View Full Version : EMD cont.

06-21-2005, 10:08 PM
Dear Readers

It seems most people agree there is something to the electromechanical
delay (EMD). The concept itself betrays its engineering origin where it
used to mean the delay of electromagnetic/mechanic switches (relays).
[Has anyone tracked the history of EMD in biomechanics?] Thinking of the
muscle as an actuator controlled via motor neurons easily provokes the
analogy with the relays. No one doubts there is sort of a turn-on-time
and a turn-off-time for the muscle. The questions are whether they can
be measured and defined in robust ways, and whether they are of any
physiological interest. For instance, i hit upon the paper

Corcos et al., Electromechanical delay: An experimental artifact (J
Electrmyogr Kinesiol, 2, 2, 1992, 59-68)

where is stated that "published values of electromechanical delay are
all so severely influenced by unknown factors of the apparatus on which
they are made that the published record is void of physiological
significance" (67). Their conclusion about the irrelevance of EMD may
apply if one uses their methodology for "measuring" it. Indeed, they
define EMD as the time from onset of DETECTABLE changes in EMG to the
onset of DETECTABLE changes in force, and what is "detectable" naturally
depends on the resolution etc; i.e., becomes device dependent. If one
adapts the threshold level (they call it T) close to the noise level,
instead of say some fraction of MVC or standard deviation of EMG and
force, the results will most likely vary with the resolution, as they
demonstrate experimentally.

Corcos et alii thus raise a valid point about the importance of how one
tries to extract parameters from data, and to what extent we obtain
intrinsic values, or values that mostly reflects properties of the
measuring device. Adopting the threshold idea for determining EMD the
authors obtain, for a simple dash-pot model, the approximative expression

t_f = sqrt {2*(T/R)* B* (1 + K_d/K_s)}

for the time t_f "to generate detectable changes in force". T is the
detection threshold; R is the rate of force development of the muscle
(one assumes that Force(t) of the contractile element grows with time t
as R*t; i.e., a linear ramp); B is the viscosity parameter of the muscle
(corresponding to the velocity term in the Hill-model); K_s is the
spring constant of the muscle SEC; K_d is the spring constant of the
measuring device (e.g. a hand grip). This expression presupposes an over
damped muscle (viscosity term dominates); indeed, the over-damped case
leads to an equation of the form dx/dt = a*t from which one obtains the
above expression setting x(0) = 0 and x(t_f) = T.

Incidentally, a recent paper

Isabelle et al., Electromechanical assessment of ankle stability (Eur J
Appl Physiol 88, 2003, 558-564)

demonstrates a method for measuring EMD of the peroneal muscles (PL)
using supramaximal electrical stimulation while the participant was
standing on a force plate. The EMD was defined as the time from onset of
PL EMG activity to the onset of the lateral ground reaction force (GRF).
For healthy people (no functional ankle instability, FAI) they obtained
10.5 ± 0.7 ms for bipedal stance, and 8.7 ± 0.6 ms for monopedal stance.
These numbers could makes sense, but we are not informed how exactly the
onsets are determined, leaving it open to the criticism of Corcos et
alii. The lower EMD value for monopedal stance is interpreted as a
consequence of "higher SEC stiffness" which would be in line with the
expression for t_f (which decreases with increasing K_d).´The effect of
mechanical coupling could have been tested making also measurement with
compliant foam between the feet and the force plate. Isabelle et alii
note the variations in the EMD values obtained by various groups but
they seem unaware of the points raised by Cocos et alii.

One would certainly expect EMG-to-force models to address the EMD-issue
as well. One example is

Lloyd & Besier, An EMG-driven musculoskeletal model to estimate muscle
forces and knee
joint moment in vivo (J Biom 36, 2003, 765-776).

They too go back to the "critically damped linear second-order
differential system" which they render as a discrete IIR-filter. This
filter involves a delay d which they set to 40 ms and refer to as EMD
and which they employ because it "improves the synchronization between
activation and the force production". Referring to the paper by Cocos et
alii they think d should be reduced to 10 ms in future models "by
modelling the delay of force production within the musculotendon unit".
So the quantity seems model-dependent, but this is not a problem but a
basic condition of the physical sciences in general. While the use of
black box models (such as based on neural networks etc) and system
identification methods may have important uses they turn a blind eye on
the physiological interpretations of the parameters.

Finally, the cross-correlation analysis of EMG-force has been up, and it
does proved a measure of phase shift between EMG and force, but its
relation to turn-on and turn-off times is probably quite convoluted,
necessitating a physical (tailored) model to make a headway on that (a
problem i am interested in).

Regards Frank Borg


This message has been scanned by F-Secure Anti-Virus for Microsoft
For more information, connect to http://www.F-Secure.com/