EBME 309 – Modeling of the muscular sub-system

Gross Muscle Architecture

Greater range of motion

and shortening velocity
Straight Fibers
Tendon

LMT Muscle
LM Fibers

More force per volume

Pennated Fibers
Joint

Tendon

Bone
(a) (b)
(a) Typical arrangement of muscle attached to bones; and (b) Five common arrangements of tendon and
muscle fiber [From Kandel, Schwartz, Jessel, Siegelbaum and Hudspeth, 2013 Figure 34.11].

Generally, muscle is made up of fibrous tissue that attaches to bones at both ends;
mostly via another set of tissue-type called tendons. With some exceptions, skeletal muscle
fibers are roughly in parallel with the direction of force; but there are several architectures
in which the fibers are at an angle to the
Spinal Cord
direction of force output. The angular (part of CNS)

Motor Nucleus B
orientation of the fibers is defined as the
Motor Nucleus A

pennation angle. Pennation angles are usually Motor Neuron A1

Tendon

small (<20 deg). Muscles that are pennated are Aponeurosis

characterized by larger forces per volume and

Muscle A
shorter excursions (amount of contraction);
while those that are straight are characterized
by smaller forces but larger excursions. A closer Muscle B

look at muscle tissue under a microscope will Muscle

Fibers

show that it is made up of several varieties of

hierarchical sub-structures. The main unit in
this hierarchy is the muscle fiber. Fibers are
bundled together to be excited by a single
neuron; which together constitute what is called
Groups of muscle fibers stimulated by a
single motor neuron constitute a motor
unit. [From Kandel, Schwartz, Jessel,
Siegelbaum and Hudspeth, 2013 Figure
34.1].
a motor unit. Each fiber can also be striated into several sub-structures called myofibrils
which, under a microscope can be shown to consist of sarcomeres. Sarcomeres are the basic

Myofibril Muscle fiber

I band A band Myofibril

Z line Z line

Sarcomere

Z line Z line

Cross-bridge

Thick (myosin) filament Thin (actin) filament

(a) (b)
(a) Sub-structures of a typical muscle fiber [From Widmaier, Raff and Strang, 2003. Figure 9.4]. (b)
When muscle contracts, sarcomere A-bands remain unchanged while their I-bands and H-zones
are reduced.
building blocks of all muscle tissue. Contraction occurs when the sarcomere molecules pull
into each other – to cause decrease in length and produce force.

Contraction occurs when the

sarcomere molecules pull into
each other – to cause decrease
in length and produce force. The
length of a muscle and its
excursion are both functions of
Cross-bridges in the thick filaments bind to actin in the thin
the state of the sarcomeres in
filaments and undergo a conformational change that propels the
thin filaments toward the center of a sarcomere. (Only two of the series while the maximum force
approximately 200 cross-bridges in each thick filament are shown.)
[From Widmaier, Raff and Strang, 2003. Figure 9.9]. produced at full activation is a
function of the number of active
sarcomeres in parallel.

Fibers (and hence motor units) in muscles come in a variety of types – the most
common is the division into slow and fast. Generally, the slow fibers are smaller in size
while the fast fibers are larger in size. As we shall see below, this has repercussions on how
muscle is recruited for movements.
Basic Muscle Input-Output Properties and Models

OVERALL STRUCTURE OF HILL MODEL OF MUSCLE

The first conceptual model

of muscle in terms of
externally observed
characteristics of muscle
tissue is attributed to A.V.
Hill (1938). Muscle model is
made up of three main
elements – contractile
element, series elastic
element (a nonlinear
spring) and parallel elastic element (another nonlinear spring). Only contractile element
responds to stimulation – hence considered the active element while the others are passive
elements. Contractile element incorporates length and velocity dependence of muscle fiber

force, which are also related to cross-bridge dynamics. Overall muscle+tendon length (Lmt)

is different from contractile element length (Lm). Input signal (u) from the central nervous
system, or other artificial stimulation device, undergoes a transformation before it reaches
the muscle active element. This transformation is defined by the activation dynamics. Let
us look at a brief overview of each of the elements in the above diagram.

ACTIVATION DYNAMICS

Neural input includes both recruitment and rate

modulation, but is often modeled as a single scalar.
Activation related to calcium release and cross-bridge
dynamics. The maximum force that can be produced by a
muscle is determined by the number of potential cross-
bridges that can be made, i.e. the number of bonds that can
be made between the myosin heads and the actin binding sites. This is determined by the
overlap between the myosin and actin filaments, and is the mechanism underlying the
contractile element length-tension characteristic. The length- tension characteristic
describes the maximal force that is possible if the muscle is fully activated, but the nervous
system controls the fraction of potential bonds that can actually be made.
Activation Dynamics describes the process of converting potential cross-bridges to active
cross bridges.

The nervous system controls force by recruiting motor units and modulating their firing
rates. Cross-bridge binding is modulated by intracellular calcium ion concentration.
Calcium (Ca) is released as a result of
neural inputs to the muscle fibers.
Active State 50 ms
[ Ca++] Each nerve action potential causes a
muscle fiber action potential
(excitation). The muscle action
Pulse
time
potential in turn depolarizes the T-
40 mV

time tubules, which cause Ca to be released

Change in calcium ion concentration in the sarcoplasmic from the sarcoplasmic reticulum
reticulum for pulses of varying durations and same
magnitude (40mV). [From Melzer et al (1986) – Figure 1]. (excitation-contraction coupling). The
Ca controls the interactions between
myosin and actin and the generation of force. When excitation is removed, force falls and
the muscle relaxes as the sarcoplasmic reticulum re-sequesters the calcium.

Rate Modulation of single motor neurons

The nervous system changes a muscle’s total active force by controlling the firing rate of
individual motor neurons. The force of each motor unit varies with stimulus rate. At low
rates, individual
twitches are seen. u(t)
interpulse
As rate increases, pulse interval (IPI)
the twitches
overlap,
increasing the
average force, time

and decreasing T=1/f

the ripple. (a) (b)
(a) Time relations between a single skeletal muscle fiber action potential and the
Maximum rate resulting contraction and relaxation of the muscle fiber force output (called a
stimulation gives twitch response) [From Widmaier, Raff and Strang, 2003. Figure 9.9]. (b) Typical train
of action potentials. IPI is inter-pulse interval and f is stimulation frequency or
a saturated, fully
stimulus rate.
fused response, (b) Terms used in stimulus pulse delivery
designated as
“tetanus”. Average force is a nearly linear decreasing function of stimulus period (IPI in the
figure).
When plotted versus frequency, the average force has
a sigmoidal shape.

Recruitment of multiple motor neurons

Each motor neuron innervates a group of muscle

fibers (motor unit) with uniform properties and all
fire together with output as a function of firing rate.
As more force is demanded, more motor neurons are
recruited. The total muscle force is the sum of the
forces of all active motor units. Recruitment and
rate modulation are different aspects of the same
thing. Both amount to changing the firing rate of the
motor units. Recruitment occurs as an abrupt firing
rate transition
from zero to
some minimum
threshold rate,
e.g. 8 Hz. or vice versa. This is a discrete transition –
rates do not go below the threshold rate without
going to zero. Above the threshold rate, the rates can
vary continuously but also have a maximum in the
range from 20-50 Hz, depending on the type of
motor unit. This continuous variation above the
threshold rate is termed rate modulation or temporal
summation of force.

Slow and Fast motor

units

In a typical
movement, the general
procedure is to recruit
the small, slow, non-
fatigable fibers in weak
contractions, to large,
(A) The time to the peak twitch force, or contraction time, is briefer in the fast-
twitch unit (upper row) while the peak force for the 100 Hz tetanus is greater in
fast, fatigable ones in the fast twitch unit (lower pane); (B) The absolute force is greater for the fast-
strong contractions. twitch motor unit at all frequencies [From Kandel, Schwartz, Jessel, Siegelbaum
and Hudspeth, 2013].
The slow fiber types produce less force/unit stimulation frequency but they are more
fatigue resistant while the fast fibers are the opposite.

Mathematical Models of Activation Dynamics

We will only describe black-box models, but structural models have also been developed.
In the black-box models, u(t) is the neural input and a(t) is the output activation level.
 What is u(t)?
- It is a scalar describing the combined
effects of all motor efferents to all motor
units
Active State 50 ms
[ Ca++]

- Therefore, it is a combination of both

recruitment and rate modulation
time

 What is a(t)?
Pulse
40 mV

time

- It lumps the activation of all fibers into

a(t)
one scalar
- It lumps together contributions of u(t)

different types of muscle fibers (e.g. fast

fibers and slow fibers)
- It is sometimes referred to as “active
act
state” or the amount of Ca++ bound to t
t0
troponin. It is the input to contractile a(t)
u(t)
dynamics.
In this situation application of excitation u(t)

causes a rise with a time constant

act .
Removing the excitation causes a decay with a deact t
 deact
time constant . There are many ways to
write the dynamic equation for this behavior. One of these is the non-linear relationship
given by the equation
da 1  1


β  (1  β )u(t) a(t)  u(t) (1)
dt τ act τ act

τ act
Here β  (2)
τ deact

Note: 0  u(t)  1 is the neural signal (excitation) from the CNS or as externally applied
by a neural stimulator. If it is assumed that τ act  τ deact  τ , then (1) reduces to the linear
form:
 da 1
 u(t)  a(t) (3)
dt τ  

This represents one first order ode that enables us determine the output (activation) for
a given input (excitation). This is the first state equation we will use for muscle tissue
dynamics.

CONTRACTILE ELEMENT

Force varies with muscle length

Tensile
Force, F
No Stimulation
(No Stim)

Passive Force-Length
change length and while keeping
the length, stimulate maximally

FP

L
Muscle Length, L

time
Active Force-Length
Tensile
Force, F
(With Stim) F0
With Stimulation

FA

FP

Muscle Length, L
time L0

Tetanic force reaches a peak, generally within the physiological length range of the
muscle. The peak force is F0, and it occurs at L0. The subscript refers to optimal. Passive
force increase exponentially, but is generally only significant for L>L0.

Functional consequences of length- tension properties.

There is an optimal posture for maximal

force production. Reconstructive surgery
involving muscle length alterations can
influence motor performance.

Length-tension properties vary with activation

 Active forces scale with F0 and with

activation
 Length axis scales with L0
Maximum muscle force F0 is proportional to physiological cross sectional area

 PCSA = (physiologic cross-sectional area)

muscle volume/fiber length
 F0 = specific force*PCSA
Lm  Specific force = 20-100 N/cm2 for most
human muscle fibers. This is the most
common method used to determine Fo.
LM

PCSA
Short fibers Long fibers
Large PCSA small PCSA

Fulcrum
Lever o Stimulate muscle isometrically to
tetanic force (To).
o Activate electromagnet to release
catch mechanism. Force on muscle
drops from isometric to a value (T)
Weight
Muscle
determined by the weight.
o Muscle contracts under constant
force (isotonic) contraction.
o Measure slope (velocity of
contraction) at instant of fall of
tf force.
o Repeat at different values of the
weight, T.
o Plot the weight (T) vs the velocity of
Quick-release experiment setup. With catch mechanism contraction.
engaged, muscle contracts isometrically. At time tf, catch o Repeat for different tetanic
mechansim is released to allow total muscle length to decrease activations a.
and lifting the weight at the other end of the lever.

Force depends on velocity

The quick-release experiment shows that force output depends on velocity of

contraction. There is a maximal velocity of shortening Vmax. Force decreases during
shortening and increases during lengthening. If muscle shortens while contracting, this is
called concentric contraction; while if it lengthens while contracting, it is called eccentric
contraction. Force-velocity relationship scales with activation level. Velocity axis scales with
Vmax beyond which no force is produced irrespective of activation.

Functional consequences of force-velocity dependence

 There is a limit to movement

speed, regardless of strength;
higher strength only means
greater acceleration
 You can lower larger forces than
you can lift.
 Power = force x velocity;
maximal at about 0.3 Vmax
 Provides damping; dF/dV = B
 Eccentric contractions often lead
to muscle injuries such as tendon
tear.

Mathematical Model of the Contractile Element

The active contractile element force is modeled as a product relationship [Hill, 1938]:

Fm  aF0LT (Lm )FV(Vm ) (4)

where F0 is the maximal muscle force, a is the activation level, LT is the normalized length
dependence scaling factor, FV is the normalized velocity dependence scaling factor, Lm and
Vm are the normalized muscle fiber length and velocity respectively. Normalized muscle
length is calculated by dividing by the optimal fiber length (i.e., Lm  Lm / L0 ) and
normalized velocity is calculated by dividing by the maximal shortening velocity (i.e.,
Vm  Vm / Vmax ).

SERIES ELASTIC ELEMENT

Linear Elastic Tendon Model

The series elastic element of the Hill muscle model includes the elastic properties of
tendon. Tendon behaves like a slightly damped nonlinear elastic element, with force
increasing with stretch, as shown in the accompanying graph from Loren and Lieber (J.
Biomechanics, 1995). These are load- strain curves obtained experimentally from several
human wrist extensor muscle tendons. The load is normalized to the maximal value for the

individual muscle (F0), while the strain is the

ratio of the incremental tendon stretch (Lt-

Lst) to the slack length of the tendon Lst.

(Lt  Lst )
ε (5)
Lst

Generally, the load-strain curve shows

Series elastic fiber strain with straight
line approximations.
toe- behavior, a low load region of
significantly
lower stiffness, followed by increasing
stiffness, as shown by the graphs. For
modeling purposes, a straight-line model is
often used to approximate the tendon elastic
properties as shown in the next graph.

Ft  Kt (Lt  Lst ) (6)

If one assumes that the maximum strain is always the

same when a muscle is contracting at its maximal force value
Force, F

F0, then a normalized model of tendon elasticity can be

derived. This model was developed by Zajac (Crit Rev BME,
1989), assuming a maximum strain of 3.25% at F0,
regardless of the muscle. Let LP be the tendon length when
the tendon force is F0. The maximum strain εmax is reached at
this force.
(LP  Lst )
εmax  (7)
Lst
At F= F0 the linear tendon model gives

F0  Kt (LP  Lst )

F0 / Kt  (LP  Lst )

Dividing both sides by LST gives the strain.

F0 / KtLst  (LP  Lst ) / Lst  εmax (8)

This can be solved for KT,

Kt  F0 / Lst εmax (9)

Thus, if you know the optimal force, you can predict the tendon stiffness if you assume a
value for the maximal strain. Tendon stiffness is inversely proportional to tendon slack
length and the maximum strain but proportional to maximal force, i.e. long tendons of
small muscles will be the most compliant.

Question for reflection: Is Zajac's assumption about the strain always being the same at
maximal force correct? See Loren and Lieber's data in the figure above.

Derivation of simulation equations and block diagram for Hill-type muscle model

The processes describing muscle contraction can be lumped into several distinct subsystems.
Activation dynamics describes the conversion of neural input signals into an activation level,
which is sometimes thought of as the amount of calcium bound to troponin. The contractile
element is also referred to as contractile dynamics, which describes the mechanical
properties of the muscle due to the interaction of the thick and thin filaments by cross
bridge cycling. The contractile dynamics are not seen directly at the tendon because of the
series elastic element, part of which arises from tendon, and part of which arises from the
elasticity of the cross-bridges and perhaps the filaments themselves.

For a Hill model of the contractile element, the contractile element force is the product of
activation, force-velocity and length-tension terms.

Fm  aF0LT (Lm )FV(Vm ) (4)

The model we desire is one in which Lmt and Vmt are the inputs, since this is what is
generally known. However, the Hill model (4) is in terms of Lm and Vm , which are inside the
muscle and are therefore not suitable as generalized coordinates in simulations.
The effect of tendon compliance (the inverse of stiffness) is to make Lm (and Vm ) a
function of both Lmt (and Vmt ) and muscle force. The greater the compliance (the lower the
stiffness), the greater the differences between muscle and muscle- tendon lengths and
velocities. In fact, if muscle force is increasing rapidly due to increasing activation, at the
same time that muscle is being forcibly lengthened by an external load, Lm can be
decreasing at the same time that Lmt is increasing (eccentric contraction).
 To simulate muscle mechanical properties
using the Hill model, we must derive the
differential equations that take into account the
series elastic properties. Since the SE and CE
elements are in series (see figure at right), their
forces are the same

Fm  Ft (10)

And their lengths add:

Lmt  Lm  Lt (11)

Differentiating (6) gives the force rate change in tendon as:

dFt dL t
 Kt  Kt Vt (12)
dt dt

Here, Vt is the velocity of tendon contraction.

(a) Combining Fiber and Compliant Tendon Equations

Since from the geometry of straight muscle:

Lmt  Lm  Lt (13)

Muscle and tendon velocities obtained by differentiating equation (13) add to give the
total velocity of the muscle:

Vmt  Vm  Vt (14)

Substituting (14) in (12):

dFt
 Kt (Vmt  Vm ) (15)
dt

Solving for fV in the contractile element model (4) gives:

Fm
fV (Vm ) 
F0  a  fL (Lm )

Invert this to get Vm as

 Fm 
Vm  fV1   (16)
 F0  a  fL (Lm ) 
Substituting into (15) and noting that Fm  Ft gives:

dFm   Fm  
 Kt Vmt  fV1   (17)
dt   F0  a  fL (Lm )  

Equation (17) is the differential equation for force change as a function of muscle
length, velocity and activation.

Note:
The variables in equation (17) are often normalized using the relations:
Fm Lmt Lst Vmt Vmax
Fm  , Lmt  , Lst  , Vmt  and Vmax  . This reduces the
F0 L0 L0 L0 L0
equation to the normalized form given by:

  
dFm 1  V  V f 1 Fm
   (18)
dt (εmaxLst )  mt

max V

 L 
a  f L  F ε L  L
mt m max st st   


Summary
 Neuro-musculo-tendon (also called musculotendon) dynamics is characterized by
two first order differential equations and hence two state variables a and Fm . The
two equations are:

da 1
 u(t)  a(t)
dt τ  

  
dFm 1  V  V f 1 Fm
  
dt (εmaxLst )  mt

max V

 L mt 
a  f L  F ε L  L
m max st st   


 For each muscle. there are 6 parameters: τ , L 0 , F0 , Vmax , L st and εmax that must
be put in these equations to fully evaluate the right hand sides in an integrator
routine to solve these equations.
 There are also two variables that must be computed at every step of the
integration: Lmt and Vmt .
 Note that we also require the two functions fL and fV1 .
 Typical approximations used for the functions fL and fV1 in normalized form are
given below. These models are obtained by curve-fitting to real data.
 3.05L2  5.98L  1.96 if 0.42  Lm  1.54
fL (Lm )   m m
(19)
 0 otherwise

Normalized Length Tension, fL(Lm) 1.0

0.8

0.6

0.4
0.2

0.0
0.4 0.6 0.8 1.0 1.2 1.4 1.6
Normalized Length, L m

 0.995e13.88(fV  1.39)  0.997e3.91fV if fV  0

fV1  Vm   (20)
0.997 otherwise

1.0
Normalized velocity, Vm

0.5

Normalized
0.0 Force-Velocity
1.2
1.4
1.0
0.6
0.4

0.8
0.2

f v(Vm)

-0.5

-1.0

 Typical values used for the parameters are:

 Parameter Explanation

Maximal active force, achieved at L . Proportional to PCSA

o
F (physiological cross sectional area = volume/ L ). Typical values for
o
o

2
specific tension are 20-100 N/cm .

L Muscle fiber length producing maximal tetanic force (optimal fiber

o
length).

V Maximal muscle shortening velocity. Typically 3-10 L /s

max o

τ Activation dynamics time constant. Typically 0.03-0.1 s.

L Slack tendon length. Difference between muscle-plus-tendon path

st
length and muscle fiber length when relaxed.

(b) Stiff or Non-Compliant Tendon Equations

The force differential equation above is for the case when the tendon is compliant and
hence transmits force by behaving like a spring. On some occasions we simplify the model
by assuming that the tendon is non-compliant or stiff. In this case the tendon acts as a
stiff rod between the fiber and the bone – hence the fiber force goes directly to the bone.
Thus the model for fiber force output now reduces basically to equation (4):

Fm  F0  a  fL (Lm )  fV (Vm )

Here now since the tendon is stiff, from (13) and (14) Lm  Lmt  Lt and Vm  Vmt . Values
of Lmt and Vmt are calculated using the position vectors and velocities of the origin and
insertion points for each muscle during the kinetic analysis of the bones to which the
muscles attach. Thus the only differential equation to integrate in the case of a stiff tendon
is equation (3):

da 1
 u(t)  a(t)
dt τ  

Values of activation from integrating this are substituted into (4) to calculate Fm  Ft .
PARALLEL ELASTIC ELEMENT

Recall that the passive force varies with the Tensile

Passive Force-Length
length of the muscle in a non-linear fashion.
Force, F
(No Stim)

This behavior is similar for all muscle tissue.

There are two ways to model this behavior –
using spline fitting with a lookup table or using
a fitted equation. The fitted equation, which is
more commonly used, takes the form:
Muscle Length, L

8(Lm 1)
e
Fp  (21)
e4