Isokinetic vs. Inverse Dynamics

[Devita, Paul]

Dear Friends,

I submitted a few questions about isokinetic and inverse dynamics
estimates of peak torques and I (and you through Biomch-L) received a number
of insightful responses. I thank everyone who responded to my original
posting and who responded to each other.

A summary of the responses is:

1. Many differences in isokinetic and inverse dynamic estimates of maximum
torque capabilities are due to methodological differences in the ways the
data were obtained.

2. There are basic physiological differences in neuromuscular function in
isokinetic tests compared to normal human movement.


These factors explain much of the differences in the values I reported in
my original post.

I list my original note and the responses in the order they were
submitted. I also responded to Vasilios Baltzopoulos' remarks below (just to
dig a deeper hole for myself).

This discussion is still open and other remarks are welcome.


Thanks again and have a happy holiday and new milleniyear,
Paul DeVita



Original note:

Dear Friends,

It seems that isokinetic and inverse dynamic estimates of joint torque
capabilities are in disagreement. Isokinetic measures of maximal torques
seem to underestimate predicted, inverse dynamic values by almost 50% (see
table below). This disagreement extends to biomechanical models of muscle
force predictions also. For example, Glitsch & Baumann (J Biomech, 1997)
predict maximum force from the quadriceps during running to be about 4500 N
with their model yet isokinetic measures are usually about 700 N for the
quadriceps. In addition, the isokinetic measures are usually derived during
slower concentric contractions than are the inverse dynamic predictions.

Why does this discrepancy exist? Are isokinetic techniques limited in some
fundamental way? Are inverse dynamic predictions flawed? Am I completely
mistaken?

Thank you for considering these questions. I will post any responses I
receive.
Paul DeVita


Peak Isokinetic Quadriceps Torques (Nm)

204 Roos et al Musc & Nerve 1999 (my estimate)
200 Lynch et al J Appl Phys 1999
175 Poulin et al Can J Sport Med 1992
165 Crenshaw et al Eur J Appl Phys 1995
163 Huston et al Amer J Sports MEd 1996 (my estimate)
158 Wong et al Arch Phys Med 1997
144 Neder et al J Orthop Sport Phys Ther 1999 (mean of males & females)
100 Wilson et al MSSE 1993
80 Iossifidou et al Int J Sports Med 1998

154 Mean
41 Sd


Peak Inverse Dynamic Knee Extensor Torques (Nm)

400 Kovacs et al MSSE 1997 - explosive drop jumping
350 Bobbert et al Eur J Appl Phys 1986 - drop-bounce jumping
300 Bobbert et al Eur J Appl Phys 1986 - drop-counter movt jumping
300 Bobbert et al J Biomech 1988 - vertical jump
264 Simpson et al JAB 1990 - medium fast running
230 DeVita et al MSSE 1992 - landing from 60 cm
210 DeVita et al MSSE 1992 - moderate running
175 DeVita et al JOB 1996 - moderate running

278 Mean
74 Sd

(These values are simply a sample of values in the literature. Perhaps they
do not represent the total literature?)

++++++++++++++++++++++++++++++

Dear Paul,

What you observed might be explained by dynamics principle. Move a static
(constant) force F over a distance u, the work W_static=F*u. Move a dynamic
force (force linearly increases from zero to F) over the same distance u,
the work W_dyn=0.5*F*u=0.5*W_static.

In your case, isokinetic measures is close to static situation. If the
muscle does same amount of work in the experiments, the extension of muscle
is the same, and the muscle can be simplified as a linear spring, dynamic
force (calculated from inverse dynamic analysis) should be twice the static
value.

Of course, these are questionable "if"s. But this might help to explain what
you observed. Hope this helps,

weixin

==
Weixin Shen, Ph.D., Senior Scientist
Science and Engineering Technology Division
JAYCOR, Inc
9775 Towne Center Drive
San Diego, Ca 92121
Tel: (858) 552-3508
Fax: (858) 552-9172
E-mail: wshen@jaycor.com

+++++++++++++++++++++++++++++++++

"Devita, Paul" wrote:

> It seems that isokinetic and inverse dynamic estimates of joint torque
capabilities are in disagreement. Isokinetic measures of maximal torques
seem to underestimate predicted, inverse dynamic values by almost 50% (see

This is a very interesting question and I am posting a reply to Biomch-L in
the hope that this topic will stimulate a public discussion.

My thoughts:

(1) Methodology

In dynamic activities that include a ground impact phase, such as drop
jumps, there is a methodological problem in determining joint moments. High
frequencies in the kinematic data are usually filtered out, but left intact
in the ground reaction force signal. So the different terms in the equation
of motion, used to compute joint moment, are not treated consistently.
Typically, this leads to high moment peaks during the impact phase which I
consider to be an artifact. Muscle torques should never show impact peaks
since muscle is soft tissue and force can only change gradually. A solution
would be to filter both the kinematics and the forces (van den Bogert AJ, de
Koning JJ,
On optimal filtering for inverse dynamics analysis. Proc. 9th CSB Congress,
Burnaby, B.C., pp. 214-215, 1996).

However, the discrepancy mentioned by Paul de Vita also exists in movements
without impact (e.g. Bobbert, 1988, vertical jump),so this is definitely not
the only source of the problem.

(2) Stretch-shortening cycle

Although this certainly contributes in many of the activities listed by Paul
de Vita, I don't think that eccentric force-velocity properties or
stretch-induced force enhancement provide a sufficient explanation. In a
later study, Bobbert showed that a 300 Nm knee moment can also be generated
during a jump that starts from a squatting position, a movement with *no*
muscle lengthening at any time (Bobbert et al., Med Sci Sports Exerc
11:1402-1412, 1996). I think that the squat jump results also show that
muscle shortening during the isokinetic test is not the cause of the smaller
torque. The muscles are also shortening, quite fast, during the squat jump.

(3) Inhibition

I propose that an inhibition must exist which limits the muscle activation
during a typical strength testing protocol. This could be a protective
mechanism, or simply a consequence of our central nervous system being
"wired" to use muscles efficiently. If extra activation does not result in a
change in movement, why even try? Movement is not influenced at all by
muscle activity during isokinetic or isometric tests. If the latter
hypothesis is true, one would expect to see larger maximal voluntary torques
being produced against inertial or viscous loads, when compared to isometric
or isokinetic tests. Is motor control naturally lazy?

There is some evidence that in addition to torque, the muscle activation can
also be larger during functional movements when compared to MVC. Jacobs and
van Ingen Schenau reported that the EMG amplitude of the Vastus Lateralis
was 200%(!) of MVC during sprint running (J Biomech 25:953-965, 1992).
Analysis of EMG amplitude is also somewhat sensitive to methodological
problems, but those results may still indicate a true mechanism.

Twitch interpolation, if I remember correctly, can be used to determine how
close the voluntary activation is to the maximal activation. I am not
familiar at all with this field of
research and not sure if those results are consistent with a 30-50%
inhibition. Perhaps someone can comment on this.

Also I am not familiar with the literature references on maximal isokinetic
knee torques. If some of those were obtained by electrical stimulation, the
inhibition theory would not apply.

Finally, I can't resist quoting the muscle physiologist Henk ter Keurs, who
once commented "it is good to be lazy and ambitious at the same time,
because it makes you efficient". Although he intended this to apply to
academic pursuits, it may just be true for muscles as well.

Ton van den Bogert
Department of Biomedical Engineering
Cleveland Clinic Foundation
9500 Euclid Avenue (ND-20)
Cleveland, OH 44195, USA
Phone/Fax: (216) 444-5566/9198


+++++++++++++++++++++++++++++++++

Paul,

The knee extension moments estimated through inverse dynamic produce have to
be viewed very carefully. The knee extension moments estimated through
inverse dynamic procedure are knee resultant extension moments that,
generally, should not be considered as the moments purely generated by knee
extensors. Both knee extensors and flexors can affect the knee resultant
extension moments. When evaluating the knee extension moments, we need to
consider the involvement of knee flexors that is different in different knee
positions in different movements. Also, we need to consider the type of
muscle contractions. With other conditions the same, an eccentric
contraction
generates a greater joint moment than an isometric contraction does (could
be up to 40% greater), and an isometric contraction generates a greater
joint moment than concentric contraction does. Further, type of movements
and subject characteristics may affect the magnitude of the knee extension
moment as well. Finally, we also need to consider the sampling frequencies
used in inverse dynamic procedure, especially if we are interested in peak
kinetic values. An inadequate low sampling frequency may result in an
under-estimated knee peak resultant extension moment. I did a small test
long time ago which showed that for an Olympic weight lifting, I could not
get the peak vertical ground reaction force estimated from film data
reasonably close to that directly measured from force plate until the
sampling frequency reached 200 frames/second. We know that knee resultant
moments are mainly affected by ground reaction forces.

We recently studied kinetics of jumping and found a large variation in peak
knee extension moment between subjects. The knee extension moments are
generally between 0.35 to 0.15 times body weight x standing height (about
300 to 130 Nm) with video data sampled at 180 frames/second. Again, these
are just knee resultant extension moment, not knee extensor moment.

Bing
Assistant Professor
Division of Physical Therapy
The University of North Carolina at Chapel Hill

(I then responded to Bing):

Bing,

If anything, the fact that inverse dynamics produces the net torque at a
joint suggests that inverse dynamic results should be less than isokinetic
results. Isokinetic quadriceps results typically are not hindered by
hamstring coactivation and represent the maximum quad-produced torque. If
this were the true maximum value for a person, an inverse dynamic extensor
result should be something less than this due to some hamstring induced
flexor torque.

Since inverse dynamics often predicts higher torques than isokinetics the
problems comparing these two methods may be even worse than they seem.

Paul

(Then Bing wrote back):

Paul,

Thanks for your responses to my responses to your original message.

The knee extension estimated through inverse dynamics procedure could be
higher than that estimated using isokinetics procedure. As I mentioned in
my previous responses, the muscle force in an eccentric contraction can be
up to 40% higher in an eccentric contraction than in isometric contraction.
The peak knee extension moments estimated through inverse dynamic procedures
in those movements might have occurred in eccetric contractions of the quad
with minimum involvement of hamstring. If that is the case, if makes
perfect sense that they are much greater than those estimated through
isokinetic procedures.

Bing

+++++++++++++++++++++++++++++++++++++++++++

Hi, Paul,

We just published two articles recently on issue of isokinetic
flexion/extension exercises of the knee. We found that the quad forces
during extension is between 4 to 5 times body weight, that's approximately
4000N. If you are interested, here are the two references:

G. Li, E. Y. S. Chao, K. R. Kaufman, H. E. Rubash "Prediction of
Antagonistic Muscle Forces Using Inverse Dynamic Optimization During
Flexion/Extension of the Knee" (June, 1999, Journal of Biomechanical
Engineering, 121, 312-322)

G. Li, K. Kawamura, P. Barrance, E. Y. S Chao, "Muscle Recruitment During
Knee Extension Exercise" (1998, Annals of Biomedical Engineering, 26,
725-733).

Best wishes!

Guoan Li, Ph.D.
Orthopaedic Biomechanics Lab
Harvard Medical School
Boston, MA
617-667-7819

+++++++++++++++++++++++++++++++++++++++++

Hello all,

I would like to add some additional thoughts to Dr. van den Bogert's and Dr.
Devita's comments on this interesting topic.

3) Along the lines of inhibition during isokinetic testing. Dr. Graham
Caldwell and I have compared the EMG and kinematics of sprinters running on
a treadmill at 30% grade and 4.5 m/s to running at the same stride frequency
as the incline (MSSE in press). We found EMG amplitudes over 400% MVC in
the mono-articular hip extensors, knee extensors, and plantar flexors as
well as in the bi-articular rectus femoris and gastrocnemius during the
stance phase of the incline condition and amplitudes of over 200% MVC (as
did Jacobs and van Ingen Schenau) during level running at ~7.6 m/s (same
stride frequency). I agree that there are some methodological problems with
EMG amplitude/MVC measures, but these amplitudes are so much higher that
there still may be something to "lazy" motor control.

4) Energy transfer mechanisms via bi-articular muscles. Jacobs et al.
(J.Biomech. 29(4): 513-523, 1996) and Prilutsky and Zatsiorsky (J. Biomech.
27(1): 25-34, 1994) have shown/suggested that energy generated at adjacent
joints can be transferred via bi-articular muscles to assist in torque
production at the joint of interest--such as the rectus femoris may do when
the hip and knee are extending concomitantly during dynamic leg extension
activities (vertical jumps, running, sprint push-offs). Adjacent joints are
typically fixed during iso-kinetic testing.

5) Differing muscle lengths of bi-articular muscles. This is especially
evident with the hamstrings during isokinetic testing of the knee. Altering
hip angles has a large effect on isokinetic knee flexor torque (Johnson et
al. MSSE 30(5s): abstract #267, p. s47, 1998) most likely due to the
hamstrings operating on different portions of the length/tension curve.
Standardization of adjacent joint angles during isokinetic testing has
sometimes been lacking.


I look forward to continued discussion on this interesting topic.

Regards,

Stephen C. Swanson, MS
Biomechanist
Institute for Sport Science and Medicine at
The Orthopedic Specialty Hospital
5848 South Fashion Blvd.
Salt Lake City, UT 84107

+++++++++++++++++++++++++++++++++++++++

Dear Paul,

Are the isokinetic values that you present only from concentric tests?
Further, are the values during the other activities that you present taken
only during the concentric
activation of the knee extensors?

Ecentric moment values are much higher than concentric ones. In our tests
eccentric moment values are approximately 350 Nm (Kellis Å., Baltzopoulos V.
(1996) Gravitational moment correction in isokinetic dynamometry using
anthropometric data, Medicine and Science in Sports and Exercise,
28:900-907). In some
cases, some trained athletes exerted very high moments of force, so that
they could stop the resistive force applied by the dynamometer! This should
be taken into consideration when comparing isokinetic moments of force with
moments exerted during movements that involve eccentric and concentric
actions of the particular muscle group.

As for the role of inhibition in isokinetics we have developed an EMG/moment
model and found that during maximum isokinetic testing the knee flexors
(antagonists) provide approximately 17% of the moment recorded by the
isokinetic dynamometer. This means that the isokinetic dynamometer provides
an underestimation of the true moment exerted by the knee extensors. Here is
the reference:

Kellis E., V. Baltzopoulos. (1996) The effects of antagonist moment on the
maximum isokinetic moment measurements of the knee extensors, European
Journal of Applied Physiology and Occupational Physiology, 76:253-259.

The reason for finding peak moments only at slow angular velocities, is that
as angular velocity increases the non-isokinetic phase increases and
therefore, the maximum moment recorded by the isokinetic dynamometer is not
correct, due to the effects of moments of inertia.

I hope the above help.

Eleftherios Kellis,Ph.D.
Department of Physical Education and Sport
University of Trikala
Greece
e-mail: leftkell@yahoo.com

++++++++++++++++++++++++++++++++++++++++++++


(from Vasilios Baltzopoulos with my comments in red):

Dear Colleagues

Although there are very interesting areas in the estimation of joint
torques using inverse dynamics or isokinetic dynamometry, I think that
the way the question was posed is wrong and leads to misunderstanding
and unnecessary confusion. Let me start with the first statement that
"It seems that isokinetic and inverse dynamic estimates of joint torque
capabilities are in disagreement". This is wrong because if the inverse
dynamics approach was applied on the actual isokinetic test movement by
measuring the force exerted on the limb by the dynamometer and using a
reasonably detailed model of the extremity used, then the joint torque
calculated will probably be very close to the joint torque measured by
the isokinetic dynamometer. Of course, there will be some differences
given the assumptions, simplifications and measurement errors involved
in both techniques. In this case both the inverse dynamics estimation
and the isokinetic measurement refer to the same single isolated joint
under the same conditions of joint velocity, joint position and subject
effort (which will affect muscle velocity, length and activation that
determine muscle and joint torque).

-The two results would be similar because the input values would be
identical. The force measured by the dynamometer would be used in both
methods and the kinematics would be identical. In any case, perhaps on one
level we can say the methods are equivalent. But in a practical sense they
are not. Isokinetics are often used to find maximum muscle strengths. If the
isokinetic torque values obtained during maximum efforts are much less than
inverse dynamic torques obtained at less than maximum effort, something is
amiss. Are isokinetic values only applicable to isokinetic tests?

A manuscript review comment recently requested that I relate gait torques in
young and elderly subjects to their maximum abilities as measured
isokinetically. The walking speed and overall effort was moderate and
comfortable for all subjects: 1.48 m/s (I know this speed seems fast for
elderly but it really was not). Elderly had peak ankle plantarflexor torque
~100 Nm and peak isokinetic values from the literature for elderly are often
around 80 Nm. Clearly our subjects were not using their plantarflexors
maximally while walking. This example shows the basic point of my original
note.

However, if the comparison refers to isokinetic studies that examined,
for example, different subject groups, at a specific fast concentric
velocity with adjacent joints in certain positions and the results are
compared with inverse dynamics estimation of a multi-joint movement
which is performed perhaps at different conditions of muscle length,
velocity and activation then it is only natural to expect differences.
If, however, the isokinetic test is performed on the same subjects and
in similar conditions of subject positioning, joint velocity, joint
position and activation compared to the action of the particular joint
during the free activity (jumping, landing, running etc.) then the
results should be similar.

-The results I reported were for young, healthy adults for both isokinetic
and inverse dynamic analyses.

-Peak knee torques during running and jumping occur at the transition from
eccentric to concentric phases and thus occur at low or nearly zero
contraction velocities. In general, peak isokinetic torques were obtained at
faster contraction velocities. Therefore, contraction velocity favors
inverse dynamic tests for peak torques.

-Knee angular position varied between the studies but in general the knee
was flexed similarly at the moment of peak torques during the inverse
dynamic and isokinetic conditions. Isokinetic testing is often performed
through ~90 degrees of knee motion, from 90 degrees of flexion to full
extension. Humans flex to about to 50 degrees of flexion in running and
about 80 degrees in jumping. The maximum torques are observed at similar
positions in isokinetic and jumping, ~80 degrees to 70 degrees of flexion
and in more extended positions in running. Therefore, muscle length seems to
be a non-factor or to favor isokinetic tests when compared with running.

-Perhaps the final part of Vasilios' paragraph is my point. Isokinetic tests
are performed at different positions, joint veloicites, joint positions, and
activations compared to human movements typically analyzed with inverse
dynamics. There are inherrent differences in isokinetic and inverse dynamic
protocols and these differences affect observed torque capabilities. I think
we give the nod here to Vasilios here and his point that with careful
consideration of all pertinent factors, apparent differences in these
torques can be accounted for. He states this idea explicitely below.

I also disagree with the selective values of peak knee joint torque (not
quadriceps torque as the dynamometer measures net joint
torque=agonist+antagonist+other torques). A good male athlete in slow
eccentric or slow concentric isokinetic tests should be able to produce
approximately 260-280 Nm of joint torque with the knee extensors
dominant. Assuming that this net joint torque includes an antagonistic
(negative) torque by the knee flexors then the actual quadriceps torque
is probably in the region of 300 Nm or more. With a moment arm of the
patellar tendon in males of approx. 0.04 m this means a tendon force of
7500 N and not only 700 N as suggested by Paul. Even values of 200 Nm
will generate 200/0.04=5000 N of tendon force. These are high load
values of 6-10 times body weight applied on the tendons during
isokinetic tests and are comparable to other dynamic activities.

-We have measured different populations of young adults ranging from
sedentary people to olympic sprinters. Co-contraction of the hamstrings is
always low (less than 10% of the quad EMG) in these people during quadriceps
tests. The additional quadriceps force to overcome the co-contraction will
be low and it will not account for a large amount of the observed
differences in isokinetic and inverse dynamic torques. It may account for
the 20 Nm difference identified above by Vasilios.

-Oh yes, I definitely erred in the estimation of maximal quadriceps forces.

There are the problems with each method as well. Inverse dynamics
estimation of joint torque is an ill-posed problem as mentioned by Ton
and others previously. There are also other issues such as the change in
joint geometry and mechanics under loading. For example we have shown
changes in tendon orientation and moment arm with contraction. It is
reasonable to assume that these changes will be specific to the loading
conditions and certainly different between isolated joint loading
compared to multi-joint activities. A rigid model of the musculoskeletal
system used typically in inverse dynamics applications will not be able
to account for these changes under different loading conditions.

There is also the impression that the torque measured by an isokinetic
dynamometer is fairly accurate because it is a direct measurement. This
is true only if the joint velocity is constant. However, if you want to
assess the joint torque at a high dynamometer velocity (e.g. 300 or 400
deg/s) then you must ensure that the subject can achieve that velocity
within the restricted range of motion during the isokinetic test and,
more importantly, that the joint velocity is constant at 300 deg/s when
the maximum joint torque is recorded by the dynamometer. This check is
almost never performed by researchers and completely ignored by the
majority of clinicians.

To summarise, I think that the comparison of joint torque values between
isokinetic dynamometry and other movements in general is invalid if the
two activities are not similar in terms of subject type and positioning
and joint position, velocity and action type. Measurement and/or model
simplifications and assumptions errors exist in both techniques and it
is not a case of which one is the right and which one is the wrong
method.

I hope that these comments are useful and help the discussion and
apologies for the length of the message.

Best wishes

Vasilios Baltzopoulos, PhD
Associate Professor
Manchester Metropolitan University

Currently at:
University of Thessaly
Trikala 42100
Greece

Tel: 0030 431 47068
Fax: 0030 431 47042
Email: baltzop@pe.uth.gr or V.Baltzopoulos@mmu.ac.uk

+++++++++++++++++++++++++++++++++++++

Dear all,

A number of interesting problems are addressed here. It may be good to
separate them carefully.

1) In my opinion the most interesting point is the question of maximum
activation. Paul Devita rightly points to this effect: maximum moments in
jumping or running are substantially higher than those on an isokinetic
dynamometer, and even more so, I might add, than the so-called Maximum
Voluntary Contraction. I am not so informed on quadriceps, but in triceps
surae (ankle plantarflexors) typical peak values are, from a paper of ours
in the press:

175 Nm isometric MVC
110 Nm in a TWO legged squat jump (one-legged seems to be higher)
155 Nm in landing after a 2-legged jump
225 Nm in sprinting

The point is even worse when it is considered that the muscle is not
isometric, but shortening at the time of peak moment. By reckoning the
effects of series-elasticity, we would estimate (rather roughly) the peak
'active state' of the triceps surae, and got:

175 Nm isometric MVC (the same of course)
270 Nm in a TWO legged squat jump
150 Nm in landing after a 2-legged jump (eccentric contraction, thus
not so much higher)
500 Nm in sprinting!

In the EMGs the effect is also clearly visible: peak rectified EMG in
running can be 200% of 'MVC' or more (like Jacobs and Van Ingen Schenau,
1992).

On this point I agree with Ton vd Bogerts remarks sub 3). In the big leg
muscles, there must be motor units (and probably the strongest ones) which
become only active in very brief actions at the highest levels of
activation. In small muscles e.g. in the hand, some 95% of maximum may be
reached. McComas in his book "Skeletal Muscle"
(Human kinetics, 1996) on p. 211 discusses the problem, but in my opinon he
underestimates the effect. Electrical stimulation is not the solution. In
the big leg muscles is
is very difficult, painful and even dangerous to try maximal stimulation. In
the subject referred to above (a very tough guy) we could reach 170 Nm, the
others did not let us go above 120 Nm. (In Scandinavia they might reach
higher levels. ) I heard a story from a well-known investigator who
stimulated the knee extensors of a colleage to such an extent, that the
patella luxated with a loud crack.

2) The inverse dynamics is quite prone to methodological errors, Bogert
point 1), but I do not suppose that it is very relevant here.
Agree with Vasilios Baltzopoulos.

3) Paul Devita gives a muscle force of 700 Nm. This is incorrect, see
remarks of Vasilios. You have probably used the moment arm of the ergometer
force transducer (some 20 cm) instead of that of the patellar tendon, 4 cm.

4) Due to the effects of the series-elastic component, isokinetic speed is
only equal to shortening speed of the muscle fibres at the peak of the
moment (when dM/dt = 0). I have investigated this problem, and the
differences are very substantial. So only a single point can be used to
construct a force-velocity curve!

There are some interesting effects here, which surely warrant further
discussion and investigation.
Yours,

At Hof
Department of Medical Physiology &
Laboratory of Human Movement Analysis AZG
University of Groningen
Bloemsingel 10
NL-9712 KZ GRONINGEN
THE NETHERLANDS
Tel: (31) 50 3632645
Fax: (31) 50 3632751
e-mail: a.l.hof@med.rug.nl

---------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
---------------------------------------------------------------