Teach-in '99 on Optical vs. Electromagnetic tracking: What people said...

Dear Dr. Chris Kirtley,

    First of all, I would like to thank you for yours considerations about
the e-mail "Teach-in #13".

    The study about knee articulation and movement, as well its mensurations
is very complicated as you have noted.

    We are studying this more than a year, and trying many ways to reach a
better method to determine the parameters defined by KAD (frontal thigh
plane, axis and center of the knee). As I have described in my last mensage,
we comeback to a more elemental point: to define the planes to decompose
general movement of the articulation (for example: frontal thigh plane).

    I believe, and if I am wrong please correct me, there are two basic
process to analyse the movement of an articulation:

    Although, this second method, which has clinics applications, propose
some basics definitions that I don't understand.
For example:
How to determine, with precision, what is frontal thigh plane? When a static
test, made with the KAD, give us a value of valgus/varus (abdution/addution)
with a large variation in cycle balance of gait, it may be minimized or
"corrected" resetting the KAD to determine a new frontal thigh plane. By
this way the projections of knee movements are remade, reducing the portion
in the frontal plane (abdution/addution). Is this analysis correct?

    We are sttuding a utilization of maximum amplitude concept of knee
rotation (movement of flexion/extention) - plane where a large angular
displacement occurs - as parameter to determine a reference plane (sagittal
plane). This study is in a very preliminary level, that is why I have
consulted the CGA forum. We are interested in establishing any variable of
the movement as reference to a  mathematics determination of each rotation
(flexion/extension, rotation and abdution/addution). However, even this
procedure will have limitations: if a person has a "rigid" knee, any
apprasial movement will be prejudiced.

    Other question, also related to the same concepts, is as the
transference of data obtained by gait analysis to a determined surgical
proceeding: derotational femoral osteotomy. Would it be necessary, to
realize this proceeding, the re-establishment of all the reference systems
used during the gait analysis by Vicon/VCM system?

    Thank you for your attention.

    Best Regards,

    Wagner de Godoy

Wagner, Chris and others,

If I've got time to pen one e-mail I must have time to pen two.

First with regard to the frontal plane of the thigh. Kadaba and
Davis both made a fairly pragmatic definition of the frontal
plane of the thigh in defining it as the plane which includes
the knee flexion axis (and the hip joint centre). That is it is
defined by the movement of the knee rather than the bony
anatomy of the femur. This is principally because with skin and
muscle movement, difficulty in placing thigh markers, KADs etc,
it simply is not possible to identify the position of the femur
to sufficient accuracy to base it on the latter.

This is fine with able-bodied subjects during gait as knee
movement is confined pretty well to a single plane. The logic
of this is to actually determine the frontal plane of the thigh
from the data from a walking test i.e. get away from this lark
of guessing where the knee flexion axis is for a static trial
with a KAD and find out where it really is from the gait data.

It has problems with pathological subjects who may have
ab-adduction instability of the knee. In this case the knee
axis (no longer knee flexion axis) may be orientated
differently to the femur and therefore as far as the femur is
concerned this approach will give a different value of internal
rotation to the able bodied comparison.

Two questions arise:

How much ab-adduction instability do our subjects have?

I think we've been influenced in this for years by seeing large
signals on the varus/valgus (sic) graphs into believing that
the knees are unstable. Its incredibly difficult to assess but
I suspect most CP kids have little medio-lateral knee
instability and that most of these signals are artefacts.
Obviously  there are patients with CP and others who do have
instability but I think a rule of thumb is probably that if
there is no reason from clinical examination to suspect
instability, always suspect when it turns up on the gait
traces. My work has been based on the principal that if you can
remove the signal by adjusting the position of the thigh marker
in the software then it is almost certainly artefactual. If you
are left with any residual signal then this may be real and you
need to consider this in your interpretation of the data.

Can we actually do any better?

Even if we do accept the possibility of some knee ab-adduction
is there a better way of defining the frontal plane of the
thigh? I suspect not. At least taking the plane including the
knee axis is fairly well defined. One could argue that after
intervention the knee axis might change in relation to the
femur but certainly with CP very little intervention is
directed at ab-adduction of the knee and any such effects will
be secondary (although there is a distinct possibility that
changing the knee kinetics which might give rise to different
patterns of knee movement).

Finally, and at last getting around to Wagner's question. How
do we define the knee axis from gait data? Two ways scream out
at the bio-engineer: the plane showing most knee flexion or
that showing least ab-adduction. On any model based on Kadaba's
or Davis' work (eg VCM) the placement of the thigh wand affects
both the orientation of the thigh axis and the position of the
knee joint centre. If the marker is too posterior the knee
joint centre will be too anterior and vice versa. Thus the knee
flexion trace is affected asymmetrically by the knee joint
centre position. If the only effect were that of thigh
orientation then picking the thigh plane to maximise knee
flexion would work, but because of this effect on knee joint
centre this approach will tend to result in thigh axis with an
overestimation of external rotation.

Movement of the knee joint centre has very little effect on the
knee ab-adduction trace (and what effect there is is
symmetrical about the mid-line). It is therefore clear that
minimising the knee ab-adduction trace will be a more effective
solution than maximising the knee flexion-extension trace.

And finally (again), the problem with this definition is that
it does't work with subjects with stiff knees (eiher in flexion
or extension). You can't simply minimise the ab-adduction trace
because whilst I believe that the knees are generally fairy
stable medio-laterally I also think there is considerable
variation between subjects in the fixed varus/valgus. (I always
minimise the waveform on the ab-adduction trace not its
absolute value). Our solution is to define a protocol for thigh
marker placement guided by our analysis of the gait data but
which is still appropriate for subjects with stiff knees. For
all my polemic above, we still regard playing with the gait
traces as a training tool and quality assurance check on
conventional methods of marker placement rather than as an ends
in itself.

Hope this helps,

Richard Baker <richard.baker@greenpark.n-i.nhs.uk>

PS To all those I've promised a copy of DynaKAD to. The latest
version will work optimally with a new version of BodyBuilder
to be available from Oxford Metrics sometime in May and I will
wait until this is available so I can check it thoroughly
before circulating you all.

Richard Baker
Gait Analysis Service Manager
Musgrave Park Hospital

Tel: +44 (0)1232 669501 ext 2155
Fax: +44 (0)1232 382008

Chris, and others,

Sorry to be so late in responding but I'm afraid work
often gets in the way of speedy replies.

Following your comparison of Polyhemus and VICON, you
might be interested in a paper which was presented as a
poster at the Dallas meeting last week comparing the
output of Cleveland Clinic and VCM marker sets (Castagno
et al, abstract to be published in next edition of Gait
and Posture). The study shows very nicely that the
differences between the two are too small to be of
clinical significance. What they also show however is
exactly the same ab-adduction artefact (thanks Raymond
for correcting my terminology). This should not be
surprising as both used exactly the same method to define
the knee joint axis (the same markers placed on the
medial and lateral epicondyles during a static trial).

What the paper illustrates is the problem lies not with
the marker set but with the definition of the knee joint

As regards your teach-in there are three possible
explanations for the residual artefact:

First is that you haven't actually got perfect alignment
of the axis systems and the artefact we are aware of is
still getting  through. I'm not sure it is just a
question of aligning the sensors in the software. What if
they weren't perfectly aligned on the patient (maybe I'm
missing something).

Second is the question of how much movement of the
sensor's axis system there is in relation to the bone.
With a conventional optoelectonic tracking system you are
only worried about the position of the marker and
although we accept this will be affected by skin movement
and movement of the underlying muscle bulk these are
generally acknowledged to be fairly small effects.
However it is not just the position of the EM sensor
which is important but also its orientation. I wouldn't
be surprised if the sensors are quite sensitive to
changes in the underlying muscle bulk in determining
segment orientation. Anyone done any work on this? (Steve
are you out there to defend your product?).

Third possibility is that you are actually getting true
signals (although the size of them  seems too large for
this to be the only explanation). Whilst the knee
movement is confined pretty well to the sagittal plane
during most of its range of motion odd things happen as
it gets towards full extension where you get the locking
mechanism and some rotation of the knee joint. This
happens in the last 10 degrees of extension and as such
isn't a feature of normal gait where he knee rarely
extends beyond 10 degrees of flexion but is obviously
included in the test you are doing. I'd suggest repeating
the test but restricting the range of movement to
between 15 and 60 degrees of knee flexion to try and
exclude this. (Although it may well exclude some of the
muscle bulk movement as the ilio-tibial band moves
forward quite markedly over the femoral condyles as the
knee moves towards full extension as well).


Richard Baker <richard.baker@greenpark.n-i.nhs.uk>
Gait Analysis Service Manager
Musgrave Park Hospital

Tel: +44 (0)1232 669501 ext 2155
Fax: +44 (0)1232 382008

Back to Teach-in