The effects of constraining OpenSim inverse

kinematics to a bone pin marker defined range

1 University of Ottawa,

Canada2 University of Copenhagen,

Denmark

INTRODUCTION

AIM

To determine the effect of applying bone pin marker

defined ranges of knee motion in OpenSim IK solutions.

ISBS XXIV Congress

July 18th-22th 2016

Tsukuba, Japan

The generic models available in the OpenSim repository contain knee

joint ranges that are not physiologically realistic3. In the NoBP

condition, nonphysiological oscillations in translations were observed

even when knee flexion angle was constant indicating that they were

the result of STA and were removed once the BP constraints were

applied. These kinetic differences will also have an impact on

quantifying joint contact and ligament loading, which will likely facilitate

more accurate injury assessments through musculoskeletal modelling.

Therefore, caution should be expressed when using the results from

musculoskeletal modelling as STA and optimizations can introduce

error in both the kinematics and kinetic solutions. This error is

amplified during ballistic and high impact tasks such as jump landing.

DISCUSSION & CONCLUSION

1. Benoit et al. Gait Posture. 24: 152-64, 2006.

2. Potvin et al. J Biomech. Submitted.

3. Xu et al. Com Meth Biomech Biomed Eng. 18: 1217-1224, 2015.

METHODS

Thirty one healthy active young adults (15 males & 16

females; 25.7± 5.7 years) participated in this study.

Participants completed 3-5 successful jump lunges

where they were asked to stand on their non-test limb

and jump forward onto a force plate, land on their test

limb and maintain balance for two seconds (Fig 1).

All data were processed through a

Matlab - OpenSim API. The

Gait4392 model was adapted to

include 37 degrees of freedom

(dof), 46 lower limb muscles,

patellae, and six dof at the knee.

Data were scaled and processed

through IK and inverse dynamics

(ID) to achieve the no bone pin

constrained (NoBP) results. Bone

pin data from Benoit et al.1 were processed to create

boundary envelopes of each knee dof as a function of

knee flexion angle to be used as the full ranges of

motion for each of the knee model's 5 other dofs (Fig

2). The same scaled models were then subjected to IK

and ID once again (BP) where the calculations were

completed at each time point (every 0.01s) and the

knee dofs were adjusted so that their kinematics were

constrained to the bone pin range associated with the

current knee flexion angle at that instant in time.

Paired sample

T-tests through

statistical parametric

mapping were used

to test kinematic and

Kinetic differences

between the two

conditions at

α < 0.05.

RESULTS

Figure 3: Mean and standard deviation

clouds for all 6 knee dofs kinematics. Black

vertical line represents heel contact while

blue horizontal line represents period of

significant different between the two

conditions

REFERENCES

Figure 1: OpenSim

model of the jump lunge

When BP constraints were applied,

significant differences were observed

for all six knee dof (Fig 3). BP results

yield a significantly more flexed

knee; although these differences

were only a few degrees. BP knee IK

results were also more adducted and

externally rotated throughout the

movement. Significant differences

were also observed for anterior/

posterior and distraction/

compression (DC trans) translations

throughout the entire jump lunge

while medial/lateral translations were

only significant pre and 50 ms post

contact.

Large discrepancies were observed

in the DC trans (Fig. 4A) shortly after

contact. In NoBP, DC trans was

shown to oscillate in a range of 8 mm

to a max of 20 mm of compression

during this movement, which is

nonphysiological (Fig. 4B). This is

likely due to the inertia observed in

the skin causing soft tissue artifact

(STA) approximately 20-150 ms after

the large impact force at contact.

Since OpenSim uses motion capture data as input while

solving inverse kinematic (IK), it is subjected to soft

tissue artifact as the commonly used surface markers

do not correctly represent the underlying rigid body

bones1. These errors cause nonphysiological

movement of bodies in the OpenSim simulations.

Figure 2: Bone pin defined ranges for all 6 knee

dof as a function of knee flexion. Potvin et al.2

Significant differences in kinetic results were also observed for all six

knee dof from initial contact onwards. After contact, BP constraints

produced a significantly greater flexor moment albeit it clinically small

(max diff: 0.18 Nm/kg). The BP solutions also solved for greater knee

abductor (max diff: 0.23 Nm/kg) and external rotator joint moments

(max diff: 0.05 Nm/kg) after contact. With respect to translation forces,

the BP solutions produced smaller posterior shear forces (max diff:

0.51 N/kg), and greater medial shear (max diff: 0.46 N/kg) and

compressive forces (max diff: 0.82 N/kg) at the knee joint.

KINEMATICS

KINETICS

Figure 4: A) Example participant displaying

the distraction/compression results with

unfiltered (blue) and filtered (black) NoBP

constraints compared to unfiltered BP

constrained results (red). B) Participant with

20 mm of compression 130 ms after contact

Ext

Flex

Add

Abd

Ant

Lat

Med

Com

Dis

Int

Ext

Post

-100 0 100 200 300-5

0

5

10

15

20

25

time (ms)

Kn

ee

Dis

/Co

m T

ran

s (

mm

) Com

Dis

A B

-40 -30 -20 -10 0

-40

-20

0

Kne

e F

lex/

Ext

Ang

le (°

)

-40 -30 -20 -10 0-10

-5

0

5

10

15

Kne

e A

bd/A

dd A

ngle

(°)

-40 -30 -20 -10 0-15

-10

-5

0

5

10

Knee Flex/Ext Angle (°)

Kne

e In

t/Ext

Rot

Ang

le (°

)

-40 -30 -20 -10 0-10

-5

0

5

Kne

e M

ed/L

at T

rans

(mm

)

-40 -30 -20 -10 0

-5

0

5

10

15

Kne

e A

nt/P

ost T

rans

(mm

)

-40 -30 -20 -10 0-20

-10

0

10

Knee Flex/Ext Angle (°)

Kne

e D

is/C

om T

rans

(mm

)

Ext

Flex

Add

Abd

Ant

Lat

Med

Com

Dis

Int

Ext

Post