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

08-02-2007, 12:35 AM
Re: Marker-set independent gait analysis

We have found this discussion of methods to capture human movement very
interesting and relevant. The following observations and comments are
based on collecting and analyzing more than 10,000 individual tests of
human movement from activities ranging from walking to a subject performing
a cartwheel (using markerless methods) that have been conducted in my (TPA)
laboratories over the past 30 years.

1. General use of marker based systems:
We concur with the previous comments that the selection of the method
should be based on the application. A quote from Albert Einstein applies:
“Everything should be made as simple as possible, but not simpler”. This
starts by focusing your question on specific variables and selecting the
best method and most direct method. (e.g. if the questions is; Did the
patient’s gait improve - then simple time-distance measures can be very

The following was learned the hard way by trying fixtures strapped to limb
segments, markers placed on bony prominences; markers distributed in
different locations along the limb and by evaluating the sensitivity and
accuracy to the number of markers.
a. Fixtures strapped on the limb introduce at least two types of
(i) Markers placed on fixtures can lead to substantial errors since the
fixture can move as a rigid body with respect to the underlying bone and
this rigid body movement cannot be detected or corrected by solidification
(ii) Tightly strapping a fixture to the limb can cause subjects to change
their natural patterns of movement. Subjects with straps placed on the
thigh and shank had a significant change in the flexion-extension moment
compared to walking without straps. (Fisher DS et al., 2003 ASME Summer
Bioengineering Conference).

b. The number of markers and distribution of markers is important.
(i) In general, increasing the number of markers on the skin can reduce error.
(ii) For the thigh and shank, the distal and proximal ends of the limb
segment produced the largest errors.
(iii) Error can be reduced by spacing the distribution of markers to avoid
multiple markers placed on single muscle planes.

c. Care must be taken in the location and definition of the anatomical
coordinate system.
(i) Accuracy of determining bony landmarks to define the anatomical system
relative to the technical marker system will influence the accuracy.
(ii) Precise location of the anatomical system is required for comparison
among studies.

These observations led to the development of the point cluster technique
(PCT) and the interval deformation technique to address the problem.
(Andriacchi TP et al., J. Biomechanical Engineering, 1998; Alexander EJ et
al., J. Biomechanics, 2001)

2. The application of the Point Cluster Technique (PCT)

We have converged on the PCT for most kinematic studies since it has proved
to be practical and produced sufficient accuracy to address the type of
questions we were pursuing. The accuracy of the PCT was tested using 2
patients with Ilizarov devices (one femur and one tibia). In addition, we
have found reasonable comparisons to published studies using bone pins. We
conducted a simulation study that introduced both random and systematic
errors to evaluate the computation robustness of this approach. We also
evaluated the method in conjunction with a method to account for both the
segment deformation and rigid body movement. Details can be found in the

A few comments on the use of the PCT:

a. The eigenvalues calculated at each time step provide a valuable metric
when attempting to identify phases of the activity where large deviations
from rigid body motion occur. Thus a retrospective quality control
evaluation can be conducted. Note: If all the markers are placed in a small
region of the segment with uniform deformation, then the eigenvalues won’t
change and this metric is not useful.

b. The method determines orthogonal coordinate axes (eigenvectors) based on
a second order tensor that is relatively insensitive to slight changes in
the position of the individual points and tends to track the rigid body
portion of the segment. Note: Marker placement is again critical since if
the markers are arranged in with axial symmetry than the direction of the
eigenvectors can become indeterminate. However, if care is taken in the
distribution of the markers then the PCT can be very useful.

3. Interpretation and scope of bone pin studies.

These types of studies are very important and provide useful insights into
the nature of skin movement. However, there is the potential to over
generalize the results from these studies.

a. While bone pins are a reasonable gold standard, there can be error
associated with kinematics derived from markers on these pins. The pins are
usually around 2.5mm in diameter. Therefore they are flexible and can bend
both elastically and plastically due to tissue pressing against the pins
during movement.

b. Care must be taken when attempting to generalize beyond the specific
marker system evaluated. For example, if only 4 markers are used and the 4
markers are not broadly distributed on the segment, then it would not be
valid to use these results to evaluate more robust methods that use broad
clusters of markers.

As discussed above, the ability to address skin movement artifact is
related to the number and distribution of markers. Therefore,
interpretation of these results should be limited to the type of marker set
used for that study. Note: A set with 4 markers would not be used for our
PCT. Thus care must be taken when attempting to generalize these results
beyond the scope of the specific marker set used in this study.

4. Validation Methods

It is also useful to look at what is the best method to validate a movement
capture technique. Should we look at precision, accuracy or something else?
While it is important to quantify these measures and use them when
interpreting results, it is important to consider the purpose of the
measurement. For many questions that we address with motion capture methods
we are interested in patterns of locomotion during activities of daily
living, since these are the patterns that influence chronic changes such as
osteoarthritis or are the conditions we want to influence with treatment. I
suggest that we consider another measure when evaluating the quality of a
measurement method. That is the “fidelity” of the test to the natural
conditions of activities of daily living. Thus for certain applications the
fidelity of the test condition could be more important than accuracy or
precision. Probably the best validation is one that shows that the gait
measurement during a specific activity is associated with some physical
changes (e.g. cartilage thinning, bone density, etc.). Thus when one
evaluates the relevance and quality of a test method the interpretation
should take into consideration the accuracy, precision and fidelity of the
measurement system.

5. A Markerless Method for the Accurate Capture of Human Movement.

A markerless system would address many of the issues raised in this
discussion. There is no dependency on marker placement, the method produces
a large collection of points on a single segment, there is no risk of
inhibiting normal movement patterns and the acquisition can be completed in
a relatively short time since time consuming marker placement is not
required. Even though the full development towards accuracy and precision
will be tied to available camera and computational technology, our initial
development and application of a markerless system has been very
encouraging. Please see the following publications for additional details.

Mündermann L., Corazza S., Andriacchi T.,The Evolution of Methods for the
Capture of
Human Movement Leading to Markerless Motion Capture for Biomechanical
Applications, Journal of Neuroengineering and Rehabilitation, 3:6, 2006.

Corazza S., Mündermann L., Chaudhari A. M., Demattio T., Cobelli C.,
Andriacchi T.P., A Markerless Motion Capture System To Study
Musculoskeletal Biomechanics: Visual Hull And Simulated Annealing Approach,
Annals of Biomedical Engineering, 34:6, 1019-1029, 2006.

Corazza S., Mündermann L., Andriacchi T., A Framework For The Functional
Identification Of Joint Centers Using Markerless Motion Capture, Validation
For The Hip Joint, [ePub] Journal of Biomechanics, 2007.


Tom Andriacchi, PhD, Stanford University
Katherine Boyer, PhD, Stanford University
Ajit Chaudhari, PhD, Ohio State University
Stefano Corazza, PhD, Stanford University
Chris Dyrby, Eng, Stanford University
Seungbum Koo, PhD, Stanford University
Sean Scanlan, MS, Stanford University