Gait & Posture 26 (2007) 482488

www.elsevier.com/locate/gaitpost

Runner-up of the Joint ESMAC and GCMAS (JEGM) 2006 Award

An exploration of the function of the triceps surae during normal

gait using functional electrical stimulation
Caroline Stewart a,*, Neil Postans a, Michael H. Schwartz b,c,
Adam Rozumalski b, Andrew Roberts a
a
Orthotic Research and Locomotor Assessment Unit, Robert Jones and Agnes Hunt Orthopaedic Hospital, Oswestry, Shropshire SY10 7AG, UK
b
Gillette Childrens Specialty Healthcare, Center for Gait and Motion Analysis, 200 East University Avenue, St. Paul, MN 55101, USA
c
University of Minnesota, Minneapolis, USA
Received 14 June 2006; received in revised form 29 November 2006; accepted 5 December 2006

This paper was selected for publication in Gait and Posture from the presentations at the Joint ESMAC and GCMAS 2006 meeting by a
Scientific Committee representing both Societies and chaired by Professor Maria Grazia Benedetti.

Abstract

Gastrocnemius and soleus have a common tendon and both are active during stance phase, where they are thought to arrest and control
tibial advance. Soleus is associated with the production of an extending moment at the knee. The two-joint gastrocnemius, which crosses the
knee joint, will have an additional contribution to the knee flexors.
Recent work using induced acceleration analysis (IAA) has demonstrated distinct differences between the actions of gastrocnemius and
soleus. This study aims to use gait analysis to provide in vivo examination of these theoretical predictions.
Functional electrical stimulation (FES) was chosen to provide a perturbation in muscle force, a close physical analogue to the theoretical
predictions of IAA. Five adult male subjects, with no gait problems, participated. Each had gastrocnemius and soleus stimulated at three
different timings during normal gait, while 3D gait data were collected. The order of testing was randomised and unstimulated trials were
randomly interspersed to act as a control.
The results show very different actions for soleus (ankle plantarflexing/knee extending) and gastrocnemius (ankle dorsiflexing/knee
flexing) in stance phase. The counterintuitive nature of the action of gastrocnemius suggests that further clinical and biomechanical
investigation into this muscles function is required. The actions of both muscles at the knee confirm published IAA predictions. In vivo
evidence such as this gives greater confidence when using model predictions. The approach adopted in this study could eventually be extended
to other muscles and patient populations.
# 2006 Elsevier B.V. All rights reserved.

Keywords: Normal gait; Gastrocnemius; Soleus; Functional electrical stimulation

1. Introduction
The triceps surae comprises the two calf muscles.
Gastrocnemius, having medial and lateral heads, originates
from the femur and inserts into the calcaneus through the
Achilles tendon. Soleus shares this common insertion point
but has its origin on the tibia and interosseus membrane. The
anatomical paths of these muscles give them plantar flexing
* Corresponding author. Tel.: +44 1691 404666; fax: +44 1691 404058.
E-mail address: Caroline.Stewart@rjah.nhs.u (C. Stewart).
moment arms at the talocrural joint. Both muscles also cross
the subtalar joint where the moment arms are small and
posture dependent [1]. Only gastrocnemius crosses the knee
joint, where it has a flexing moment arm.
Based on their similar anatomy the calf muscles are
generally thought to have similar functions, particularly at the
ankle. Active testing on the examination couch produces a
predictable plantarflexion response. Clinically, passive ankle
dorsiflexion range increases when gastrocnemius is relaxed
by knee flexion, especially when this muscle is pathologically
contracted (Silverskiold test). During dynamic activity,

0966-6362/$ see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.gaitpost.2006.12.001
 C. Stewart et al. / Gait & Posture 26 (2007) 482488
however, the roles of the gastrocnemius and soleus become
more complex. EMG data show that both muscles are active in
stance, with soleus turning on before gastrocnemius. The
delay between the reported onsets of the two muscles varies
between studies [2,3]. Various authors have described the
effects of this activity at the knee and ankle.
The general consensus is that soleus, assisted later by
gastrocnemius, restrains the forward progression of the tibia
over the talus in second rocker, hence reducing dorsiflexion
[35]. The effect at the knee is more complex. Gage [6] uses
the term plantarflexion/knee extension couple to describe
the action of soleus in advancing the ground reaction force
and producing a knee extending moment. Exaggeration of
this mechanism is proposed as a possible cause of knee
hyperextension in pathological gait [7]. Gage [6] reports a
relatively late onset for gastrocnemius in stance and does not
associate the muscle with a knee extension mechanism in
normal gait. For gastrocnemius which, unlike soleus, crosses
the knee joint, the flexing moment arm is likely to mitigate
any similar coupling which might occur.
Studies have been carried out looking at patients with calf
muscle deficiency [8] and also normal subjects after tibial
nerve blocks [8,9]. The aim was to understand the normal
function of the muscles and the effects of nerve blocks are
particularly interesting. Simon et al. and Sutherland et al.
both highlight the restraint of tibial advancement during
second rocker and question the appropriateness of assigning
a propulsive action to the muscles. There are, however, two
main limitations with the work. Firstly, the block affects
several muscles and it is impossible to separate out the
individual contributions of gastrocnemius and soleus.
Secondly, it is likely that compensation mechanisms will
be seen, even in the very early period after the block has
taken effect.
During walking the musculoskeletal system forms a
dynamic, linked chain of segments. In this situation, a
muscle does not only affect the joints it crosses. Each muscle
has the potential to affect the whole kinetic chain, which can
produce counterintuitive effects. Computer simulation
techniques are being developed to predict these whole
mechanism effects. Forward dynamics [10] allows theore-
tical movement patterns to be generated from muscle force
inputs. Induced acceleration analysis (IAA) provides a
simpler, more stable, approach which gives information
about dynamic function in terms of the potential of a joint
moment [11] or muscle force [12] to accelerate the joints and
segments of the musculoskeletal system.
Recent work using forward dynamics/IAA techniques has
yielded some surprising results for the calf muscles [13,14].
These simulations have been used to analyse the contribution
of the muscles to support (vertical acceleration), propulsion
(acceleration in the direction of progression) and the initiation
of swing phase. Both studies emphasise the importance of
looking at the individual components of the calf muscle
separately and both revealed clear differences between the
actions of gastrocnemius and soleus. An example of this is the
483
induced accelerations at the knee at midstance. Neptune et al.
[14] showed that, while soleus accelerated the joint into
extension, gastrocnemius had the opposite effect.
The implementation of IAA models is mechanically
correct, but the practical application and significance of IAA
techniques has been questioned. Chen [15] points to
differences in results and interpretation between studies.
This study treats IAA as a perturbation analysis. IAA can be
viewed as predicting the accelerations, which would result
from an incremental change in muscle force. As yet no
experiments have been carried out to demonstrate a direct
link between the predictions of the models and the response
of a real person in the laboratory. If confirmatory evidence
could be provided then IAA models could be used with
greater confidence. What is required is a technique for
producing a physical perturbation in muscle force to
compare with the theoretical predictions of IAA.
Functional electrical stimulation (FES) uses low level
electrical currents to induce muscle contractions. The
stimulus is supplied via skin mounted electrodes, usually
placed over the belly of the muscle or the nerve that supplies
it. FES is a convenient intervention in that it can be applied
randomly, with almost instantaneous but short-lived effects.
Increasing the force produced by a muscle with FES gives an
indication of the dynamic function of that particular muscle
under specific kinematic conditions. The use of FES in this
way is similar in concept to the earlier work carried out using
nerve blocks, with the difference that FES seeks to augment
rather than reduce muscle forces. FES has, however, two
main advantages. It is possible to target stimulation to
gastrocnemius and soleus separately and its rapid action,
which can be switched on and off, minimises the
development of compensations.
This study aims to use FES to clarify the dynamic
function of the calf muscles during normal gait, increasing
our understanding of the biomechanical action of the
muscles. If the results confirm the reports of recent IAA
simulation work then researchers will be able to have much
greater confidence in the predictive power of these models.
2. Subjects
Five adult male subjects consented to take part in the
study. (Ethical approval was obtained from the local
research ethics committee.) Subjects had an average age
of 38 years (2256), height of 180 cm (176183) and weight
of 87 kg (67101). No subject had any pathology affecting
their walking and this was confirmed through the normal gait
data collected.
3. Methods
Testing was only carried out on one lower limb, selected
at random. One subject specifically requested that his left leg
484 C. Stewart et al. / Gait & Posture 26 (2007) 482488

be tested, due to a previous but currently asymptomatic

injury on the right, and this was accommodated. The result
of this process was that one subject had his right leg tested
(Subject 1) and the rest their left.
One pair of self adhesive FES electrodes (Nidd Valley
Medical) was placed over the belly of the lateral gastro-
cnemius muscle. The active electrode (cathode) was placed
proximally with the indifferent electrode placed directly
below. A second pair of electrodes was placed over the soleus
muscle, below the distal end of the gastrocnemius. The active
electrode was again placed proximally. The positions of the
electrodes were based on Baker et al. [16]; however, the exact
positioning had to be dictated by palpation of each individual
subject. This was done with the subject lying prone on the
examination couch. The final position was chosen to give
strong plantarflexion for each muscle, with no noticeable
(palpable) contraction of gastrocnemius when soleus was
stimulated and vice versa. The positions of the electrodes in
one subject are illustrated in Fig. 1.
An Odstock two channel stimulator (Department of
Medical Physics, Salisbury District Hospital) was used. The
stimulator could be controlled by foot switches to provide
several different stimulation algorithms. The stimulation
waveform comprised asymmetric biphasic pulses (fre-
quency 40 Hz, maximum current amplitude 70 mA). The
pulse width was adjusted to elicit as strong a contraction as
possible while remaining comfortable enough for the subject
to walk naturally.
Two foot switches were used to control the onset and
cessation of stimulation. One switch was placed under the
heel and one under the first metatarsal head on the tested
limb. As the subjects walked barefoot the foot switches were
secured using double sided tape.
A Vicon 612 system was used for the data collection, in
combination with a single Kistler force platform. A full body
marker set (PluginGait, Vicon) was used for each subject and
an initial set of data was collected to include six clean strikes
on the force platform for the tested limb. All data were
collected at a comfortable, self-selected walking speed.
Two muscles were tested, gastrocnemius and soleus, with
three stimulation patterns (1: initial contact to foot flat, 2:
foot flat to toe off, 3: heel off to toe off) giving six test
combinations. Randomisation was used to determine which Fig. 1. Typical positions of FES electrodes.
muscle was tested first and also the order of data collection
for each timing interval. of the Vicon system, allowing the exact timing to be
Data were collected in the same fashion for each of the measured and checked. The sampling rates used were 60 Hz
six test conditions. Each trial was randomly assigned as a for camera data and 1080 Hz for analogue and force plate
stimulated or non-stimulated walk to prevent any voluntary data.
compensation or anticipation. For the stimulated walks the The data were processed using Vicon standard software
FES was only used for a single gait cycle, the one over the (Workstation/PluginGait, Polygon). The first five good trials
force platform. The application of the FES was controlled were used for subsequent analysis. A good trial was one
by the foot switches, in combination with a remote, where continuous data were available for at least one
manually operated switch. Data collection continued until complete gait cycle starting with left and right initial
at least six clean force plate strikes were obtained for the contacts, where any gaps in the marker trajectories were
stimulated and unstimulated conditions. The stimulus small (less than 10 frames) and force plate strikes were
signal was recorded through one of the analogue channels clean.
 C. Stewart et al. / Gait & Posture 26 (2007) 482488 485

4. Results
Each subject was able to tolerate a strong contraction
in both gastrocnemius and soleus. A hand held dynam-
ometer was used to estimate the moment production in
each case and the magnitudes were in the order of
510 Nm.
Unsurprisingly during gait the most marked effects were
seen for the second stimulation interval (foot flat to toe off),
where the FES burst duration was the longest. The changes
observed in the sagittal plane were also greater than those in
the coronal or transverse planes. As a result only sagittal
plane data for the lower limb during for the second
stimulation interval will be presented here.
Fig. 2 gives a set of data from one of the subjects (Subject
4). This subject illustrates the general trends most clearly. In
every graph the mean of the stimulated curve lies outside the
range of 1 standard deviation for part of the gait cycle.
Because force produces acceleration, rather than displace-
ment, there will be a time delay before peak displacement is
observed. Effects, which occur much later in the gait cycle,
however, particularly in swing phase, may be compensa-
tions, as the subject will have had sufficient time to make
volitional alterations to their gait pattern. The direct effects
of the stimulation perturbations are likely to peak around
midstance and may be interpreted as the dynamic action of
the muscle. During third rocker the force in the calf muscles
is very high. It is, therefore, less likely that a significant

Fig. 2. Sagittal plane kinematic data for Subject 4 for stimulation of gastrocnemius (left hand column) and soleus (right hand column). The band between the
dotted lines represents the interspersed unstimulated trials (1 standard deviation). The dark line is the mean of the stimulated trials.
486 C. Stewart et al. / Gait & Posture 26 (2007) 482488

Fig. 3. The differences between the means of the five stimulated trials and the interspersed five unstimulated trials for Subject 4. The differences are scaled by
the standard deviation of the unstimulated trials. The horizontal axis gives the percentage of the gait cycle (0100%). Knee flexion and ankle dorsiflexion are
positive.

change in kinematics will result from the augmentation of
force produced by the addition of FES.
For each point in the gait cycle 10 data points were
available, 5 from the stimulated trials and 5 from the
unstimulated trials. The difference between the means of the
stimulated and unstimulated trials was then calculated. This
was then scaled by dividing by the standard deviation of the
five unstimulated points. The effect of this is to make the
measured change independent of the measurement units and
directly related to the normal step-to-step variation. Fig. 3
illustrates the resulting curves for knee and ankle kinematics
of Subject 4, with the y-axis in standard deviations and the
duration of the stimulation indicated.
The upper graph shows the effect at the knee.
Gastrocnemius stimulation leads to a positive change
(greater flexion) and soleus stimulation to a negative change
(greater extension). The lower graph shows the effect at the
ankle. Again, gastrocnemius stimulation leads to a positive

Fig. 4. Action of gastrocnemius and soleus during second rocker for all five subjects. A point is marked on the graph when the mean of five stimulated traces
differed from the mean of five interspersed unstimulated traces by at least 2 standard deviations.
 C. Stewart et al. / Gait & Posture 26 (2007) 482488
change (greater dorsiflexion) and soleus stimulation to a
negative change (greater plantarflexion). The direction of
the change is interpreted as the dynamic action of the
muscle. Subject 4 clearly has opposing actions at the knee
and ankle for gastrocnemius and soleus. The action of
gastrocnemius as a knee flexor and ankle dorsiflexor is
particularly surprising.
A threshold of significance was set at 2 standard
deviations. An effect, which produces a change of this
magnitude will be large. The period of interest was taken to
be from foot flat to toe off and for this region the areas where
a 2 standard deviation change is observed are marked with a
solid horizontal line.
Fig. 4 gives the equivalent results, in summary, for all five
subjects. It shows the regions where stimulation led to a
change of at least 2 standard deviations, the sign of the change
being interpreted as the dynamic action. Each point on the
figure represents 2% of the gait cycle. The trends observed for
Subject 4 can now be analysed for the whole group.
At the knee, all subjects showed significantly increased
knee flexion when gastrocnemius was stimulated. The time
taken to reach the required threshold varied but the trend is
clear. A delay is unsurprising given the time taken for the
resultant acceleration to produce a significant displacement.
Three subjects show the opposite action for soleus, with
stimulation causing greater knee extension. Inspection of the
raw data for Subject 2 showed that the same trend was also
present throughout the stimulation period; however, the
magnitude peaked at 1.4 standard deviations and so failed to
reach the threshold. Subject 5 showed the opposite effect,
with soleus stimulation leading to increased knee flexion.
At the ankle, four subjects showed significantly increased
dorsiflexion when gastrocnemius was stimulated. Again the
time to reach the threshold varied but the trend was clear. For
Subject 4 the threshold was achieved relatively late but, as
can be seen from Fig. 3, a lower magnitude change occurred
much earlier. Subject 1 failed to produce a change in ankle
angle on gastrocnemius stimulation. Stimulation of soleus
produced increased ankle plantarflexion in all subjects.
From these data, an overall trend emerges, with
gastrocnemius and soleus appearing to have opposing
actions at the ankle and knee in second rocker. For only one
stimulation condition (soleus), at one joint (the knee), for
one subject (Subject 5) do the results directly contradict this
trend and this requires further discussion. There is also
variability in the magnitude of the response between
subjects.
5. Discussion
The results of this study appear to show antagonistic
activity in gastrocnemius and soleus during stance phase of
normal gait. The five subjects did, however, respond
differently to the stimulation and this variability is also
discussed below.
487
The data from all five subjects demonstrate that soleus
acts to restrain dorsiflexion after foot flat, in agreement with
the general consensus in the literature. When the muscle is
stimulated dorsiflexion is decreased. This action is, however,
entirely absent in this group when gastrocnemius is
stimulated. In fact, four of the five subjects show the
opposite effect. Gastrocnemius activity appears to promote,
rather than restrain, tibial advance. This role of gastro-
cnemius as a dorsiflexor is surprising, given its anatomy and
no other evidence has been found in the literature confirming
this finding. Further musculoskeletal modelling work may
help to explain these observations, in particular IAA
simulations focusing on stance phase ankle control.
At the knee opposing actions are again observed in
gastrocnemius and soleus. The common understanding of
the plantarflexion/knee extension couple is that calf muscle
activity can promote knee extension in stance phase by
advancing the ground reaction vector. Increased knee
extension is clearly demonstrated for soleus in three of
the subjects. Once again gastrocnemius seems to have the
opposite action, promoting knee flexion in all subjects. Here,
there is supporting evidence from Neptune et al.s [14]
model-based study, which also reported the two muscles
acting in opposition at the knee. It would appear that, when
the foot is on the ground, ankle dorsiflexion and knee flexion
are coupled, as are ankle plantarflexion and knee extension.
In the case of gastrocnemius the effects at the knee seem to
dominate the apparent anatomical action at the ankle.
Not all of our subjects showed the general trend described
at both joints and for both muscles. Two subjects failed to
reach the two standard deviation threshold, Subject 2 at the
knee for soleus stimulation and Subject 1 at the ankle for
gastrocnemius stimulation. Two standard deviations represent
a high threshold and in one of these cases a clear change could
be seen in the raw data (not presented here) at a lower level.
Subjects 1 and 2 had the lowest stimulation moments, as
measured with the dynamometer, which may have con-
tributed to the failure to observe an effect. The stimulation
magnitude does depend on individual subject tolerance levels.
The variability seen in Subject 5 is harder to explain. In
his case soleus stimulation appeared to lead to increased
knee flexion. One possibility is that the stimulation current
was also affecting gastrocnemius. The effect at the ankle is,
however, entirely consistent with the general trend for both
muscles. There does not seem to be a marked difference in
gait patterns between the subjects, which were consistently
within normal limits.
Overall, it is likely that the variability observed arose
from differences in the magnitude of the stimuli, the
electrical isolation of the muscles, the mass properties of the
limb segments, the individual musculoskeletal geometry or
tissue interconnections. It will be interesting to see whether
the differences between subjects are also reflected in their
individual IAA results. A larger study would be necessary to
explore the extent of variability within the normal
population and the factors influencing it.
488 C. Stewart et al. / Gait & Posture 26 (2007) 482488
Despite the variability, a general trend emerges and it is
interesting to explore the implications of the observations
made. Gastrocnemius and soleus, both inserting into the
Achilles tendon, are working apparently antagonistically at
two adjacent joints. It is difficult to think of another similar
example. Given that the muscles work synergistically in
much of stance phase how could this action be explained?
The main effect would appear to be loading of the Achilles
tendon in second rocker so it is possible to speculate that the
muscles may be storing elastic energy for release during
third rocker.
It is difficult to draw any firm conclusions about
pathological mechanisms from this work using normal
subjects. The calf muscles are, however, often associated
with the pathological gait patterns and implicated in the
development of equinus or calcaneus deformities of foot and
ankle and crouch or hyperextension patterns at the knee. It is
tempting to conclude that equinus with hyperextension can
be blamed on overactivity of soleus (or weakness of
gastrocnemius) and crouch and calcaneus on overactivity of
gastrocnemius (or weakness of soleus). Unfortunately such
simple conclusions are often comfounded by the complex
interaction of many factors, both primary pathology and
secondary compensations.
The work described here for normal subjects will need to
be repeated for a range of different pathologies. It will be
interesting to see what FES stimulation and IAA modelling
reveal in patient groups with weakness, control deficits and
skeletal deformities.
The conclusions of this study need to be considered in the
light of its limitations, some of which have already been
discussed in the context of the variability observed. When
stimulating muscles it is possible to define precisely the
stimulation timing and waveform. It is, however, difficult to
correlate this directly with muscle force, which develops
more slowly. Ideally stimulation would produce short,
discrete, instantaneous step changes in muscle force
allowing the effects to be assessed at precisely controlled
stages of the gait cycle. Experience from this study indicates
that this is unrealistic. It is also possible that soft tissue
interconnections can cause attenuation of the force between
origin and insertion. This is a particular risk for gastro-
cnemius where the electrodes were placed high to achieve
isolation. Testing on the couch provided reassurance here, as
gastrocnemius stimulation produced much stronger plantar-
flexion than knee flexion.
The stimulation current will also cross boundaries,
affecting other adjacent muscles. Soleus, in particular,
cannot be totally isolated but we feel it is unlikely that any
cross-activation produced a contraction strong enough to
influence the results in most cases, with the possible
exception of Subject 5. The different results obtained from
gastrocnemius and soleus provide some reassurance that
satisfactory isolation has been achieved. Stimulation of the
common peroneal nerve could produce a flexor pattern, as
observed for gastrocnemius. It is, however, very unlikely
that sufficient current reached the nerve to have any effect
and no dorsiflexion, or tension in the dorsiflexors or
hamstrings was observed during testing on the couch.
This study has cast new light on the role of the triceps
surae muscles in normal gait. Evidence is presented for
gastrocnemius and soleus having antagonistic actions at
ankle and knee, in at least some normal subjects. The role of
gastrocnemius, which produced dorsiflexion in four out of
five subjects, is particularly surprising. FES stimulation
provides a sudden perturbation in muscle force, making its
results suitable for direct comparison with those from
models based on IAA. This study has provided experimental
evidence for the differing roles of gastrocnemius and soleus
revealed by IAA and reported in the literature.
References
[1] Klein P, Mattys S, Rooze M. Moment arm length variations of selected
muscles acting on the talocrural and subtalar joints during movement:
an in vitro study. J Biomech 1996;29(1):2130.
[2] Sutherland DH. An electromyographic study of the plantar flexors of
the ankle in normal walking on the level. J Bone Joint Surg Am
1966;48(1):6671.
[3] Perry J. Gait analysis: normal and pathological function. Thorofare,
NJ, USA: SLACK Incorporated; 1992.
[4] Bleck E. Orthopaedic management in cerebral palsy. Oxford: MacK-
eith Press; 1987.
[5] Sutherland DH, Olshen R, Biden E, et al. The development of mature
walking. Oxford: MacKeith Press; 1988.
[6] Gage JR. Gait analysis in cerebral palsy. Oxford: MacKeith Press;
1991.
[7] Chambers HG, Rose J. Dynamic electromyography. In: Chambers HG,
Rose J, Gage JR, editors. The treatment of gait problems in cerebral
palsy, Cambridge: MacKeith Press; 2004. p. 13445. Chapter 9.
[8] Simon SR, Mann RA, Hagy JL, Larsen LJ. Role of the posterior calf
muscles in normal gait. J Bone Joint Surg Am 1978;60(4):46572.
[9] Sutherland DH, Cooper L, Daniel D. The role of the ankle plantar
flexors in normal walking. J Bone Joint Surg Am 1980;62(3):35463.
[10] Anderson FC, Pandy MG. Dynamic optimization of human walking. J
Biomech Eng 2001;123(5):38190.
[11] Kepple TM, Siegel KL, Stanhope SJ. Relative contributions of the
lower extremity joint moments to forward progression and support
during gait. Gait Posture 1997;6:18.
[12] Schwartz M, Lakin G. The effect of tibial torsion on the dynamic
function of the soleus during gait. Gait Posture 2003;17(2):1138.
[13] Hof AL, Otten E. Assessment of two-dimensional induced accelera-
tions from measured kinematic and kinetic data. Gait Posture 2005;22
(3):1828.
[14] Neptune RR, Kautz SA, Zajac FE. Contributions of the individual
ankle plantar flexors to support, forward progression and swing
initiation during walking. J Biomech 2001;34(11):138798.
[15] Chen G. Induced acceleration contributions to locomotion dynamics
are not physically well defined. Gait Posture 2006;23(1):3744.
[16] Baker LL, Wederich CL, MeNeal DR, Newsam C, Waters RL.
Neuromuscular electrical stimulation: a practical clinical guide.
Downey, CA: Ranchos Los Amigos Rehabilitation Centre, Ranchos
Los Amigos Hospital; 2000.