Validation of Radiocarpal Joint Contact Models Based on Images from a Clinical MRI Scanner

1
Johnson J E;
2
McIff T E;
2
Lee S-P;
2
Toby E B; +
1
Fischer K J
+
1
University of Kansas, Lawrence, KS,
2
University of Kansas Medical Center, Kansas City, KS
Senior author fischer@ku.edu

INTRODUCTION:
A means to evaluate in vivo joint mechanics during functional loading
would be a very useful tool in orthopaedic biomechanics. The contact
area and contact pressure distribution data could reveal the in vivo
effects of injury and the efficacy of surgical repair/reconstruction
subsequent to an injury. Long-term data from injuries, may provide
insight into the level of changes in joint mechanics that will result in
joint degeneration [1]. With such data, in vivo joint mechanics data
could also be used clinically to determine risk of OA following injury.
Joint modeling based on MRI imaging during functional loading has the
potential to provide in vivo joint mechanics data, and preliminary studies
indicate feasibility [2]. The objective of this study was to validate this
image-based modeling approach in the radiocarpal joints, by comparison
with specimen-specific cadaveric experimental data. We hypothesized
that the contact areas from image-based modeling would be within 15%
of experiment data and contact areas measured directly from MRI, and
that joint contact forces would be within 20% of experiment data.

METHODS:
Three cadaver specimens were dissected to allow loading of the FDP,
FDS, and FPL tendons to simulate grasp, with loading of the ECR and
ECU to prevent wrist flexion. Loading fixtures were non-ferromagnetic,
so that wrist loading could be performed in an MRI scanner. The volar
wrist capsule was sectioned along the radiocarpal border to allow
insertion of a Tekscan pressure sensor into the radiocarpal joints to
measure contact pressures, area and force during simulated grasp.
Each specimen was imaged at the wrist without loading in a 3T MRI
scanner. A custom surface coil and a constructive interference steady
state (CISS) sequence were used to obtain images of the cartilage and
bones, with a 0.5 mm slice thickness and 0.15 mm in-plane pixel size.
Unloaded images were used to construct surface models of the radius,
scaphoid, and lunate, which included the cartilage surfaces.
Each specimen was also imaged at the wrist during the application of
the digital flexor and wrist extensor tendon loads. These images were
acquired with half the resolution, to bring scan time low enough for
future in vivo active grasp scans (196 seconds).
The radius, scaphoid, and lunate bones without cartilage (from both
image sets) were isolated on a black background in separate image sets
for image registration. The radius was selected as the fixed reference.
Registration of the loaded radius to the unloaded radius provided the
transformation between the two image sets. The images of the loaded
scaphoid and lunate were transformed into the coordinate system of the
unloaded images. Then the unloaded scaphoid image set was registered
to the (transformed) loaded scaphoid image set (and similarly for the
lunate). This provided the kinematic transformations of the carpal bones
from the unloaded state to the loaded configuration.
The carpal kinematics were implemented with the surface contact
models in the Joint_Model software [3]. The area of surface penetration
defined the contact area, the local interpenetration defined the local
cartilage deformation. Based on uniform cartilage thickness (1 mm) and
an effective cartilage modulus of 4 MPa, local contact pressure was
calculated from local surface penetration. Contact pressures were
integrated over the contact area to obtain contact force. Contact area was
compared to contact area directly measured from the MRI images and
from the experimental Tekscan data. Contact force and peak contact
pressures were compared to those from the experimental Tekscan data.

RESULTS SECTION:
Results are currently analyzed for only one specimen. Qualitatively,
the relative size and location of scaphoid and lunate contact on the
radius correspond well (Fig. 1). Quantitatively, the model radioscaphoid
contact area matched the Tekscan data precisely (within 10%) and was
within 15% of the directly measured contact area (Fig. 2). Model
radiolunate contact area was also within 15% of both the Tekscan and
the directly measured contact area, except for Specimen 2. Tekscan and
direct area measurements matched each other very precisely.
Force data from the model was off by somewhat more than 20% for
both radioscaphoid and radiolunate contact, but values were reasonable
(Table 1). Peak pressure data from the model showed the largest
discrepancies, as they were substantially lower than the Tekscan values.

Table 1: Complete data comparisons for validation study
Spec Pair Measurement Tekscan Model Direct
Force (N) 35.3 20.7
Area (mm
2
) 39.9 49.2 43.4
RL
Peak Pressure (MPa) 4.59 0.98
Force (N) 39.13 187.1
Area (mm
2
) 105.2 103.9 89.9
1
RS
Peak Pressure (MPa) 3.15 2.98
Force (N) 15.2 13.3
Area (mm
2
) 108.9 58.3 100.2
RL
Peak Pressure (MPa) 1.02 0.88
Force (N) 24.0 58.6
Area (mm
2
) 52.0 69.3 55.9
2
RS
Peak Pressure (MPa) 1.94 1.74
Force (N) 79.2 94.1
Area (mm
2
) 130.6 110.8 124.8 RL
Peak Pressure (MPa) 2.98 1.43

Force (N) 46.2 32.1

Area (mm
2
) 60.5 65.8 63.4
3
RS
Peak Pressure (MPa) 5.10 1.00


DISCUSSION:
Contact area data is quite consistent and indicates reliability of all
measurement methods. Contact force appears reasonably accurate
compared to experimental measures, though it did not fall within the
20% threshold of the experimental measurement (about 2 times the
accuracy of experimental measurements). Lower image resolution in the
scans acquired with loading made direct measurement of contact area
somewhat more difficult than prior experiments. Peak pressure is most
difficult to precisely determine, as it is highly dependent on both
geometry and kinematics. Accurate peak pressures are, however,
important for the proposed applications of image-based modeling. While
the method is not perfect, it could provide reasonably good in vivo data.

REFERENCES:
1. Andriacchi, T.P., et al., A Framework forOsteoarthritis at the Knee.
Annals of Biomedical Engineering, 2004. 32(3): p. 447-457.
2. Pillai, R.R., et al., MRI-based modelingduring active light grasp.
J Biomech, 2007. 40(12): p. 2781-2787.
3. Kwak, S.D., et al., Formulation for Models of Diarthrodial
Joints. Comp Meth Biomech Biomed Eng, 2000. 3(1): p. 41-64.

(a) (b)
Figure 1. Tekscan pressure data (a) vs. MRI-based model data (b) for
lunate contact (right) and scaphoid contact (left). Contact areas and
relative pressure intensities are consistent between the measures.
0
20
40
60
80
100
120
140
RL RS RL RS RL RS
S1 S2 S3
Area (mm2)
Tekscan Model Direct
0
20
40
60
80
100
120
140
RL RS RL RS RL RS
S1 S2 S3
Area (mm2)
Tekscan Model Direct

Figure 2. Contact area from all specimens and all measurement
methods show very consistent measures.
Poster No. 1862 55th Annual Meeting of the Orthopaedic Research Society