Professional Documents
Culture Documents
and
NURI AKKA~
Department of Civil Engineering. Middle East Technical University. Ankara. Turkey
Abstract-A mathematical dynamic model of the two-dimensional representation of the knee joint IS
presented. The profiles of the joint surfaces are determined from X-ray films and they are represented by
polynomials. The joint ligaments are modelled as nonlinear elastic springs of realistic stiffness properties.
Nonlinear equations of motion coupled with nonlinear constraint conditions are solved numerically. Time
derivatives are approximated by Newmark difference formulae and the resulting nonlinear algebraic
equations are solved employing the Newton-Raphson iteration scheme. Several dynamic loads are applied to
the center of mass of the tibia and the ensuing motion is investigated. Numerical results on ligament forces.
contact point locations between femur and tibia. and the orientation of tibia relative to femur are presented.
The results are shown to be consistent with the anatomy of the knee joint.
253
254 MANSSOURH. MOEINZADEH,AI.I ERKAN ENGIN and NURI AKKAS
of the knee joint have also been investigated via a external load conditions. These multi-segmented
nonlinear finite element model (Andriacchi et al., models play a significant role in both vehicle crash
1977). Both material and geometric nonlinearities are victim studies and aerospace-related applications. Our
included in the three-dimensional model of the joint present model includes both geometric and material
and its associated components. Numerical predictions nonlinearities. The highly nonlinear set of governing
are consistent with experimental observations. The equations are solved numerically and the numerical
models of Andriacchi et al. (1977) and Wismans et al. procedure is presented in sufficient detail. The effect of
(1980) are essentially the same in the sense that both the duration and the shape of the dynamically applied
deal with the quasi-static response of the knee joint. loads on the response of the joint is discussed.
As seen from the discussion presented above, mathe- Although the mathematical development to be pre-
matical modelling of the knee joint has not yet reached sented is general, the numerical results will be for the
the stage it deserves. It is interesting to note that a human knee. Detailed discussions of various anat-
biodynamic model of the knee joint is, to the authors’ omical and functional aspects of the human knee joint
knowledge, not yet available in the literature. It is more can be found in Gray (1973), Engin and Korde (1974)
appropriate to study via a dynamic model theresponse and Engin (1978).
of the joint to dynamically applied loads. The artificial
restrictions of the quasi-static inverse method, such as FORMULATION
the necessity to specify the preferred configuration, can
be eliminated if the dynamics of the problem is In the present work the knee joint will be modelled
incorporated into the model. The most general and by two rigid bodies connected by nonlinear elastic
realistic dynamic model of the knee joint should, of elements simulating the ligaments. It is assumed that
course, be three dimensional, however, it is believed the femur is rigidly fixed and the tibia is undergoing a
that a simpler two dimensional model should be general plane motion relative to the femur. Due to the
helpful and rewarding in an understanding of the two-dimensional nature of the model proper com-
problems that may be associated with the three- ponents of the ligament forces in the plane of motion
dimensional dynamic model. Moreover, by restricting are considered. The coefficient of friction between the
the motion to two dimensions it will hopefully be articulating surfaces. owing to the presence of the
possible to dwell upon the essentials of the problem synovial fluid, is known to be very low (Radin and
which will, in turn, lead to a sound development of the Paul, 1972). Accordingly, the friction force between the
more realistic three-dimensional model. femoral and tibia1 articulating surfaces will be
The purpose of the present work is to present the neglected.
development of a two-dimensional biodynamic model
of the knee joint. Such a biodynamic model can be a Geometry and kinematic conditions
part of a multi-segmented total-human-body model The position of the moving segment (tibia) relative
which is used in predicting biodynamic response of the to the fixed segment (femur) is described by two
human body subjected to expected or unexpected coordinate systems as shown in Fig. 1. The origin of the
Fig. 1. Coordinate systems locations and relative positlons of the tibia and femur are shown for the two-
dimensional dynamic model of the knee joint.
Two-dimensional dynamic modelling of human knee joint 255
F, = MEDIAL COLLATERAL
F,= LATERAL COLLATERAL
F,= ANTERlOR CRUCIATE
F,= POSTERIOR CRUCIATE
Fig. 2.I2orces acting on the moving tibia are shown for the two-dimensional model of the knee joint.
These forces are discussed in some detail in the flexion angle. Zero strain condition for the ligaments
following paragraphs. can be partially justified if an appropriate starting
Only four major ligaments of the knee joint will be flexion angle under no external load is chosen.
considered in the present work although consideration Let (&,), be the position vector in the base (?,j?) of
of any other ligament presents no difficulty. These the insertion point of the ligament m in the moving
ligaments are the lateral collateral (LC), the medial tibia. The position vector of the origin point of the
collateral (MC), the anterior cruciate (AC), and the same ligament m in the fixed femur is denoted by (F,,,)/
posterior cruciate (PC) (Fig. 2). The ligaments are in the base (El?. Here the subscripts t andfoutside the
modelled as nonlinear elastic springs. To be precise, the parentheses imply ‘tibia’ and ‘femur’, respectively. The
following force-elongation relationship is assumed for current length of the ligament is given by
each ligament:
&I = J[(~~)~-~~---(~~)ll[(V,)f-j?o-~(~~)tl.(13)
F, = K,(L, - 1,)’ for L, > I,, (12a)
The unit vector, I,,,, along the ligament m directed from
in which I(,,, is the spring constant, L, and 1, are, the tibia to the femur is
respectively, the current and initial lengths of the
ligament m. The tensile force in the mth ligament, thus, x, =+,-T.-T@.),]. (14)
designated by F,. It is assumed that the ligaments m
cannot carry any compressive force; accordingly; Thus the axial force in the ligament m, in its vectorial
F,=O for L,<l,. (12b) form, becomes
The stiffness values, K,, are estimated according to the 7, = F&t,,, (15)
data available in the literature (Trent et al., 1976;
Kennedy et al., 1976), and the values used in the present where F, is given by equations (12).
work are given in Table 1. These are the same values as The coordinates of the insertion points of the
those used in Wismans et al. (1980). Initial strains in the ligaments in both the tibia and the femur are de-
ligaments are taken as zero since, at present, there is no termined from the available information in literature
accurate data available on these strains as a function of (Wanget at., 1973;Crowninshieldetaf.. 1976)andclose
anatomical study of the knee joint. These coordinates
on the tibia are denoted by $,, and yk; and those on the
Table 1. Ligament stiffnesses
femur are denoted by x, and y,. The values used in the
Ligament K,(Nmme2) present work are summarized in Table 2. Obviously,
these values are for a specific specimen used in our
LC 15 study.
MC 15 Since the friction force between the tibia and the
AC 30
femur is neglected, the contact force will be in the
PC 35
direction of the normal to the surface at the point of
Two-dimensional dynamic modelling of human knee joint 251
contact. The contact force Ratting on the tibia is given NUMERICAL PROCEDURE
by
The governing equations of the initial value problem
M=yNi?, at x=x, (16) at hand are the three equations of motion (19), the
contact condition (10) and two geometric compat-
where N is the magnitude of the contact force and ibility conditions (7). The main unknowns of the
(d2f,/dx2) problem are x0, y,, a, A’, x, and XL. The problem is,
1’= ,&f,j&2, at x = xc (17) thus, reduced to the solution of a set of simultaneous
nonlinear differential and algebraic equations.
7 is either + 1 or - 1 and it ensures thecorrect direction The first step in arriving at a numerical solution of
of the contact force acting on the tibia. these equations is the replacement of the time deriva-
The final set of forces entering the equations of tives with a temporal operator; in the present work, the
motion are the external forces (Fig. 2) whose resultant Newmark operators (Bathe and Wilson, 1976) are
is given as: chosen for this purpose. For instance, f, is expressed in
the following form:
F, = (F,),:+ (F&j: (1%)
where [K] is an n x n coefficient matrix, {A) is a vector which is an exponentially decaying sinusoidal pulse of
of incremental quantities and {D} is a vector of known duration to, and amplitude A. A dynamic loading in the
values. Although the true value of n is 6, because of the form of equation (24) is extremely difficult to simulate
length and complexity of the equations new variables experimentally; however, the study of these two func-
were defined. thus, during actual computation process tions will hopefully be helpful in understanding the
n became 22. effect of rise time of the dynamic load on the joint
The iteration process at a fixed time station con- response. Equation (25) is a more realistic forcing
tinues until the delta quantities of all the variables function and it has been previously used as a typical
become negligibly small. In the present work, a representation of the dynamic load in head impact
solution is accepted and iteration process is terminated analysis (Engin and Akkas, 1978).
when the delta quantities become less than or equal to The following are obtained as a function of time
0.01 ‘>,, of the previous values of the corresponding from the computer program developed for this study;
variables. The converged solution of each variable is the coordinates of x0, y, of the center of mass of tibia;
then used as the initial value for the next time step and the flexion angle CC;the coordinates x, and XL of the
the process is repeated for consecutive time steps. contact point in (x, y) and (x’, y’) coordinate systems
The only problem that the Newton-Raphson pro- respectively; the magnitude, N, of the contact force;
cess may present in the solution of dynamic problems the elongations of the ligaments and the ligament
is due to the fact that the period of the forced motion of forces, F,.
the system may turn out to be quite short. In this case it The initial values for x0, y, and a are obtained by
becomes necessary to use very small time steps; specifying the location of the starting contact point.
otherwise, a significantly large number of iterations is Here, the following values are used for the coordinates
required for convergence. In our application of the of the contact point at t = 0:
method, only 5-6 iterations were necessary for conver-
gence most of the time. But there were also cases where x, = - 4.2 cm, xi = 2.5 cm
more iterations were required for convergence. This which yields
was especially true at instants at which there was a
sudden sharp change in the response of the tibia. Such x, = - 20.16 cm, y, = 17.49cm, and u = 234.79”
behavior manifested itself usually when the tibia This angle of rotation CC, corresponds to a flexion angle
started moving in the opposite direction due to the fact of 54.79” for which the ligaments of the knee joint are
that the pulling force of a ligament(s) became the in a relatively relaxed condition. It was reported as
governing force of the problem. Shorter time steps early as 1907 by Pringle that the position of maximum
required fewer iterations, as expected, even at the relaxation of the knee joint ligaments was approxi-
situation described, The time increment used in the mately halfway between full flexion and full extension
present work is At = 0.0001 s. of the knee joint.
The effect of pulse duration on the response of the
knee joint motion is studied by taking t, = 0.05, 0.10
NUMERICAL RESULTS AND DISCUSSION
and 0.15 set for both rectangular and exponentially
The numerical results to be presented are only for an decaying sinusoidal pulses. The effect of pulse ampli-
external force acting on the tibia without the presence tude, A, is also examined by taking A = 2ON, 6ON,
of an external moment. It is assumed that the force is lOON, 140N and 180N for both types of pulses. In
always perpendicular to the longitudinal axis of the principle, it is possible to plot and cross-plot the results
tibia (y’ -axis) and passes through its center of mass. Let in numerous ways. However, because of the limitations
this force be denoted by F,(t). A parametric study of on the space available, in the following a few typical
the effect of various combinations of moment and results will be presented. More extensive results can be
force acting simultaneously on the response of the found in Moeinzadeh (1981).
knee joint may prove to be rewarding. However, for Ligament forces as functions of flexion angle of the
the present work we will only consider an external knee joint for the two previously described forcing
force and believe that this will be sufficient to illustrate functions are presented in Figs. 3(a) and 3(b). Results
the capabilities of the model. The effect of the shape of indicate that when the knee joint is extended by a
the forcing function on the knee joint response will dynamic application of a pulse on the tibia, lateral
be studied by considering the following two functions collateral, medial collateral and anterior cruciate liga-
for F,(t): ments are elongated while the posterior cruciate
ligament is shortened. The load carried by the anterior
F,(t) = A[zf(t)-H(t-t,)] (24) cruciate ligament is substantially higher than those of
the lateral collateral and medial collateral ligaments.
which is a rectangular pulse of duration to, and The variation of the lengths of ligaments and the
Two-dimensional dynamic modelling of human knee joint 259
90
C- ANTERIOR CRUCIATE
P POSTERIOR CRUCIATE
C- LATERAL COLLATERAL
b MEDIAL COLLATERAL
C- ANTERIOR CRUCIATE
V-POSTERIOR CRUCIATE
I
I I O- LATERAL
P-MEDIAL
COLLATERAL
COLLATERAL
50 40 30 20 IO 50 40 30 20 0 -IO
tai FLEXION ANGLE, a=’ (b) FLEXION ANGLE,‘:”
Fig. 3. Ligament forces as functions of flexion angle, for an externally applied (a) rectangular and
(b) exponentially decaying sinusoidal pulses of 60 amplitude and 0.05 s duration.
forces carried by them during normal knee motion has rature. It is important to note that the dynamic model
been the subject of various studies reported in the presented in this study is an idealized representation of
literature and in these studies several different a very complex anatomical structure: thus, static
opinions and conclusions have been expressed in experimental studies may not support some of the
regard to the biomechanical role and function of predictions of the model. Additional disagreements
various ligaments of the knee joint. The function of the may also be due to approximate locations of the
anterior cruciate as depicted in the dynamic knee- attachment sites of the ligaments in particular. and
model developed in this study is to resist anterior two-dimensional nature of the model in general.
displacement of the tibia. This function is in general In Figs. 4(a) and 4(b), a few representative plots of
agreement with the experimental and clinical studies of forces in the anterior cruciate and lateral collateral
Kennedy and Fowler (1971); Girgis et al. (1975); Van ligaments are plotted as a function of time for two
Dijk et al. (1979); and quasi-static model analyses of different forcing functions with varying pulse dur-
Crowninshield et al. (1976); and Wismans et al. (1980). ations. Although not presented here, similar curves
The role of the posterior cruciate as predicted by the may be obtained for other pulse durations and pulse
model is to resist posterior displacement of the tibia. magnitudes. Generally, for a fixed amplitude, the
This function of the posterior cruciate ligament is in shorter the pulse duration, the sooner the tibia reaches
agreement with the experimental studies of Kennedy its turning point (i.e. direction of motion reverses) and
and Grainger (1967); Edwards et al. (1970): Girgis et al. for a given pulse duration, the smaller the amplitude,
(1975); Crowninshield et a/. (1976) and Wismans et al. the sooner the turning point is reached. In Figs. 3-5 the
(1980). values in parentheses indicate the flexion angles at the
The present dynamic model also predicts that the corresponding times. Note that, for illustrative pur-
medial and lateral collateral ligaments offer very little poses up to 6” of hyperextension was allowed.
resistance in the flexionextension motion of the knee Generally, one expects only 1-3” of hyperextension to
joint. The major role of these ligaments is to offer be anatomically tolerable beyond which joint failure
varus-valgus and partial internal-external rotational becomes unavoidable.
stability. The model shows that as the knee joint is In Fig. 5, contact forces as a function of time are
extended under influence of a dynamic load the lateral plotted. These forces are in response to the different
collateral and medial collateral ligaments elongate at forcing functions with varying amplitudes and pulse
different magnitudes. This prediction is in general durations. Note that the magnitudes of the ligament
agreement with the experimental results of Smillie and the corresponding contact forces in response to a
(1970); Edwards et al. (1970); and Wang et al. (1973). particular forcing function are comparable.
The model shows good agreement with the quasi- Femoral and tibia1 contact point locations as a
static experimental investigations reported in the lite- function of flexion angle are plotted in Fig. 6. In these
260 MANSSOUR H. MOEINZADEH, ALI ERKAN ENGIN and NUR~ AKKAS
4””
1. A two-dimensional mathematical dynamic model
LATERAL COLLATERAL of the knee joint has been developed to describe the
I relative dynamic motions between femur and tibia and
320
the forces in the joint. The model includes the curva-
ture effects of the articular surfaces as well as the
200 nonlinear behavior of the major ligaments of the joint.
C-69 2. A rectangular and an exponentially decaying
g 240 sinusoidal pulse ofduration, t,, , and amplitude, A, were
_ applied to the tibia and the numerical results from this
2 model were presented to illustrate the effects of
g 200
duration and shape of the dynamically applied loads
e
on the response of the joint. Special attention has been
5 160 given to the ligament and contact forces, the location
of contact points and anterior-posterior displace-
:
3 120
ments. The results were compared with the available
experimental data from the literature to establish the
validity of the model.
3. The model predicts that when the knee joint is
extended, by a dynamic application of a pulse on the
tibia, lateral collateral, medial collateral and anterior
cruciate ligaments are elongated while the posterior
cruciate ligament is shortened. The load carried by the
anterior cruciate ligament is substantially higher than
(‘4 TIME, (set)
those of the lateral collateral and medial collateral
Fig. 4. (a) Anterior cruciate and (b) lateral collateral ligament ligaments. However, because of the planar nature of
forces as a function of time for externally applied rectangular the model, the ligament functions predicted by this
pulses of 60N amplitude and durations of 0.05, 0.10 and model may not be quite reflective of the actual function
0.15 s.
of the knee joint ligaments and should be viewed only
as an approximation.
figures the values in the parentheses indicate the total With this model the influence of the variations of
elapsed time of the motion since its initiation. As the initial strain of the ligaments and their attachment sites
flexion angle decreases, it can be seen that the curves (i.e. insertions and origins) on the response of the
representing femoral contact points have a steadily model may be determined and a parametric study
increasing positive slope, while the curves of tibia1 addressing to these points may reveal the sensitivity of
contact points change slope from positive to negative the model to the variations of the coordinates of the
or vice-versa at various flexion positions. This phenom- insertions and origins of the ligaments.
Two-dimensional dynamic modelling of human knee joint 261
2700 - 7oc ,-
I (-6’)9
2400 -
60C )-
2100 -
3 5oc ,-
-1, =0.05s
f
- 1800 - c-t,=o.ios
Llo:O. I55 2
z
ki 400
g 1500 -
B
LL
z (25.5”)
L 1200 - ;3 300
F
Z
8 900 -
200
600 -
I’ I I
Od 0 0.10 0.20 0.30 0.40 0.50
0 0.04 0.00 0.12 0.16 0.20 0.24
W TIME, (set)
(a) TIME, (set)
2800 ,
DA= 20N
C-A= 60N
&A= IOON
PA. 140N
&A= l8ON
D-A= 20N
C-AC 60N
L-A= IOON
WA: 140N
C-A= 180N
0
0 0.10 0.20 0.30 0.40 0.50 0.60 0 0.10 0.20 0.30 0.40 0.50 0.60
(d TIME, (set) (4 TIME, (set 1
Fig. 5. Contact forces as a function of time for an externally applied (a) rectangular pulse of 60 N amplitude
and durations of 0.05, 0.10 and 0.15 s, (b) exponentially decaying sinusoidal pulse of 60 N amplitude and
durations of 0.05, 0.10 and 0.15 seconds, (c) rectangular pulses of 20 N, 60 N. 100 N, 140 N and 180 N
amplitude and 0.05 s duration and (d) exponentially decaying sinusoidal pulses of 20 N, 60 N. 100 N. 140 N
and 180 N amplitude and 0.05 s duration.
In a similar way the mathematical models of the complex behavior, with various portions behaving
other major articulating joints can be developed. differently under given conditions or configurations.
However, a special attention should be given in These are described in the literature as, in the case of
modelling of the ligaments. Some ligaments, particu- the broad ligaments of the hip joint, having anterior
larly the thick band or large cord-like ligaments, have and posterior fibers, or medial and lateral components.
262 MANSSOUR H. MOEINZADEH,ALI ERKAN ENGIN and NURI AKKAS
b-
(1=0.447s) 02X*
11.0.232s)
t, I-
2
3 1 1 I I I
LIJO '
FEMORAL CONTACT POINT, xc
z
“0-l -
Y.
(1=0.447sl
-4
&-“D50 40 30 20 IO 0
(b)
FLEXION ANGLE, a”
(4 FLEXION ANGLE, a0
Fig. 6. Femoral and tibia1 contact points as a function of flexion angle for externally applied (a) rectangular
and (b) exponentially decaying sinusoidal pulses of 60 N amplitude and 0.05 s duration.
Fig. 7(a)
Two-dimensional dynamic modelling of human knee joint 263
____
PLOT CIF XO,YO.XC,XPC,RND FLEXION
RNGLE RLFfl VS. TIME ::
:
d D
8
2
d :
YO
m IOON z
d :
$ 8
-6 IIlL 1:P
0.15s c:
: LL
:c
‘.g
2” CL
3 c
-.z $
26.
‘0.00 0.06 0. 2 0. I8 0.24 0.30 0.36 0.7;
L TIME [SEC1
3 z
a-
>
8
d.
Fig. 7. Continuous plots of tibia1 center of mass coordinates (X0, YO). femoral (XC) and tibia1 (XPC)
contact points and the flexion angle (ALFA) as functions of time for an externally applied (a) rectangular
pulse of 20 N amplitude and 0.15 s duration and (b) exponentially decaying sinusoidal pulse of 100 N
amplitude and 0.15 s duration.
The mathematical model should include additional Acknowledgemrnts- This paper is based upon a dissertation
elastic elements to reflect the contributions of various submitted by the first author to the graduate school of The
fibers of the ligaments. Unfortunately proper exper- Ohio State University, in partial fulfillment of the require-
ments for the degree of Doctor of Philosophy. The research
imental data to determine the constitutive behavior of
was supported by the Mathematics and Analysis Branch of
these thick band or broad ligaments do not exist. the Aerospace Medical Research Laboratory at Wright-
Finally, in advancing the understanding of ligament Patterson Air Force Base under Contract No. F 33615-78-C-
and articular surface functions of human joints under 0510 to the second author.
dynamic external loading conditions, the method of
joint modelling presented in this paper and the numeri- REFERENCES
cal procedure utilized in the solution of equations of
the two-dimensional dynamic knee joint model may be Andriacchi, T. P.. Mikosz, R. P., Hampton, S. J. and Galante.
extended to the solution of equations of three- J. 0. (1977) A statically indeterminate model of the human
dimensional dynamic model of an articulating joint. In knee joint. Biomechanics Symposium AMD, 23, 227-229.
Bathe, K. J. and Wilson. E. L. (1976) NumerrcuI Mrrhods in
fact, the three-dimensional articulating surface equa- Finite Elemrnr ~4nulyais. Prentice-Hall. Englewood Cliffs,
tions for the elbow, hip, knee and ankle joints as well as NJ.
the origins and insertions of the major ligaments of Crowninshield, R.. Pope, M. H. and Johnson, R. J. (1976) An
these joints have been already provided in Moeinzadeh analytical model of the knee. J. i3iomrchamc.s 9, 397405.
Edwards, R. G., Lafferty, J. F. and Lange, K. 0. (1970)
(1981). Development and results obtained from the Ligament strain in the human knee joint. 7’runs ASME,
analysis of the three-dimensional dynamic joint model M/inter Annual Meeting.
will be presented in a future paper. Engin. A. E. (1978) Mechanics of the Knee Joint: Guidelines
264 MANSSOUR H. MOEINZADEH,ALI ERKAN ENGIN and NURI AKKAS
for osteotomy in osteoarthritis. In Orthopaedic Mechanics: contact-area determination and implications. J. Bone. Jr
Procedures and Devices, (Edited by Chista, D. N. and Roaf, Surg. 54-A, 349-356.
R., p. 60. Academic Press, London. Menschik, A. (1974) Mechanik des Kniegelenkes, lc Teil.
Engin, A. E. and Akkas, N. (1978) Application ofa fluid-filled Z. Orthop. 112, 481495.
spherical sandwich shell as a biodynamic head injury model Moeinzadeh, M. H. (1981) Two and three dimensional
for primates. Aviation, Space. Env. Med. 49, 12&124. dynamic modeling of human joint structures with special
Engin, A. E. and Korde, M. S. (1974) Biomechanics of normal application to the knee joint. Dissertation, The Ohio State
and abnormal knee joint. J. Biomechanics 7, 325-334. University, Columbus, OH.
Girgis, F. G., Marshall, J. L. and Monajem, A. R. S. (1975)The Pringle, J. H. (1907) Avulsion of the spine of the tibia. Ann.
cruciate ligaments of the knee joint. Clin. Orthop. Ref. Res. Surg. 46, 169.
106, 216-231. Radin, E. L. and Paul, I. L. (1972) A consolidated concept of
Gray, H. (1973) Gray’s Anatomy. (Edited by Warwick, R. and joint lubrication. J. Bone Jt Surg. 54-A, 607616.
Williams, P. S.), 35th edition. Longman, London. Smillie, I. S. (1970) Injuries of rhe Knee Joint. Livingston,
Huson, A. (1974) Biomechanische Problme des Kniegelenks. London.
Orthopiide 3, 119-126. Trent, P. S., Walker, P. S. and Wolf, B. (1976) Ligament length
Huson, A. (1976) Quelques consequences Fonctionelles du patterns, strength and rotational axes of the knee joint.
croisement des ligaments articulaires. Bull. Ass. Anat., Paris, C/in. Orthop. 117, 263-270.
60, 71-80. Van Dijk, R., Huiskes, R. and Selvik, G. (1979) Roentgen
Kao, R. (1974) A comparison of Newton-Raphson methods stereophotogrammetric methods for the evaluation of
and incremental procedures for geometrically nonlinear the three-dimensional kinematic behaviour and cruciate
analysis. Comput. Str., 4, 109-1097. ligament length patterns of the human knee joint.
Kennedy, J. C. and Fowler, P. J. (1971) Medial and anterior J. Biomechanics 12, 727-731.
instability of the knee. J. Bone Jt. Surg. 53-A, 1257- Walker, P. S. and Hajek, J. V. (1972) The load-bearing area in
1270. the knee joint. J. Biomechanics 5, 581-589.
Kennedy, J. C. and Grainger, R. W. (1967) The posterior Wang, C. J., Walker, P. S. and Wolf, B. (1973) The effects of
cruciate ligament. J. Trauma 7, 367-376. flexion and rotation on the length patterns of the ligaments
Kennedy, J. C., Hawkins, R. J., Willis, R. B. and Danylchuk, of the knee. J. Biomechanics 6, 587-596.
K. D. (1976) Tension studies of human knee joint liga- Wismans, J., Veldpaus, F., Janssen, J., Huson, A. and Struben,
ments. J. Bone Jt. Surg. 58-A, 35&355. P. (1980) A three-dimensional mathematical model of the
Kettelkamp, D. B. and Jacobs, A. W. (1972) Tibiofemoral knee-joint. J. Biomechanics 13, 677-686.