View Full Version : Re: Marker-set independent gait analysis

07-11-2007, 12:23 AM
Dear all,

I would like to add a few remarks to the important discussion that is taking
place on biomch-l.

The general aim of movement analysis is the relative bone pose estimation.
To fulfil this objective, two different pieces of information are necessary:
bone morphology and bone movement. In most gait analysis protocols, these
two "requirements" are accomplished simultaneously positioning skin markers
directly on the bony landmarks of interest and tracking their movement. This
approach has two main drawbacks: first the bone morphology description

on few points that are hardly identifiable in a repeatable manner (this
raises a repeatability problem in the identification of the joint axes) and
secondly, most of these anatomical landmarks are close to joints and thus
markers located on them undergo large displacements with respect to the
underlying bones (accuracy problem).

In my opinion, and according to the work proposed by Cappozzo and Colleagues
(1995, 2005), to tackle the problem of the bone movement reconstruction it
would be better to split the "big" problem in two "small" subproblems:

1) Bone morphology acquisition

2) Bone movement reconstruction

This approach allows to formulate the movement analysis problem in a clearer
way thus facilitating its solution.

With respect to point 1 (Bone morphology acquisition):

It has been demonstrated (Della Croce et al. 1999) that the anatomical
landmarks identification by palpation is a very tough mission and it is
associated with a very high variability. Probably, if the anatomical
landmarks identification is carried out always by the same operator, the
repeatability could be very high but, of course, it is not a practical
solution. For those anatomical landmarks that can be thought to coincide
with rotation centres, the functional approach may be used and, under
favourable circumstances, it may contribute to improve both accuracy and
repeatability of the identification of the anatomical landmark involved.
However, axes identification remains a problem. An approach which could
improve the overall quality of the analysis has been recently proposed by
Donati and colleagues (2007) and it is based on the use of bio imaging. A
digital bone model is fitted with as many points of the subject's bone that
can be made available through a digitization procedure.

With respect to point 2 (Bone movement reconstruction):

It is has been demonstrated by Benoit and colleagues (2005) in an in vivo
study using bone pins that "although skin-marker derived kinematics could
provide repeatable results this was not representative of the motion of the
underlying bones". In other words also artefacts may be repeatable. Other
scientists (Cappello and colleagues, 2005 ) have demonstrated that errors
strongly affect the angles of "those joints characterized by a small range
of motion, such as knee ab/adduction and internal/external rotation. This
may be critical in the exploitation of gait analysis data for clinical
decisions". To overcome the aforementioned limitation, it is probably
necessary to take a step back and set as a priority answering or arriving to
an agreement to these questions:

What is the best way to reconstruct the movement of the bone?

Should we link the adjacent segments with unreal 3 DOFs constraints or can
we analyse the movement of each segment as a 6 DOFs body?

How many markers should be used for each body segment?

Where should markers be placed to minimize artefacts?

What is the most appropriate analytical technique to process the movement
data (optimal pose estimators, global optimization, STA compensation

Finally, another very important issue is the definition of the anatomical
system of reference. In fact, unfortunately, the different gait protocols
(Saflo, Helen-Hayes, Cappozzo et al., 1995, etc,etc) use different
anatomical frames definition and, therefore, their description of joint
kinematics differ independently from the quality of the estimate. This
circumstance impedes a direct comparison among the results obtained using
different protocols.

Thanks to everybody

Andrea Cereatti

Andrea Cereatti, PhD

Department of Human Movement and Sport Sciences University Institute for
Movement Science

Piazza Lauro de Bosis 6, 00194 Rome Italy

tel (+39) 06.36733.506 fax (+39) 06.36733.517