and position p<0.001, effort and position p=0.029). The main effect of effort was significant (p=0.002) while participant and position were not statistically significant (p=0.186 and p=0.094). The sum of squares from the different initial effort levels is 0.41 while the sum of squares for the interaction of participant and initial effort is 0.11. R squared was 87.3%.

Data averaged across participants is shown in Figure 5. Despite changes in p values, the same effects were significant when the change in force during the voluntary effort maintenance task was subtracted from the combined force increment. Post-hoc analyses provided more insight into which positions/effort levels were statistically significant for individual participants (p<0.05). In the near position, participant 3 was not statistically significantly different from stimulation alone at low or high effort and participant 4 was not significantly different at low effort as shown in Figure 4b. Subjects 2, and 5 were statistically different in all positions and at all effort levels. Participant 1 was not statistically significantly different in the far position at low or high effort, though the data exhibited a similar downward trend and an ANOVA applied to that set of data revealed significant differences at the two effort levels. Table III shows that in seven of the ten instances, the increment in force during stimulation alone was significantly greater than during the combination of low effort and stimulation (p<0.05) with a mean difference between the force increments for the two tasks of −18.0%. The force increment in six of those seven cases was still significantly greater when the change in force during the voluntary effort maintenance task was subtracted. At moderate (50%) effort, eight of the stimulation alone force increments were significantly greater than combined effort and stimulation force increments (p<0.05) with a mean difference of −44.9%. All eight of the force increments were still significantly greater when the change in force from the voluntary effort maintenance was subtracted from the combination trial.

Figure 5.

Averaged stimulation response and standard deviations across participants.

Table 3.

Summarized results showing the number of instances where the change in force in the combined voluntary effort and stimulation task is different than the change in force during stimulation alone as well as the number of instances when the difference between those changes is different than the change in the maintained voluntary force during voluntary effort alone without either stimulation or feedback.

	Change significantly different from stimulation alone (p<0.05)	Combination change as a proportion of stimulation alone	Mean combination change difference	Mean voluntary change (no stimulation)	Combination increment minus voluntary maintenance significantly different from stimulation alone (p<0.05)
Low Effort Combination	7/10	82.0%	−18.0%	−2.8%	6/10
Moderate Effort Combination	8/10	55.1%	−44.9%	−9.0%	8/10

The dependence of the normalized force increment on the voluntary component of the force (effort) is described well by the equation (Ft-Fv)/Fs=(1+a*Fvn) where Ft represents the total force after the onset of stimulation, Fv represents the voluntary force prior to the onset of stimulation, Fs is the force due to stimulation alone, Fvn is the normalized voluntary force (magnitude of the voluntary force normalized by the maximum voluntary force (Fvn=Fv/vFmax)) and a is the slope parameter characterizing how the force increment changes with voluntary effort. This equation is similar to the model used by Perumal et al (17), except with the assumption that the change in force is linear over the range measured. The equation was fit to each participant’s data separately, forcing the line through 1 on the vertical axis. The average value for the slope a is −0.23 (s.d. = 0.13). The mean difference in slope (a) between the near and far positions was 0.04, but a paired t-test did not indicate statistical significance (p=0.63).

While Figures 4b, 4d, 5 and Table III show a decrease in the response to stimulation as voluntary effort increases, in all instances the stimulation response was statistically greater than zero (p<0.05), indicating that stimulation increased the total force in all cases.

The target force directions among the set of participants covered a range of directions, as shown in Table II. Across the participants, the target force directions followed a trend, with the more medial preferred directions (+z rotations) having smaller downward components (+y rotations). The directional rotations are with respect to the coordinate frame shown in Figure 1A. None of the participants had a target force direction in the same direction as the stimulated triceps force.

While the data set is incomplete for calculating the full set of inverse torque calculations, we can make some general observations about the directions of forces being generated by the elbow and shoulder based on the directions of endpoint forces. The average angle between forces generated by the shoulder and elbow in the near and far positions were 62° and 94° respectively.

Discussion

The above results support the first hypothesis that electrical stimulation can produce greater forward force even in the presence of some levels of voluntary effort. This increase in force should lead to increased movement produced with a neuroprosthesis. However, the results also support the second hypothesis that the force response depends on the effort level of the participant. The results also indicate that there may be individual variations in the relative dependence on voluntary force. On average, the force increment caused by stimulation was reduced by 15% in the presence of 20% voluntary effort and by 36% in the presence of 50% voluntary effort. While the results indicate that the total force is not a simple summation of voluntary and stimulated forces, the differences between individuals implies that a more targeted set of experiments to specifically determine the mechanisms that result in these changes and what is causing the variations across individuals would be useful.

These results are similar to those reported in able bodied participants. Langzam et al (24) studied the tibialis anterior, and observed that higher stimulation levels were necessary to produce the same total torque levels when superimposed on higher volitional torques. Their experimental and analysis methods were based on an explicit assumption that the voluntary and stimulated motor neurons did not overlap, as well as the quantitative estimation of voluntary contraction levels from the EMG of the stimulated tibialis anterior, whereas our experimental and analysis methods did not. Perumal et al (17) observed that stimulation superimposed on voluntary effort decreased the stimulation response with increasing voluntary effort in the quadriceps. They assumed that motor unit recruitment overlap was the only mechanism contributing to the change in force. We studied a different muscle group in an impaired group of participants, and also observed significant variations between participants and similar decreases in force increment. Both the current study and the earlier studies observed less than linear summation of the stimulated and voluntary force responses that was force dependent. The similarity of the conclusions in the perspective of the differences in approach gives increased confidence in the robustness of the effect.

These experiments did not examine the mechanism creating these changes, but there are a few that could be causing this difference. One possible mechanism contributing to the reduction in incremental force is that some motor units are activated both by voluntary effort and by electrical stimulation (overlap). If the motor units being recruited by the stimulation overlap with the volitionally activated motor units in the same muscle, the increment in force from stimulation will be limited to the additional recruitment of inactive motor units, and the possible increase in excitation frequency of volitionally activated motor units. Even though the brain recruits in order of smallest to largest motor units (25) and it is generally accepted that neuromuscular electrical stimulation has been shown to recruit motor units in the opposite order using direct nerve stimulation (26), the response to surface stimulation has also been observed to be more mixed and unselective (26, 27). Despite the differences in recruitment order, rate modulation continues after full recruitment with the result that full voluntary recruitment can occur well before maximal force is achieved (28–30). This interpretation is consistent with the observation (Figures 4b and 4d) that the reduction in the incremental force is larger at higher effort levels, since we would expect both greater overlap of recruitment, and higher motor unit firing rates at the higher force level. It has been demonstrated during fatigue tests employing maximal voluntary activation that short trains of supramaximal stimulation do not produce a force increment (31). Thus, we would not expect an increase in force if either stimulation or voluntary effort alone was fully activating the muscle. If the primary mechanism for the decrease in force increment is motor unit activation overlap, variations in the extent of overlap between the directions of the volitional force and stimulated force, including a lack of overlap in the near position for participant three, could partially explain the variations in force dependence across participants. In addition, participant three’s preferred voluntary force direction in the near position was the most different from the stimulated direction of all of the combinations of participants and positions. It is possible that overlap was low because the shoulder was contributing significantly to the force output and the elbow was not contributing much extension.

Another possible mechanism is that stimulation elicits reflexes that are modulated by effort. Surface stimulation would be expected to excite both cutaneous and proprioceptive afferents, and there are widespread reflexes in the arm following stroke that are augmented by effort (23, 32, 33). Reflexes could either enhance or reduce the total force. Since the changes in force increment are not significantly different (p<0.05) between the near and far arm positions based on the slope of the model, this indicates that a position dependent reflex effect is not significantly contributing to the difference in force increments. It should be noted though that this study is not powered to account for β error, so while we cannot say that the force increments for the two positions are significantly different we also cannot confidently say that they are the same.

Another explanation for the force increment caused by stimulation decreasing with increasing effort is a failure to maintain voluntary effort in the presence of both stimulation and a loss of feedback. However, our analysis above indicates that failure to maintain effort due to loss of feedback is unlikely to be the primary explanation. While there was a decrease in the voluntary force generated when feedback was removed, the difference between the changes during stimulation alone and combined voluntary effort and stimulation was significantly greater than the change in force when feedback was removed during voluntary effort alone. This supports the hypothesis that the changes are a result of more than just the feedback removal during the task. There could also be a change in voluntary effort as a result of participants feeling and responding to the sensation of electrical stimulation. Participants participate in multiple training sessions before the test sessions to allow them to become comfortable with the sensation of electrical stimulation, however the current study cannot statistically verify that sensation is not a mechanism for the decrease in force increment.

The variance in combined force reduction across subjects could partially be a reflection of the variance in synergy pattern expression. The maximum net force is partially limited by antagonist muscle co-contraction in the synergy patterns and participants potentially being unable to generate maximal contractions post-stroke. Both of these would prevent knowing the maximum forces produced by individual muscles. The target forces and stimulated forces are scaled to the maximum net force, not the actual maximum individual muscle force. Thus, variation of the intensity of synergies across subjects could lead to variation of effects. Similarly, if motor unit overlap is the mechanism responsible for the reduction, variation in recruitment and rate modulation patterns across subjects (34–36) could lead to variability in the amount of voluntary effort that is replaced by electrical stimulation.

The decrease in stimulation response with increasing voluntary effort is significant statistically for effort and the interaction between participants and effort, with a weaker dependence on the interaction. The sum of the squares of the effort is 0.41 as compared to 0.11 for the interaction between effort and participant indicating that effort is generally the stronger contributor. In future studies, percutaneous stimulation, measurement of EMG levels across muscles and an experimental design that simplifies force analysis could provide insight into the mechanisms responsible.

The experimental design could be improved by standardizing the arm position/orientation by joint angle rather than end point position. This would allow choosing a consistent arm orientation across subjects making it easier to back calculate joint torques. We cannot rule out contributions of shoulder torques to the endpoint force. Calculating the joint torques would enable assessments of the force contributions of different joints, similar to the method used by Keller et al (9). The current experimental design could not distinguish between forces generated by shoulder and elbow torques as previously mentioned in the discussion. Inverse joint torque calculations are sensitive to small changes in limb configuration, and we did not measure the horizontal arm orientation with adequate precision. As described in the setup, the small forces transmitted through the arm support would also affect the calculated joint torques, but the impact of these forces in our experimental setup is less than the impact of errors in the shoulder/elbow positions.

Similarly, it would be ideal to have the same voluntary force target direction for each participant. Targeting joints that have fewer muscles crossing them would further simplify the interpretation of the data. Lastly, able bodied trials incorporating these design changes would help establish a baseline before evaluating these summations in groups who have altered reflexes and central inputs. These changes would provide greater understanding of the mechanisms, which is important in the future design of neuroprosthesis to exploit voluntary force augmentation.

The relationship between the response to stimulation and the underlying level of voluntary effort has potential ramifications for the design of stroke neuroprostheses that integrate a user’s voluntary effort with electrical stimulation. While there is a decrease in the force increment as voluntary effort increases, the force increment is still substantial (64%) at moderate effort for a stimulation level that produces 20% vFmax when exerting no effort. Partial effort does not completely block the stimulation response indicating that voluntary effort can be used as a command signal and FES can still augment movements as long as effort level is considered while designing post-stroke neuroprosthesis control schemes. Limiting effort to limit the expression of synergy patterns will allow FES to have a greater effect. This type of neuroprosthesis would use EMG recorded at low effort levels as the command signal for large levels of stimulation that would be the primary movement generators. The user’s focus would be to generate suitable EMG levels for a command signal rather than attempting to generate maximum effort. As observed in the results from Keller et al. (9), shoulder abduction generates elbow flexion that can be difficult to overcome by elbow extensor stimulation. Our preliminary studies support the hypothesis that reducing voluntary effort during reach and hand opening and supplementing that effort with stimulation can generate hand opening even during reach (37). By generating most of the shoulder abduction with stimulation instead of voluntary effort, elbow and hand stimulation may be able to have a greater net effect.

One approach in designing neuroprostheses for stroke is to rely as much as possible on the user’s residual voluntary ability to move and only add stimulation to supplement the voluntary effort as needed. This approach minimizes the extent of intervention (number of channels, intensity of stimulation) and maximizes the potential therapeutic benefit of requiring greater voluntary control. However, there are potential benefits to reducing voluntary effort and increasing the use of stimulation in a neural prosthesis. First, lower voluntary effort reduces the intensity of synergistic contraction patterns (6). Synergy patterns scale somewhat proportionally to effort, increasing the forces in multiple muscles (5). In order to decrease the undesired synergy response and maximize the stimulation response, it is beneficial to reduce the effort exerted on the part of the user. Thus, less stimulation force would be required to overcome undesired contractions of antagonists. For example, relaxation of the arm allows FES to open the hand by reducing synergistic finger flexor activation. While FES has been hypothesized to reduce antagonist contractions by reciprocal inhibition, experimental evidence does not support that hypothesis (15). Secondly, there is no loss of maximal force potential by limiting voluntary effort, since lower voluntary effort increases the forces that can be elicited by FES in the same muscle. Maximal stimulation and maximal voluntary activation produce the same force (31). Third, lower effort reduces the intensity of stretch reflexes, which slow down and limit the extent of voluntary movement in stroke (23, 33).

If the mechanism of reduction in the stimulation response as a result of increased effort is a result of overlap between activated motor units, this has another implication for the design of neuroprostheses for stroke. For systems which use EMG from stimulated muscles as part of the control signal (38), one must be aware that some of the asynchronous action potentials that were activated by the central nervous system will be replaced by synchronous action potentials activated by stimulation. Even if the synchronous action potentials (M-waves) are removed from the signal (38–41), the residual EMG is not entirely indicative of the user’s effort in that muscle.

FES, as a functional neuroprosthesis, has the potential to increase function by improving reach and hand opening (37) augmenting the gains in function that are achieved by physical rehabilitation and robotic therapies. Neuroprosthetic and therapeutic approaches are not mutually exclusive approaches to increasing either reach (42) or hand function (43) after stroke, and combined approaches may provide the best results. Assistive forces from an FES system may enable stroke survivors to utilize some of these therapies that they were previously unable to participate in. Similarly, participating in these therapies may increase volitional movement and disconnect synergy patterns (42), thereby allowing more assistance from volitional effort, providing a more robust command signal, allowing for finer movements in response to stimulation, and progressively decreasing a reliance on the stimulation.

Future stroke neuroprostheses should be designed with an understanding of the relationship between effort, stimulation level, and motor output. Doing so can allow the user to derive optimal benefit from the device. These results indicate that even in the presence of moderate voluntary effort, FES can increase post-stroke force production, decreasing the impairment on the affected side.

Conclusions

Electrical stimulation is capable of increasing endpoint force even in the presence of voluntary effort post-stroke. The stimulation response is dependent on the level of voluntary effort and to a lesser extent on the individual participant, though the contributions of different mechanisms are unknown. This change in force response should be taken into consideration in the design of future post-stroke neuroprostheses.

Abbreviations

FES

Functional Electrical Stimulation

EMG

Electromyograms

DOF

Degree of Freedom

FMA

Fugl-Meyer Motor Assessment

vFmax

maximum voluntary force