Effect of Whole-Body Vibration on Neuromuscular Activation of Leg Muscles During Dynamic Exercises in Individuals With Stroke

Abstract

Liao, L-R, and Pang, MYC. Effect of whole-body vibration on neuromuscular activation of leg muscles during dynamic exercises in individuals with stroke. J Strength Cond Res 31(7): 1954–1962, 2017—This study examined the leg muscle activity during exposure to different whole-body vibration (WBV) intensities while performing various dynamic exercises in patients with chronic stroke. Thirty patients with chronic stroke performed a series of dynamic exercises under 3 conditions: (a) low-intensity WBV (peak acceleration: 0.96 units of gravity of Earth [g]), (b) high-intensity WBV (1.61 g), and (c) no WBV. Neuromuscular activation was recorded with surface electromyography (EMG) on bilateral biceps femoris (BF), vastus lateralis, tibialis anterior (TA), and gastrocnemius (GS) in both legs and was reported as EMG_rms (root mean square) normalized to % maximal voluntary contraction. The EMG amplitude of all tested muscles was significantly increased by adding WBV during dynamic exercise (p ≤ 0.05). The EMG amplitude of BF, TA, and GS during exposure to high-intensity WBV was significantly greater than low-intensity WBV (p ≤ 0.05). The increase in EMG amplitude caused by WBV was exercise dependent in GS and TA (p ≤ 0.05). The EMG response to WBV in GS and BF in the affected leg was significantly greater than the corresponding muscles in the unaffected leg (p ≤ 0.05). The extent of WBV-induced muscle activity was dependent on the dynamic exercise, WBV intensity, and muscle trained among patients with chronic stroke.

Introduction

A wealth of literature has investigated how whole-body vibration (WBV) influences the neuromotor function in both young and older adult populations (^{4–7,9,19,31}). It has been demonstrated in surface electromyography (EMG) studies that adding WBV to exercise regimen can significantly augment the level of muscle activation during static and dynamic exercise in young adults (^{1,5,8,16,17,35,38,42}), suggesting that WBV may have potential impact on neuromuscular function after long-term training. A number of meta-analyses have indeed revealed some positive impact on postural control, muscle strength, and mobility in elderly population associated with the use of WBV (^22,23,39). However, the most optimal WBV parameters for improving various clinical outcomes are yet to be identified (^22,23,39). Individuals with stroke may be beneficiaries of WBV therapy, as they also sustain deficits in muscle weakness (¹¹), balance, and functional mobility (^13,25,40).

Increasing research effort has investigated the efficacy of WBV therapy in the stroke population in the past few years. Recently, a systematic review by Liao et al. (²⁶) revealed inadequate evidence to refute or support the clinical application of WBV therapy in enhancing motor function in the stroke population, primarily because of the conflicting results reported. It seems that more foundational issues have to be more thoroughly investigated before further randomized controlled efficacy studies are conducted (²⁶). One of these issues pertains to how different WBV intensities affect the extent of neuromuscular activation, and how the muscle responses to different WBV intensities vary with different exercises performed (i.e., static vs. dynamic exercises). Liao et al. (^27,28) demonstrated that EMG amplitudes in 4 major leg muscle groups, namely, gastrocnemius (GS) and tibialis anterior (TA), biceps femoris (BF), vastus lateralis (VL) muscles, were significantly augmented by applying WBV (by up to 6.5–25.0%) during various static exercises among individuals with chronic stroke, and that the muscle responses to WBV were quite similar in both the affected and unaffected lower limb. However, the extent to which leg EMG activity is enhanced by WBV during dynamic exercises in individuals with stroke has never been investigated. This is an important issue because dynamic muscle strength is impaired after stroke, and is strongly correlated with gait performance, balance, and activity participation (^12,33). Therefore, dynamic exercises, such as semi-squat, deep squat, forward lunge, and single-leg standing were also often used in WBV training (^24,34,41).

This study aimed to examine the EMG responses in the VL, BF, TA, and GS of both legs during exposure to different WBV intensities while performing various dynamic exercises among individuals with chronic stroke. Knowing how the EMG responses vary with different WBV intensities and exercise conditions can help determine the optimal WBV intensities and choice of exercise to induce the strongest neuromuscular activation, and is thus essential for designing effective WBV exercise training protocols for individuals with stroke. The hypotheses of this study were as follows: (a) higher WBV intensity would lead to greater EMG amplitude of the tested muscles during different dynamic exercises; (b) the changes in EMG amplitude caused by WBV would be associated with the dynamic exercise performed; (c) the effects of WBV on EMG amplitude on the affected and unaffected side would be similar.

Methods

Experimental Approach to the Problem

A one-group experimental study with cross-over design was undertaken to examine the EMG amplitude of bilateral VL, BF, TA, and GS among individuals with chronic stroke during exposure to various WBV intensities when simultaneously performing different dynamic exercises (Figure 1). Therefore, the dependent variables were the EMG amplitudes of the 4 muscle groups on each side, whereas the independent variables were WBV intensity and exercise.

Figure 1.: Exercise protocol. For exercise (A–E), the feet were placed at shoulder width, and the trunk was in upright position. A) Dynamic semi-squat: bilateral knees flexion and extension at a range between 10° and 45°. B) Dynamic deep squat: bilateral knees flexion and extension at a range between 10° and 70°. C) Dynamic forward weight shift: bilateral knees maintained at 10° flexion, leaning body forward so that the heels came off the platform as much as possible, then return to original position. D) Dynamic backward weight shift: bilateral knees maintained at 10° flexion, leaning body backward so that the forefoot came off the platform as much as possible, then return to original position. E) Dynamic weight shift side to side: bilateral knees maintained at 10° flexion, shirting body weight onto the affected side, and then shifting onto the unaffected side. F) Dynamic forward lunge: leaning body forward and weight shifting onto the affected leg as much as possible, flexion and extension of the affected knee at a range between 10° and 45°. Repeat after switching the positions of the 2 legs.

Subjects

Sampling

Between June 2013 and March 2014, participants with stroke were recruited from a stroke patient organization in the community through convenience sampling. The following eligibility criteria had to be met for inclusion: diagnosis of hemispheric stroke in chronic stage (i.e., onset >6 months ago), community-dwelling, having motor impairment in the affected leg, as demonstrated by the total leg and foot score ≤13 based on the Chedoke-McMaster Stroke Assessment (Impairment Inventory) (¹⁴), and capable of performing the 6 dynamic exercises (Figure 1) specified in this study. Exclusion criteria included low back pain, spinal diseases, severe cardiovascular conditions (e.g., cardiac pacemaker), peripheral vascular disease, severe musculoskeletal conditions (e.g., rheumatoid arthritis), vestibular dysfunctions, kidney or bladder stones, or pregnancy.

Sample Size Estimation

Previous studies showed that WBV significantly augmented the EMG amplitude of various leg muscle groups (GS, TA, BF, and VL) in the paretic leg and nonparetic leg during static exercises, yielding large effect sizes (f = 0.66–1.85) (^27,28). We took a more conservative approach and assumed a medium to large effect size (f = 0.35). After accounting for a 10% attrition rate, we aimed to recruit a minimum of 27 participants to yield a power of 90% at a significance level of 0.05.

Ethical Approval

Ethical approval was obtained from the institutional review board of the university, and the Declaration of Helsinki was followed in all experiments. Written informed consent was obtained from each participant before any data collection took place.

Demographic Characteristics

Thirty participants (9 women; mean ± SD age: 56.8 ± 10.1 years [age range = 29–77 years]) completed all experimental procedures (Figure 2). None of the participants were younger than 18 years. Overall, the participants had moderate motor impairment (median Chedoke-McMaster Stroke Assessment lower extremity score = 7, interquartile range = 6.8–9) (Table 1).

Figure 2.: Study flow chart. Thirty individuals with stroke completed all experiments. WBV = whole-body vibration.

Table 1.: WBV testing protocol.*

Procedures

Whole-Body Vibration Intensities

All participants underwent a single session of WBV on a vibration platform (Jet-Vibe System; Danil SMC Co. Ltd., Seoul, Korea) that delivered uniform synchronous oscillations at frequencies of 20–55 Hz with fixed amplitudes associated with each frequency. We used synchronous WBV in this study, as intended to induce higher level of neuromuscular activation than side-alternating WBV (³⁷). All experiments were conducted in a university exercise performance laboratory.

All participants were exposed to 3 WBV conditions: (a) low-intensity WBV (20 Hz frequency with 0.60 mm amplitude, peak acceleration: 0.96g, where g represents Earth's gravity at 9.8 m·s⁻¹) (²⁰); (b) high-intensity WBV (30 Hz, 0.44 mm, 1.61g); and (c) no WBV (Table 2). The sequence of WBV intensity was randomized through drawing ballots. A triaxial accelerometer was used to verify the output of the WBV device (Model 7523A5; Dytran Instruments Inc., Chatsworth, CA, USA).

Table 2.: Characteristics of participants (n = 30).*

Exercise Conditions

Participants performed 6 dynamic exercises for each WBV intensity (Figure 1). These exercises involved the use of major leg muscles that are essential for daily functions such as balance, transfers, walking, and stairs management (^{2,12,15,30,33,36}), and were used in previous WBV research studies (^24,34,41). Before actual data collection took place, participants were given the opportunity to practice the exercises until they were able to perform them accurately and consistently. Typically, only 5 repetitions (1 repetition: 1.5-second down and 1.5-second up) for each exercise were needed to become familiarized with the movement and rhythm required. A rest period of 1 minute was given after the practice of each exercise before data collection. A metronome was used in conjunction with verbal commands to guide the rhythm during each repetition of exercise. For actual data collection, each exercise was 45 seconds in duration for each set (i.e., 15 repetitions), and 3 sets were performed (¹⁶). A 10-second pause was given between each set of the same exercise. A rest of 1 minute in duration was given after the 3 sets were completed. In addition, after all 6 exercises for a particular WBV intensity had been completed, another rest of 5 minutes in duration was given. The sequence of the 6 exercises was determined randomly by drawing ballots before actual EMG measurement. An electrogoniometer (Type XM180; Penny and Giles Biometric Ltd., Blackwood, Gwent, United Kingdom) was used to monitor the knee angle changes during exercises. This was performed to ensure that the knee flexion and extension movements required in the prescribed exercises were done properly. Participants were asked to hold onto the handrail of the vibration machine gently for maintaining equilibrium only. The whole session was monitored by the researcher to ensure the movements were done properly. Participants were instructed to report to the researcher if any fatigue was experienced. The rate of perceived exertion (RPE) (³) was monitored before and after each exercise. If fatigue or RPE ≥15 was reported, a longer duration of rest period would be provided. The testing would not resume until the subjects felt ready again and RPE returned to ≤8 (i.e., no exertion at all or extremely light exertion).

Measurement of Neuromuscular Activation

Surface EMG was recorded form the VL, BF, TA, and GS on both sides during exposure to the 3 WBV protocols. After skin preparations, bipolar bar electrodes were attached to the skin atop the respective muscles (Bagnoli EMG system; Delsys, Inc., Boston, MA, USA), according to standardized guidelines (¹⁸). The ground electrode was placed on the head of the fibula in the affected lower limb.

All participants underwent maximal voluntary contraction (MVC) testing for the 4 muscle groups on each side in sitting. The hip and knee joints were placed in 90° flexion. For each muscle, 3 MVCs were attempted, with an interspersed 1-minute break between each attempt. For each muscle, the peak EMG amplitude of each trial was averaged to yield the mean. This mean value was then used to normalize the EMG data acquired from the WBV exercise trials. Analysis of the EMG data generated from the 3 MVC trials demonstrated very good test-retest reliability (intraclass correlation coefficient_2,1 = 0.85–0.98).

Electromyographic Data Analysis

A sampling rate of 1000 Hz was used (Bagnoli-8; DelSys, Inc.,), and the EMG data were stored directly to hard disk for offline analysis. A 20–500 Hz band-pass Butterworth filter was used to process the EMG signals (LabView software version 7.0; National Instruments Corp., Austin, TX, USA). We had visually inspected the EMG data and performed some frequency spectrum analyses. Some noise associated with the frequencies of the mechanical WBV signals was identified. Therefore, to improve the signal-to-noise ratio, the EMG signals at the same frequencies as the WBV used (20 and 30 Hz) and their common harmonic frequency (60 Hz) were selectively removed from the data (MyoResearch XP, Master Package version 1.06 software; Noraxon USA, Inc., Scottsdale, AZ, USA) (²⁷). After full-wave rectification of the signals, the EMG root mean square (EMG_rms) was then computed in 100-millisecond windows around each data point (¹). For all trials, the data acquired during the middle 30 seconds were selected for computation of the EMG_rms values (¹⁶). The EMG_rms was normalized to the peak EMG amplitude recorded in MVC (i.e., %MVC). For each unique combination of WBV intensity and dynamic exercise, the data of the 3 trials were averaged to provide the mean EMG_rms (in %MVC) for subsequent analysis.

Statistical Analyses

The dependent variable was normalized EMG_rms measured from the 4 muscles in both legs. Four separate 3-way repeated measures analysis of variance (ANOVA) models (3 WBV intensities; 6 exercises; paretic leg vs. non-paretic leg) were used to examine the EMG data of the 4 leg muscles respectively (IBM SPSS software version 20.0; IBM Statistics, Armonk, NY, USA). This was followed by contrast analysis with Bonferroni correction if the ANOVA model generated significant results. A significance level of p ≤ 0.05 was used for ANOVA. A more stringent p < 0.01 was used for post hoc analysis because of the increased risk of making a type I error due to multiple comparisons.

Results

Complete data sets of 30 participants were obtained. None of the subjects reported fatigue or RPE ≥15 during testing.

Three-Way Analysis of Variance

The intensity × exercise × side interaction was not significant for the VL (F_{5.496,150.344} = 0.642, p = 0.683), BF (F_{4.460,156.628} = 1.073, p = 0.376), TA (F_3.672,94.615 = 1.277, p = 0.285), and GS (F_{2.888,138.914} = 1.098, p = 0.353), suggesting that the interaction of all 3 factors (i.e., intensity, exercise, and side) did not determine the EMG responses. The following sections would address the 3 hypotheses of the study.

Main Effect of Vibration Intensity

The effect of WBV intensity for the VL (F_{1.634,1708.653} = 7.497, p = 0.003), BF (F_{1.354,2942.099} = 41.283, p < 0.001), TA (F_{1.584,625.499} = 15.639, p < 0.001), and GS (F_{1.282,5532.252} = 27.522, p < 0.001) were all statistically significant, suggesting that increased WBV intensity led to an increase in neuromuscular activation of these muscles (i.e., hypothesis 1). Post hoc analysis revealed that in both legs, imposing either low-intensity or high-intensity WBV stimulation during dynamic exercise resulted in significantly greater EMG activity compared with the corresponding exercises when no WBV was added (p ≤ 0.05). The EMG amplitude induced by the high-intensity protocol was also significantly greater than that induced by the low-intensity protocol for all (p ≤ 0.05) tested muscles except the VL (p > 0.05) (Table 3).

Table 3.: Effect of WBV intensity on neuromuscular activation.*

Intensity by Exercise Interaction

The TA (F_{4.809,254.511} = 2.794, p = 0.021) and GS (F_{4.410,283.364} = 2.636, p = 0.032) showed a significant WBV intensity × exercise interaction, suggesting that the muscle responses to the 3 WBV intensities were associated with the specific dynamic exercises performed (Table 3) (i.e., hypothesis 2). However, such interaction effect was not found in the VL (F_{5.923,197.072} = 0.720, p = 0.632) and BF (F_{3.780,278.255} = 1.956, p = 0.110).

Intensity by Side Interaction

The intensity × side interaction was significant for the GS (F_{1.179,2326.829} = 10.501, p = 0.002) and BF (F_{1.206,1242.577} = 14.931, p < 0.001) (Table 3), indicating that the 2 sides showed different responses to WBV (i.e., hypothesis 3) (Figure 3).

Figure 3.: Normalized EMG amplitude in different WBV and exercise combinations. The normalized EMGrms of the VL (A), BF (B), TA (C), and GS (D) in each WBV and exercise condition is expressed as %MVC. The error bars denote 1 SEMs. Six dynamic exercises were performed while undergoing 3 different WBV conditions: dynamic semi-squat (DSS), dynamic deep squat (DDS), dynamic forward weight shift (DFWS), dynamic backward weight shift (DBWS), dynamic weight shift side to side (DWSSTS) and dynamic forward lunge (DFL).* denotes significant difference between low-intensity WBV and control condition (i.e., no WBV).† indicates significant difference between high-intensity WBV protocol and control condition.# denotes significant difference between the high-intensity and low-intensity protocols. The results were significant if P < 0.002 was due to Bonferroni adjustment. NWBV = no WBV; WBV = whole-body vibration; HWBV = high-intensity WBV; LWBV = low-intensity WBV; VL = vastus lateralis; %MVC = percent maximal voluntary contraction; EMG = electromyography; EMG_rms = EMG root mean square; GS = gastrocnemius; TA = tibialis anterior; BF = biceps femoris.

Discussion

This is the first report of acute effects of different WBV intensities on neuromuscular activation during different dynamic exercises in individuals with stroke. The key findings were that (a) adding WBV to dynamic exercise caused a significant increase in EMG amplitude of leg muscles, and higher WBV intensity was associated with more pronounced increase in EMG amplitude of the BF, TA, and GS muscles; (b) the augmentation of EMG activity evoked by WBV was exercise dependent in the TA and GS; and (c) the EMG response to WBV in the GS and BF of the affected leg was greater than the corresponding muscles of the unaffected leg.

Our first hypothesis was confirmed because the EMG amplitudes were significantly augmented by the imposed WBV across all tested muscle groups. With the exception of the VL muscle, the high-intensity WBV protocol led to greater level of neuromuscular activation when compared with the other 2 WBV conditions (Table 3). Therefore, our findings are generally well aligned with previous investigations on healthy adults, which also demonstrated that a higher WBV intensity would result in more pronounced increase in EMG amplitude (^{1,5,21,29,32,35,37,38}). The actual degree of increase in EMG amplitude induced by WBV differed across various studies, primarily because of differences in WBV protocols (e.g., types of vibration, frequency, and amplitudes), exercise (e.g., type of exercise, additional load, and duration), processing methods of raw EMG data, and characteristics of the participants (e.g., stroke vs. able-bodied, severity of stroke impairment, etc.).

Liao et al. (^27,28) used the same WBV intensities as those in this study to investigate the neuromuscular activation of leg muscles during static exercise among individuals after chronic stroke. Their report showed that the EMG amplitudes of the BF, VL, TA, and GS muscles on the affected side were significantly increased up to 25.0, 11.5, 13.5, and 20.2%MVC, respectively, by adding WBV. In this study, we found an increase of EMG amplitude up to 8.9, 4.5, 3.7, and 10.7%MVC in the corresponding muscles. The results thus indicated that the influence of superimposed WBV on EMG amplitude was less remarkable during dynamic exercise than static exercise (²⁷). Whether the 2 WBV protocols used here would induce more gain in muscle strength in addition to the dynamic leg exercise protocol after a longer intervention period would need to be further tested.

This study demonstrated that the EMG amplitudes of the BF, TA, and GS during the high-intensity WBV protocol were significantly greater than the low-intensity WBV protocol during dynamic exercises. However, the actual difference in EMG amplitude induced by the 2 protocols used here was only around 2–3%MVC and 1–2%MVC for the affected and unaffected sides respectively. This was similar to what was found in previous studies that investigated EMG response in individuals with stroke during static exercise (^27,28). Further increasing the intensity from 0.96g to 1.61g led to only modest increase in EMG amplitude, highlighting that the relationship between EMG response and increasing WBV intensity was not linear.

The intensity by exercise interaction was found to be significant in the TA and GS muscles, thus partly confirming our second hypothesis. It indicated that neuromuscular activation of these 2 muscles was exercise dependent. In particular, the WBV-induced muscle activity for the TA and GS was substantially less with the dynamic backward weight-shift and dynamic forward weight-shift exercises respectively (Figures 3C, D). This was likely related to the relatively high EMG amplitude during these exercises in the control condition where no WBV was applied.

In contrast, the VL and BF muscles did not show any intensity by exercise interaction (Table 3). Some previous studies investigated the WBV intensity by exercise interaction effects in healthy adults but the results were inconsistent (^10,17,38). For instance, in the study by Di Giminiani et al. (¹⁰), different vibration frequencies did not seem to influence the EMG responses during various exercises. However, Roelants et al. (³⁸) identified a more substantial increase in EMG activity of the VL muscle when one-leg-squat exercise was performed simultaneously when WBV was applied, as opposed to high-squat and low-squat exercise. Recently, Liao et al. (^27,28) found a similar WBV intensity by exercise interaction in the VL, BF, TA, and GS muscles of both legs during static exercises in individuals with stroke. The greater neuromuscular activity in the GS and TA induced by WBV relative to the VL and BF muscles found in this study may be related to the attenuation of WBV signals as they were transmitted from the feet upward to other parts of the body. The energy of WBV signals may be attenuated by the muscles of the shank (TA and GS) before reaching the thigh (VL and BF), and therefore the difference in effective intensity of WBV delivered to the regions above the knee among the 3 WBV conditions may be less.

The third hypothesis was partly confirmed as there was a significant intensity by side interaction in the BF and GS, but not the VL and TA. It is clear from Figure 3 that the WBV-induced augmentation of EMG activity of the BF and GS was more remarkable in the affected leg than the corresponding muscles in the unaffected leg, indicating possible preferential activation of the affected side by WBV. This is in discordance with the findings of 2 previous reports, which showed similar EMG responses to WBV in the affected and unaffected lower extremities during static exercises in individuals with stroke (^27,28). The discrepancy in results may be related to the difference in EMG amplitude during static and dynamic exercises. In the control condition, the EMG amplitude in BF and GS was in the range of 10–25%MVC on the paretic side, compared with only about 5%MVC on the nonparetic side during different dynamic exercises (Figure 3). During static exercises, the corresponding %MVC values were similar in the 2 legs (2–15%MVC) (^27,28). Dynamic exercises, requiring controlled movements at a specified pace, are more challenging than static exercises, which involve only the maintenance of particular postures. Dynamic exercises may have “forced” the participants to use the paretic leg more, and thus induced a higher level of EMG activity on the same side. This may in turn enhance the response of the paretic leg muscles to WBV compared with the nonparetic leg.

The lack of significant results in the VL and TA muscles could be due to inadequate statistical power. Indeed, the corresponding p values obtained were not remote from the level of significance, at 0.112 and 0.069 respectively. Significant results might have been obtained had greater sample size been used. Although we had tried the best we could to ensure adequate power in our study, the sample size calculation was based on the data obtained from previous research on the EMG response to WBV during static exercises in individuals with stroke (^27,28). As aforementioned, the increase in EMG magnitude induced by WBV during dynamic exercises reported in this study turned out to be less pronounced than that during static exercises reported previously (^27,28). Therefore, we may have overestimated the effect size. Because of the smaller effect sizes reported here, a larger sample size would have been required to detect a significant effect.

Our findings may not be generalized to acute or subacute individuals with stroke because the participants were all individuals with chronic stroke in this study. We only compared the neuromuscular activation among 3 WBV conditions. The EMG response to WBV intensities higher than 1.61 g is unknown.

There was a possibility that fatigue may have an impact on the EMG amplitude as the testing session progressed. However, measures were taken to minimize such effects. First, the practice duration was brief and a rest period was given before actual data recording took place. Second, intermittent rest periods were provided during actual data collection. Third, the sequence of testing was randomized. Fourth, the RPE was monitored. Indeed, none of the subjects reported fatigue or RPE ≥15. Thus, the impact of fatigue on our results should be minimal.

There may be cumulative effect of WBV on EMG amplitude as the testing session progressed. As aforementioned, intermittent rest periods were given. The order of exercises was also randomized. Although we could not completely rule out the possibility of a cumulative effect, the above measures taken should have kept such effect to a minimum.

Although we tried our best to monitor the movements as the participants performed the exercises, subtle changes in posture and leg weight-bearing pattern may go undetected.

This study only evaluated the EMG activity during exposure to WBV when performing different dynamic exercises, and was not designed to examine whether the WBV intervention would translate into long-lasting changes in muscle activation or functional performance after a training period of longer duration.

Practical Applications

The WBV protocols used in this study caused a modest increase in EMG amplitude of the leg muscles on both the affected and unaffected sides during dynamic exercises, although the response in the paretic leg seemed to be greater. It is suggested that the WBV protocols used here can be used to augment leg muscle activity during functional dynamic exercises. The WBV-induced muscle activation was dependent on the choice of intensity, exercise, and their interaction. Therefore, these factors should be carefully considered when prescribing WBV intervention to individuals with stroke. In particular, dynamic semi-squat or dynamic deep-squat can be used in combination with either low-intensity or high-intensity WBV to achieve better activation of the paretic VL muscle (Figure 3A). To achieve the highest EMG level in the TA and GS muscles on the paretic side, high-intensity WBV in combination with dynamic backward (Figure 3C) and forward weight shift exercise (Figure 3D) should be used. Similarly, for the paretic BF muscle, high-intensity WBV is more effective than low-intensity WBV but the choice of exercise is less critical (Figure 3B).