Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/clinbiomech
Abstract
Objective. The objective of this study was to assess the capability of an escapement-driven inverted pendulum with springs and
damping model to estimate the eects of impairments (e.g. spasticity, muscle weakness) on the dynamics and patterns of locomotion
of children with spastic cerebral palsy.
Methods. Kinematic data of six children with spastic hemiplegic cerebral palsy and six matched, typically developing children
were collected at ®ve dierent self-selected overground walking speeds (`very slow' to `very fast'). Changes in forcing, stiness and
gravitational potentials were estimated during the stance phase of each leg according to the model's equation of motion.
Results. Signi®cantly greater stiness and decreased forcing was observed in the more aected limbs of children with spastic
hemiplegic cerebral palsy and compared to typically developing peers. The forcing term of the non-aected limb was greater than
that of the matched typically developing children.
Conclusions. Results support the claim that disabled individuals with losses in dynamic resources (stiness, muscle forcing ca-
pability) exploit and develop the remaining resources in their adapted gait patterns. It was suggested that clinical interventions
aimed at normalizing a gait pattern may be contraindicated, and that rehabilitation might be more eective if focused at the level of
dynamics.
Relevance
Pattern formation is seen as an optimal solution based on the individuals' action capabilities and dynamic properties under
environmental and task demands. This perspective could lead to the development of interventions that address these dynamic
variables with the objective of improving the functional capabilities of children with cerebral palsy. Ó 2001 Elsevier Science Ltd. All
rights reserved.
our premise that the changes in dynamic action capa- that the observed gait patterns in CP are alternative
bilities that occur as a result of upper motor neuron functional solutions to the dynamic demands of con-
disease lead to predictable changes in movement pat- tinued locomotion.
terns. The patterns are both a direct re¯ection of the
available resources and adaptations that facilitate the
use of those resources. Features of spastic cerebral palsy 2. Modeling locomotion as an autonomous inverted
include increased co-contraction [6], hyperactive stretch pendulum system
re¯exes [7], decreased muscle contractility [8], and in-
creased tendon to muscle length within a muscle [9]. The Several investigators have applied the concept of
dierences due to the neurological impairment are sug- pendular motion to describe walking [1,15]. Walking has
gestive of a stier system, that might be potentially been successfully modeled as a regular pendulum-spring
better capable of conserving elastic energy. Weakness to predict the preferred locomotor frequency of typically
and poor timing of muscle concentric contractions [10] developing adults and quadrupeds [16,17]. Recently, an
may result in a system that is capable of producing only escapement-driven, inverted pendulum with springs
limited muscle power [11]. Movement patterns of chil- model (EDIPS) with viscous damping was developed to
dren with CP may be a re¯ection of strategies that ex- account for the dynamics of the stance phase [18]. The
ploit the changed dynamic action capabilities. For following Newtonian equation of motion describes the
example, during walking raising the center of mass behavior of the model in which spring and damping
higher and dropping the body weight from an increased forces act horizontally, at a distance b from the origin,
height onto a plantar ¯exed foot would present a bio- and escapement forces act horizontally at a distance L
mechanically advantageous means to increase the load- from the origin
ing torque on a sti gastrocnemius-soleus muscle±
mL2 hd Fd
hu ; h_u L cos hu ch_u
b cos hu
2
tendon complex. The leg could then rebound in the
manner of a pogo stick. Thus, the running pattern often kb sin hu b cos hu mLg sin hd ;
1
used to describe CP gait [12] would facilitate elastic
energy return just as it does in a typical running pattern where hd is the angular displacement from downward
[4,5]. This interpretation would suggest that gait pat- vertical, hu the angular displacement from upward ver-
terns adopted by individuals with cerebral palsy are not tical, hu hd p, h_d ; hd the angular velocity and accel-
limitations but adaptations to the changed dynamics. eration, m the mass of the system, L the equivalent
To investigate the proposed changes in the dynamic distance from the axis of rotation to the center of mass
action capabilities of children with spastic hemiplegic of the body, k the system stiness, c the viscous dam-
CP a model is needed that captures the muscular con- ping, b the distance from the insertion of the spring and
tractile forces that are generated during gait. It should damping forces to the axis of rotation (ankle), g the
also capture the energy conservation that is accom- acceleration due to gravity, and mL2 hd is the net inertial
plished by exchanges between potential and kinetic en- moment of the system, mLg sin hd the restoring torque
ergy due to gravitational, elastic and inertial forces, and due to gravity, kb sin hu cos hu the restoring torque due to
any losses of energy due to less than perfect energy ex- the elastic properties of the tissues, Fd
hu ; h_u L cos hu the
changes during locomotion. Our previous work has escapement (state-dependent forcing), and ch_d b cos hu is
shown promising results on modeling locomotion. the moment due to damping (see Fig. 1).
However, the study was limited only to the analysis of Due to the presence of a dissipative term represented
the stiness of children with CP during locomotion at a by the moment due to damping, a forcing moment
constant speed under varying stride frequencies [13]. The acting on the center of mass is necessary to maintain the
purpose of the present study was to assess the capability oscillation. This state-dependent forcing behavior re-
of a forced, inverted pendulum with springs and dam- sembles that of the escapement mechanism of a pendu-
ping model to estimate expected dierences in the lum clock, where a periodic input of energy is applied to
complete dynamics of locomotion (force, stiness, the the system at an appropriate phase of the oscillation
gravitational eect and their relative contributions) at [19]. Thus, in such an autonomous system, the instant of
dierent speeds in children with spastic hemiplegic CP energy input is uniquely determined by the motion of
and to relate them to observed kinematic patterns. As the system. However, we need to estimate kb2 in order to
emphasized by Bernstein [14], the analysis of the dy- estimate, in turn, the escapement's forcing moment that
namic reactive forces arising from the movement is es- is applied during the phase from onset of push-o to the
sential to the understanding of co-ordination. end of toe-o.
Bernstein's perspective suggests that movements are not In normal walking, most of the power generated
completely determined by eector processes, but also by during each cycle is produced during the push-o phase
the internal and external forces acting on the individual. of walking [11]. In this escapement system, as in normal
In this spirit, the present study examined the possibility gait, the energy is assumed to be applied by the impul-
S.T. Fonseca et al. / Clinical Biomechanics 16 (2001) 793±805 795
kb2 . A viscous element (damping) is included to rep- The advantage of the inverted hybrid pendulum
resent the energy lost in each cycle. Finally, a forcing model is that it includes the major dynamic elements for
moment represents the muscular actions that periodi- locomotion. This allows the model to be used for
cally generate an accelerating moment on the center of walking, running or to be applied to describe the types
mass during push-o. of gait seen in children with CP. Because the equation
In normal walking, two thirds of the total power can be satis®ed in a number of ways, dierent gaits may
generated during each stride is provided by the ankle be produced by varying the elements' contribution (e.g.
extensors at the push-o phase of walking [11]. There- increased stiness).
fore, most of the periodically applied forcing moment is
generated at this phase in the cycle. The resulting net
inertial moment
mL2 h summed during this propulsive 4. Model predictions
phase re¯ects the contribution of all elements in Eq. (3)
including the forcing moment and restorative torques. The dynamics of walking in children with CP have
Because kb2 can be calculated from Eq. (7) and mLg can been described as having running-like characteristics.
be estimated from anthropometric measures, the kine- Increased re¯ex activity, co-contraction, and changes
matic measures of angular acceleration of the center of in the mechanical properties of connective and mus-
mass can be used to calculate the net inertial moment. cular tissues would be expected to result in greater
This procedure allows the estimation of the forcing stiness and help provide the exchange between elastic
moment according to the following equation: and kinetic energies that characterizes this motion.
Because the elastic restoring torque
kb2 of the model
X
n X
n X
n X
n X
n
Fd L mL2 hu cb2 h_u kb2 hu mLghu ; is a global measure of the system's stiness, this
i1 i1 i1 i1 i1 variable should capture the increased stiness of the
8 aected limb during the stance phase. Obusek et al.
[21] demonstrated a linear scaling between the gravi-
where n is the total number of sample periods between tational and the elastic restoring torque during leg
the beginning of plantar ¯exion (start of push-o phase) swing. As weight was added to the legs of adult sub-
and toe-o (end of push-o phase). jects, the measured leg stiness increased proportion-
If it is assumed that the damping coecient (cb2 ) ally. To avoid the eect of anthropometric dierences
remains constant for a given subject, changes in the among children, a stiness to gravitational component
damping moment over time will be dependent only on ratio
kb2 =mLg, instead of kb2 should provide a
the changes in angular velocity of the system. Dividing measure that can better dierentiate relative stiness of
Eq. (8) by the summed angular velocity during the same children of dierent masses and lengths. Increased
time period will produce a constant to be added to the stiness to gravitational component ratios are expected
forcing moment, according to the following: in the aected side of children with hemiplegic CP
Pn Pn Pn Pn compared to non-aected side and typically developing
2 2
i1 Fd L 2 i1 mL hu i1 kb hu i1 mLghu children (Prediction 1).
Pn _ cb Pn _ :
hu
i1 hu i1 Characteristics of children with CP are poor power
9 production of the gastrocnemius-soleus, and increased
resistance to passive motion of the limbs. The ratio be-
In any analysis aimed to assess the same individual for tween forcing and stiness, is predicted to be lower in
dierent conditions, we further assume the damping the aected leg compared to the non-aected leg of
term
cb2 to remain constant across conditions. Al- functionally limited children and compared to typically
though the values obtained for forcing moment are not developing individuals (Prediction 2).
actual values, they represent a true proportional esti- Walking speed is a multiple of frequency and stride
mate of forcing for those conditions. Thus, a propor- length. Increases in speed are due to linear increases in
tional estimate of normalized forcing can be obtained stride length and stride frequency [18]. Stiness is asso-
for each individual as follows: ciated with stride frequency and stride length or ampli-
Pn Pn Pn Pn tude is associated with forcing magnitude. Since children
2 2
Fd L i1 mL hu i1 kb hu i1 mLghu with CP have a limited capability of forcing, increases in
Pi1 / P :
n
h_u
i1
n
h_u i1
speed would be achieved largely through increases in
10 frequency (stiness). Therefore, the rate of change (re-
gression slope) of stiness in relation to walking speed is
While Eq. (9) allows an estimation of the overall gen- predicted to be higher on the aected side of these
erative and dissipative forces, Eq. (10) can be used to children than on the non-aected side or typically de-
assess the ability of a given subject to generate forcing veloping children (Prediction 3). In addition, the limited
moment as the speed of walking increases. forcing capability in children with CP should manifest in
S.T. Fonseca et al. / Clinical Biomechanics 16 (2001) 793±805 797
the rate of change of forcing moment in relation to height. The children's characteristics are presented in
speed of walking. In this case, the slope of a linear re- Table 1.
gression between forcing and speed in the aected leg of The participants were recruited from schools, clinics
functionally limited children is predicted to be closer to and hospitals in the Boston area. All children were paid
a ¯at line as compared to typically developing individ- for their participation in this study. Prior to participa-
uals and to the healthy leg of children with CP (Pre- tion in this study, children, parents or guardians signed
diction 4). a consent form in accordance with the policies of Boston
University Institutional Review Board.
5. Methods
5.2. Instrumentation
5.1. Subjects
Three-dimensional (3D) kinematic data were col-
Nine children with spastic hemiplegic CP and seven lected with an Optotrak 3020 System (Northern Digital,
typically developing children participated in this study. Waterloo, Ont., Canada) with two banks of position
Two children with CP were excluded from the study for sensors. Each sensor consisted of a three camera bank
inability to complete all speed conditions. In addition, and was connected in series with the other sensor and to
one child with CP and one typically developing child an IBM compatible computer through the Optotrak
were excluded due to technical problems during data System Control Unit.
collection. In the ®nal analysis, three male and three Commercially available force sensing resistors (In-
female children participated in each group. Children's terlink Electronics, Camarillo, CA, USA) linked in se-
age ranged from 8 to 15 years old. ries and ®tted within a custom-made shoe insole served
Selection criteria for typically developing children as foot switches. The switches were linked to an elec-
included no prior history of cardiovascular, neurological trical circuit [22] that produced 3.5 V when the switches
or orthopedic disease. Selection criteria for children with were fully loaded by the children's weight. The voltage
CP included a physician diagnosis of spastic hemiplegia output of the circuit was fed to the Optotrak Analog to
with no history of cardiovascular disease and no surgery Digital Unit (ODAU) and synchronized with kinematic
in the past 24 months. In addition, these participants data to provide event identi®cation. Data were recorded
were capable of independent walking without crutches, at the rate of 100 Hz.
walkers or braces. In the functionally limited group, Pairs of photoelectric cells were placed 2 m apart
three hemiplegic children had greater left side involve- before the beginning of the viewing volume. A digital
ment and three had greater right side involvement. timer connected to the photoelectric cells provided the
Groups were matched for age, gender, body weight and participants' walking speed.
Table 1
Group characteristics
Child ID Group Aected/corresponding Sex Age (yr) Height (m) Weight (kg)
to aected
lyzed, until the push-o was reached. This period cor- In all conditions the mean angular excursion of the
responds to the phase of concentric muscular activation center of mass about the ankle joint from the vertical
of the gastrocnemius muscle during push-o and de®nes during the push-o was small. The minimum angular
when the forcing moment is applied to the system. The excursion from the vertical was 11.2° at the very slow
program solved the forcing moment for each stride and speed condition of CPA and the maximum excursion
the mean value for each speed condition was used for was 18.5° at the very fast condition of TDCA.The
analysis. Considering all conditions analyzed, the max- maximum angular excursion of the center of mass from
imum number of frames included as the forcing phase the vertical around the ankle joint of 18.5° supported the
was 32 and the minimum was 8. use of the linear approximation of the equation of mo-
tion. FFT analysis performed on the angular displace-
5.5. Statistical analysis ment of the center of mass time series indicated that
most of the observed power was concentrated at the
Fast Fourier Transform (F F T ) analyses were con- walking frequency. Considering the total spectral power
ducted to determine the spectral power distribution of as 100%, the percentage of the power found at the stride
the angular displacement of the child's center of mass. frequency relative to the total power was calculated. It
This procedure veri®ed the condition of harmonicity of was found that the individual percentage of the total
the system. power at the stride frequency ranged from 89.9% to
Analyses of Variance (A N O V A ) with one between- 100%. Mean results for percentage of total power at the
subject eect (groups) and two within-subject eect (leg stride frequency varied from 93.6% at the very slow
and speed condition) were used to compare the following speed (CPNA) to 97.95% at the very fast speed
dependent variables: (1) speed, (2) elastic to gravitational (TDCNA). A second smaller peak was found for most
restoring torque ratio and (3) normalized forcing mo- of the children. This ®nding suggests that a source of
ment to stiness ratio between children with CP and energy is tapped every half cycle, consistent with an
typically developing individuals during the stance phase escapement due to push-o of each foot. The percentage
of walking. This statistical procedure allowed the testing of total power for this second peak varied from 2.05%
of the main eects group and leg for each condition, as (very fast) to 5.98% (very slow). A representative sample
well as interaction eects. Pre-planned focussed contrasts of the F F T plots is illustrated in Fig. 2.
were performed to locate signi®cant dierences accord-
ing to proposed hypothesis. The level of signi®cance was 6.1. Speed comparisons
set at P 6 0:1 for all comparisons. This signi®cance level
is appropriate for small sample studies by protecting There was a signi®cant dierence between groups for
against type I error and yet facilitating the process of speed (F
1; 50 6:59, P 0:0133) and for group by
®nding true group dierences [25]. Bonferroni correction condition interaction (F
4; 50 2:58, P 0:049). Fo-
for multiple comparisons in terms of speed conditions cussed contrast analysis demonstrated that typically
was applied to contrast results. The procedure changed developing children
CPA CPNA walked signi®-
the signi®cance level from 0.1 to 0.02. cantly faster than functionally limited ones
General linear models were used to test dierences in
TDCA TDCNA at fast and very fast speed condi-
slopes of normalized forcing moment, and elastic re- tions F
1 133:64, P < 0001 and F
1 715:50,
storing torque by speed of walking between children P < 0001, respectively). Fig. 3 depicts mean and stan-
with CP and typically developing children. dard error for all groups and conditions.
Linear regression analyses were used to test the as- Since stiness
kb2 has been demonstrated to vary
sociation between the child's speed (independent vari- according to the walking speed [34], all measures in-
able) and measures of normalized forcing moment and cluding stiness were normalized with respect to speed
stiness (dependent variables). for all subsequent analyses focussing on group com-
parisons.
For speed normalized ratio kb2 =mLg, k
6. Results
kb2 =speed
k :
11
mLg
The following abbreviations are used throughout the
results section: aected leg of functionally limited chil-
dren (CPA), non-aected leg of functionally limited 6.2. Stiness to gravitational component comparison, k
children (CPNA); leg corresponding to the aected (Prediction 1)
(matched) control leg of typically developing children
(TDCA), and leg corresponding to the non-aected Analysis showed that children with CP presented
(matched) leg of typically developing children signi®cantly higher k on their aected side than on the
(TDCNA). non-aected side in all speed conditions
P < 0:001.
800 S.T. Fonseca et al. / Clinical Biomechanics 16 (2001) 793±805
Fig. 2. Representative sample of FFT on the angular displacement time series for the aected leg of functionally limited children, CPA (a), non-
aected leg of functionally limited children, CPNA (b), leg corresponding to the aected (matched) leg of typically developing children, TDCA (c)
and leg corresponding to the non-aected (matched) leg of typically developing children, TDCNA (d).
Fig. 4. Mean and standard error of stiness
kb2 to gravitational Fig. 5. Mean and standard error of normalized forcing to stiness ratio
component
mLg ratio normalized by speed for all groups according normalized by speed for all groups according to speed condition.
to speed condition. Groups are: aected leg of functionally limited Groups are: aected leg of functionally limited children, CPA (j),
children, CPA (j), non-aected leg of functionally limited children, non-aected leg of functionally limited children, CPNA (d), leg cor-
CPNA (d), leg corresponding to the aected (matched) leg of typically responding to the aected (matched) leg of typically developing chil-
developing children, TDCA () and leg corresponding to the non-af- dren, TDCA () and leg corresponding to the non-aected (matched)
fected (matched) leg of typically developing children, TDCNA (
). leg of typically developing children, TDCNA (
).
P 0:018 and preferred
P 0:014 speed conditions. P 0:0003), and for the side corresponding to the non-
No dierence was observed at slow, fast and very fast aected side of the typically developing group
conditions
P > 0:089. Comparisons between CPNA (R2 0:981, slope 1421.50, P 0:0011). The slope
and typically developing children
TDCA TDCNA comparisons demonstrated a signi®cant dierence be-
demonstrated that functionally limited children pre- tween CPA and TDCA (F
2; 6 66:79, P 0:0001),
sented signi®cantly lower normalized Fd =kb2 than non- and between CPA and CPNA (F
2; 6 43:25,
functionally limited children
P < 0:001, except for the P 0:0003). No dierence was observed between CPNA
very slow
P 0:482 and very fast
P 0:046 speed and TDCNA (F
2; 6 0:18, P 0:839), and between
condition. Fig. 5 presents mean and standard error for TDCA and TDCNA (F
2; 6 0:407, P 0:683). In
all groups and conditions. summary, the results are in agreement with Prediction 3.
Slope comparisons on the mean stiness values for each
6.4. Stiness/speed relationship (Prediction 3) group and leg indicated that the aected side of the
functionally limited children had signi®cantly steeper
6.4.1. Stiness/speed regressions slopes than the non-aected side and typically develop-
Regression analyses for each individual between ing individuals. These results are summarized in Fig. 6.
stiness and speed demonstrated a very strong and sig-
ni®cant linear relationship for all children. Values for R2 6.5. Forcing/speed relationship (Prediction 4)
ranged from 0.946 to 0.999 for CPA, from 0.965 to 0.997
for TDCA, from 0.908 to 0.998 for CPNA, and from 6.5.1. Forcing/speed regressions
0.968 to 0.989 for TDCNA. In contrast to the stiness results, individual regres-
sion analysis between forcing and speed demonstrated
6.4.2. Stiness/speed slope comparisons non-uniform results. The R2 values ranged from 0.003 to
A general linear model was used to compare the 0.443 for CPA, from 0.430 to 0.976 for CPNA, from
slopes of the linear regression between stiness and 0.619 to 0.982 for TDCA, and from 0.007 to 0.858 for
speed. The regression lines of the group's mean value for TDCNA. For some children there was a marked drop in
each speed demonstrated a very high association be- forcing moment at very fast speeds. Since the very fast
tween speed and stiness for the aected side of the speed is located at the extreme of the regression line, this
functionally limited group (R2 0:994, slope 1595.25, condition has a strong leverage eect reducing the as-
P 0:0001), and for the non-aected side of the func- sociation between the variables. Removal of the very
tionally limited group (R2 0:984, slope 1503.18, fast speed produced much stronger and signi®cant re-
P 0:0009). Similar results were found for the side lationship between normalized forcing and speed for
corresponding to the aected side of the typically de- CPNA and TDCA groups. The R2 values ranged from
veloping group (R2 0:992, slope 1346.53, 0.853 to 0.976 for CPNA, and from 0.825 to 0.994 for
802 S.T. Fonseca et al. / Clinical Biomechanics 16 (2001) 793±805
Fig. 6. Scatter plot with ®tted regression lines between stiness and
Fig. 7. Scatter plot with ®tted regression lines between normalized
speed. Groups are: aected leg of functionally limited children, CPA
forcing and speed. Groups are: aected leg of functionally limited
(j, y 1592:0x 563:4, R2 0:995), non-aected leg of functionally
children, CPA (j, y 82:3x 127:5; R2 0:69), non-aected leg of
limited children, CPNA (d, y 1503:2x 266:3, R2 0:984), leg
functionally limited children, CPNA (d, y 582:9x 86:1, R2 0:76),
corresponding to the aected (matched) leg of typically developing
leg corresponding to the aected (matched) leg of typically developing
children, TDCA (d, y 1346:5x 334:5, R2 0:990), and leg corre-
children, TDCA (, y 422:11x 265:57, R2 0:95), and leg corre-
sponding to the non-aected (matched) leg of typically developing
sponding to the non-aected (matched) leg of typically developing
children, TDCNA (
, y 1421:5x 290:0, R2 0:980).
children, TDCNA (
, y 263:82x 0:87, R2 0:69).
TDCA. For TDCNA the coecients improved for two (F
2; 6 2:643, P 0:1502). In summary, the results
children and one child remained with R2 of 0.01. Ex- are in agreement with Prediction 4. Typically developing
cluding this child, the R2 values ranged from 0.74 to children and CPNA had signi®cantly steeper slopes than
0.89. The results for CPA were not aected by the re- that observed on the aected side of functionally limited
moval of the very fast condition except for one child. children. No dierence between sides was observed in
The R2 values ranged from 0.015 to 0.653. typically developing children. These results are pre-
In summary the relationship between forcing and sented in Fig. 7.
speed for individual regressions was strong and signi®-
cant only after the removal of the very fast speed for the
typically developing individuals and for the non-aected 7. Discussion
leg of children with spastic CP. The aected side of the
functionally limited children did not show a signi®cant The results supported the dynamic predictions in
association between forcing and speed. children with spastic hemiplegic CP derived from the
EDIPS model and from the changes in the kinematics
6.5.2. Forcing/speed slope comparisons and kinetics of the gait of CP reported previously in the
The regression lines of the group's mean value for literature. CP are relatively stier and generate less ac-
each speed demonstrated a strong association between tive force on the aected side (Prediction 1, Fig. 4). The
speed and forcing for the side corresponding to aected greater stiness is due to increased re¯ex gain and co-
and non-aected of the typically developing group contraction and to alterations of the connective and
(R2 0:947, slope 422.11, P 0:039 and R2 0:875, muscle tissues that are characteristic of CP [6±8]. It has
slope 263.82, P 0:019, respectively). The non-af- been suggested that uneven structural growth of bone
fected side of the functionally limited group demon- and muscles in CP may result in longer bones and ten-
strated only a moderate linear relationship (R2 0:765, dons and shorter muscle bellies leading to increased re-
slope 582.99; P 0:052). The aected side of the sistance to a passive stretch [8,9]. These anatomical/
functionally limited group also displayed a moderate structural changes in children with spastic CP poten-
association between forcing and speed (R2 0:698, tially provide them with an enhanced ability to store and
slope 82.303, P 0:078). The slope comparisons return elastic energy. The decreased force generation on
demonstrated signi®cant dierences between CPA and the aected side captured in the model estimates is
TDCA (F
2; 6 41:724, P 0:0003), between CPA consistent with reports of limited power production and
and CPNA (F
2; 6 46:040, P 0:0002), and between poor timing in the ankle plantar ¯exors [10,11].
CPNA and TDCNA (F
2; 6 17:915, P 0:003). No CP and typically developing children displayed simi-
dierence was observed between TDCA and TDCNA lar preferred, slow and very slow speeds. However, in
S.T. Fonseca et al. / Clinical Biomechanics 16 (2001) 793±805 803
Acknowledgements [17] Kugler PN, Turvey MT. Information, natural law, and the self-
assembly of rhythmical movement. Hillsdale: Lawrence Erlbaum
Associates; 1987.
This work was partially supported by the Brazilian [18] Obusek J. The force driven oscillator model in the control of
government agency CNPq through a four year schol- human walking speed and stride frequency. Doctoral Disserta-
arship and by the Dudley Sargent Research Fund Grant tion, Boston University, Boston, 1995.
conferred to the ®rst author. The work was also partially [19] Minorsky A. Introduction to non-linear mechanics. Ann Arbor:
supported by a grant from the Physical Therapy J.W. Edwards, 1947.
[20] Andronov AA, Vitt AA, Khaikin SE. Theory of oscillators. New
Foundation awarded to the second author. York: Dover; 1966.
[21] Obusek J, Holt KG, Rosenstein R. The hybrid mass-spring
pendulum model of leg swinging: stiness as a determinant of
cycle period. Biol Cybern 1995;73:139±47.
References [22] Hausdor JM, Ladin Z, Wei JY. Footswitch system for
measurements of the temporal parameters of gait. J Biomech
[1] Mochon S, McMahon TA. Ballistic walking: an improved model. 1995;28:347±51.
Math Biosci 1980;52:241±60. [23] Winter DA, Patla AE. Signal processing and linear systems for
[2] Blickhan R. The mass-spring model for running and hopping. the movement sciences. Waterloo Biomechanics: Waterloo, Can-
J Biomech 1989;22:1217±27. ada; 1997.
[3] Cavagna GA, Margaria R. Mechanics of walking. J Appl Physiol [24] Jensen RK. Body segment mass, radius and radius of gyration
1966;21:271±8. proportions of children. J Biomech 1986;19:359±68.
[4] McMahon TA, Cheng GC. The mechanics of running. How does [25] Ottenbacher KJ, Biocca Z, DeCremer G. Quantitative analysis of
stiness couple with speed. J Biomech 1990;23:65±78. the eectiveness of pediatric therapy: emphasis on the neurode-
[5] McMahon TA, Valiant G, Frederick EC. Groucho running. velopmental treatment approach. Phys Ther 1986;66:1095±101.
J Appl Physiol 1987;62:2326±37. [26] Paz AC, Burnett SM, Braga LW. Walking prognosis in cerebral
[6] Milner-Brown HS, Penn RD. Pathophysiological mechanisms in palsy: a 22-year retrospective analysis. Dev Med Child Neurol
cerebral palsy. J Neurol Neurosurg Psychiatry 1979;42:606±18. 1994;36:130±4.
[7] Myklebust BM. A review of myotatic re¯exes and the develop- [27] Zarrugh MY, Todd FN, Ralston HJ. Optimization of energy
ment of motor control and gait in infants and children: a special expenditure during level walking. Euro J Appl Physiol
communication. Phys Ther 1990;70:188±203. 1974;33:293±306.
[8] Hufschmidt A, Mauritz KA. Chronic transformation of muscle in [28] Jeng SF, Holt KG, Fetters L, Certo C. Self-optimization of
spasticity: a peripheral contribution to increased tone. J Neurol walking in non-disabled children and children with spastic
Neurosurg Psychiatry 1985;48:676±85. hemiplegic cerebral palsy. J Mot Behav 1996;28:15±27.
[9] Tardieu G, Huet de la Tour E, Bret MD, Tardieu C. Muscle [29] Sutherland DH, Davies JR. Common gait abnormalities of the
hypoextensibility in children with cerebral palsy I. Clinical and knee in cerebral palsy. Clin Orthop Rel Res 1993;288:139±47.
experimental observations. Arch Phys Med Rehabil 1982;63:97± [30] Knutson LM, Soderberg GL. EMG use and interpretation in
102. gait. In: Craik RL, Oatis CA, editors. Gait analysis theory and
[10] Giuliani CA. Dorsal rhizotomy for children with cerebral palsy: application. St. Louis: Mosby; 1995. p. 307±25.
support for concepts of motor control. Phys Ther 1991;71:248±59. [31] Lehmann JF, Price R, deLateur BJ, Hinderer S, Traynor C.
[11] Olney SJ, MacPhail HA, Hedden DM. Work and power in Spasticity: quantitative measurements as a basis for assessing
hemiplegic cerebral palsy. Phys Ther 1990;70:431±8. eectiveness of therapeutic intervention. Arch Phys Med Rehabil
[12] Gage JR. Gait analysis. An essential tool in the treatment of 1989;70:6±15.
cerebral palsy. Clin Orthop Rel Res 1993;288:126±34. [32] Lin JP, Brown JK, Brotherstone R. Assessment of spasticity in
[13] Holt KG, Fonseca ST, LaFiandra MT. The dynamics of gait in hemiplegic cerebral palsy I: proximal lower-limb re¯ex excitabil-
children with spastic hemiplegic cerebral palsy: theoretical and ity. Dev Med Child Neurol 1994;36:116±29.
clinical implications. Hum Mov Sci. (in press). [33] Carmick J. Clinical use of neuromuscular electrical stimulation
[14] Bernstein N. The co-ordination and regulation of movements. for children with cerebral palsy, part I: lower extremity. Phys
Oxford: Pergamon Press; 1967. Ther 1993;73:505±13.
[15] Morawski, JM, Wojcieszak, I. Miniwalker ± a resonant model of [34] Carmick J. Managing equinus in children with cerebral palsy:
human locomotion. In: Proceedings of the VI International electrical stimulation to strengthen the triceps surae muscle. Dev
Congress of Biomechanics, Copenhagen, 1978. Med Child Neurol 1995;37:965±75.
[16] Holt KG, Hamill J, Andres RA. The force-driven harmonic [35] Comeaux P, Patterson N, Rubin M, Meiner R. Eect of
oscillator as a model for human locomotion. Hum Mov Sci neuromuscular electrical stimulation during gait in children with
1990;9:55±68. cerebral palsy. Pediat Phys Ther 1998;9:103±9.