No announcement yet.

Myoskeletal Inverse Dynamics (BIONET Topic 2)

This topic is closed.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Myoskeletal Inverse Dynamics (BIONET Topic 2)

    Dear List Members,

    First, many thanks for the overwhelming response to those who reacted to
    the problem statement of the above topic.

    Although any contribution is welcome, it is necessary to keep the
    discussion focused on the current topic, as has been pointed out
    recently by Dr. Leardini and Dr. van den Bogert. There is no need to
    deviate from the topic currently under discussion because the next
    topics planned for the discussion forum are (abbreviated) "HUMAN BODY
    wide range of problems occurring in HUMAN MUSCULOSKELETAL BIOMECHANICS
    which, as should be remembered, is the general theme of the present
    sequence of discussion sessions.

    This does, however, not mean that other topics outside this range could
    not be debated but this should be done at a later stage to avoid
    discussions drifting out of hand. Virtually all contributions submitted
    until now fit into one or more of the above mentioned topic categories
    except, perhaps, Dr. Duda's microscopical cell-cell interactions. But
    even molecular dynamics is part of the current discussion series because
    the elucidation and modeling of the power stroke generation within the
    muscle fiber myosin-subfragment-1 unit after association with actin is
    presently a "hot topic" in skeletal muscle modeling.

    I shall now briefly comment on the opinions posted until Monday, 14
    January. The views expressed are very diverse and range from rather
    traditional and conservative to enthusiastic, visionary, and open for
    new approaches.

    The emphasis in the discussions lay undoubtedly on CLINICAL GAIT
    ANALYSIS. The CONSERVATIVE VIEW seems to be that present gait analysis
    procedures and models are, on the whole, satisfactory and do not really
    need to be improved substantially. It was suggested that accelerations
    are small in pathological gait and therefore inverse dynamics solutions
    in clinical gait analysis are trustworthy so that clinicians need not
    worry unduly. Concern was expressed that the present adequate models are
    not appropriately incorporated into clinical practice, that measures of
    validity of presently applied methods are missing, that the estimation
    of joint centers (or rather axes) of rotation are the single greatest
    error source, and that both bioengineers and clinicians lack
    biomechanical understanding.

    The discussants expressing more PROGRESSIVE VIEWS recognise the
    deficiencies of currently used models and methods in clinical gait
    analysis. A need for improvements is seen in the calculation of joint
    kinetics, in the presently used anthropomorphic models, in the
    estimation of joint axes locations, in the computation of moment arms,
    and in the development of new techniques for determining
    subject-specific model parameter values. In contrast, the necessarity of
    detemining individually and adequately for each subject the inertial and
    anthropometric segment parameters was no issue for the conservative gait

    Most interestingly, some of the (apparently) younger discussants
    enthusiastically proposed visionary concepts for future biomechanical
    modeling which strongly agree with my own views on Second Generation
    Biomechanics. A basic concept, relating to the objectives of human
    motion analysis, will be briefly discussed below. Congratulations to
    these colleagues of the younger generation. My hope is that they will be
    the ones who implement the next phase in the development of

    In conclusion, I would like to express MY PERSONAL OPINION on the views
    just summarized. It should be clear that motion analysis (which I feel
    is a more appropriate term than "movement" analysis) is not restricted
    to the analysis of pathological gait and sports motions but, quite
    generally, is applicable to any type of human motion such as
    (predominantly passive) car crash victim behavior, work motions,
    reconstruction of accidents, observation of non-gait-rehabilitation
    processes, etc. On the other hand, there is no clear definition of the
    term "motion analysis". If taken to be opposite of motion synthesis
    (synonymous with motion simulation or forward dynamics) then motion
    analysis presupposes the use of the inverted dynamical system equations
    (inverse dynamics) and therefore can not consist of procedures that
    compare kinematic quantities (such as joint angle histories) only.

    and its implications it should be realized that this formulation
    provides us with a means of assessing the VALIDITY of a specific inverse
    dynamics model used for motion analysis. In fact, if the dynamic body
    model is formulated as a multi-body system with a fictitious hinge at
    the body center and having a fixed number of degrees of freedom, and
    external constraints are accounted for by additional algebraic
    equations, then this formulation results in a combined system of
    differential-algebraic equations (which is state of the art). In this
    case, the first three equations of the inverted system can be used to
    compute the three spatial components of the ground reaction forces as
    would be predicted by the inverse dynamics model for a specific motion
    pattern used as model input. Obviously, these model-predicted ground
    reaction force histories can be compared with the ones actually measured
    by the force plates. Under the (justified) assumption that the measured
    ground reaction forces resulting from the dynamics of the motion of the
    real biosystem (the patient or subject) are reasonably accurate, the
    discrepancy between model response and that of the real biosystem is a
    measure for the validity and quality of the model and the input data.
    Small discrepancies imply good models (and input data), large
    discrepancies mean the opposite. A typical example of such an inverse
    dynamics model validation, showing dramatic model deficiencies which
    usually remain undetected, can be found in Hatze, H. (January 2002): The
    fundamental problem of myoskeletal inverse dynamics and its
    implications, Journal of Biomechanics 35/1, pp. 109-115. There it can be
    seen that this problem certainly did not originate from a purist's brain
    hidden in an academic back room or an ivory tower far removed from
    reality, but from every-day practical work in the motion research

    It is also not too surprising that, as one discussant remarked,
    the simplest animals (including humans) are able to consistently
    reproduce their movement patterns ...". Apart from the fact that I (and
    probably many others) sadly lack the feeling of belonging to the class
    of simplest animals, do these creatures most likely not engage in
    extensive model building and inverse dynamics evaluations, and therefore
    miss out on the opportunity and the intellectual pleasure of coming
    across this problem.

    Finally, I would like to respond to the accusation made by some
    discussants that frequently in biomechanical research the formulation of
    SPECIFIC HYPOTHESES, or the establishing and utilisation of simplifying
    principles, is missing. This is, unfortunately, true to a large extent
    for some areas of biomechanical research conducted today and is as
    unacceptable as it is in any other scientific discipline. At the
    beginning of any investigation there should always be a clear concept of
    the processes involved and the aims to be pursued.

    As far as the OBJECTIVES OF MOTION ANALYSIS are concerned, MY BASIC
    HYPOTHESIS underlying and guiding all my research and publications on
    active motions over the past 35 years, was always the postulated
    determines the form of any type of active (but not passive) human
    motion. More specifically, I am convinced that there ALWAYS exists a
    PERFORMANCE CRITERION, however complex, which the biosystem (the
    subject) attempts to minimise or maximise in the execution of a given
    task. In some cases we know these criteria, in others we don't. It has,
    for instance, been shown that if a person is to walk a long distance
    without being under time pressure, it will adopt a combination of step
    length and step frequency that minimizes the metabolic energy expended.
    In other words, the nervous system generates optimal neural control
    patterns that control the muscles in such a way that a (in a specific
    sense) optimal motion results. This is then the "best" motion under the
    given circumstances and for a specific individual.

    Suppose that such an OPTIMAL TARGET MOTION is available for a given
    individual, for instance by using the optimal control solution of an
    adequate human neuromusculoskeletal system model. Then the objective of
    motion analysis is to compare the recorded present-status motion of the
    subject with the optimum, investigate the reasons for possible
    discrepancies and, ideally, implement measures (treatment) for
    improvement until a satisfactory status has been achieved.

    At present, this is wishful thinking. Neither do we, in general, know
    the performance criteria the nervous system uses (which is one of the
    topics to be discussed soon), nor do we have sufficiently complex and
    adequate simulation models of the human neuromusculoskeletal system
    (two other discussion topics, including skeletal muscle modeling), or
    proper methods for determining the necessary subject-specific parameter
    sets to individualize the model (also a discussion topic). Analogous
    remarks apply to inverse dynamics.

    If, however, such comprehensive models and methods could be developed,
    their utilisation would open up entirely new opportunities. An EXAMPLE
    OF CLINICAL APPLICATION could be the following. A spastic child exhibits
    the typical pathological gait pattern characteristic of this condition.
    The child's individual set of anthropometric, inertial, articular,
    myodynamic, and myocybernetic parameter values is determined
    experimentally and used to individualize the general simulation model.
    The performance criterion to be minimized is the (appropriately defined)
    "difference" between the currently observed spastic gait pattern and a
    "normal" gait pattern. This is a combined optimal control and parameter
    optimization problem in which certain anthropometric and possibly
    myodynamic parameters are changed as part of the optimization procedure,
    as are the neural inputs to the skeletal muscles. The result could be a
    modified (but for the target gait pattern optimal) parameter set
    requiring, for instance, for its practical implementation the surgical
    lenghtening of certain muscle groups (such as the triceps surae) as well
    as the surgical translocation of specific muscle origins and (or)
    insertions. In addition, the optimization procedure would yield the
    optimal, now "normal looking" gait pattern specific for this child.
    Clinical gait analysis would then be used to continually observe during
    the rehabilitation phase the post-surgical progress toward the new
    optimal and normal gait pattern.

    This scenario may sound like a somewhat futuristic vision. I am,
    however, convinced that such developments could be successfully
    completed by a combined effort of the biomechanics community. Fairly
    advanced submodels and parameter identification techniques exist already
    and are used routinely.

    Herbert Hatze

    ************************************************** ******
    Prof. Dr. Herbert Hatze
    Head, Department and Laboratory of Biomechanics, ISW,
    University of Vienna

    Auf der Schmelz 6 Tel: + 43 1 4277 48880
    A-1150 WIEN Fax: + 43 1 4277 48889
    AUSTRIA e-mail:
    ************************************************** ******

    To unsubscribe send SIGNOFF BIOMCH-L to
    For information and archives: