Biomedical Engineering Reference
In-Depth Information
4 Discussion and Conclusions
The CV of the congruity indices was equal or better than the CV of the existed
congruity indices [
8
]. However, it is difficult to compare since our congruity
quantification was a local measure and limited to the CA and also different methods
were used to locate the cartilage. Our congruity measure was local and limited
to CA, which is likely the most interesting region. We employed mean curvature
flow in a level set formulation since it was proved to be a better choice in terms
of accuracy at computing the curvature limited by voxel resolution [
16
]. However, a
tibiofemoral shape model could allow computation of the congruity at coarse scale
and we would like to investigate this in future.
The congruity measure is dependent on the scale at which the local normal
curvatures computed. This is because at higher curvature scales, the curvature
values were lower thereby resulted in low congruity values. At lower scales the
higher incongruities found in the edges of the contact region whereas at higher
scales, the incongruities shift to the central regions of the CA. This may be due to
that at higher scales, the edges become smoother and have less curvature. Never-
theless, the cross-validation decided the best curvature scale and iteration number
(parameters) for each diagnosis task. Another limitation is the computational time.
We would like to compute them with voxel super sampling that might increase the
robustness of the measure. A shape model may be a good alternative to this.
We are limited to over CA since the scans were acquired only at a specific angle.
Ideally, the overall CI is the integration of the CI at different angles of flexion/
extension since CA changes with angle of flexion or in other words during the daily
functional activities.
The medial knee joint congruity was quantified noninvasively from MRI and
was capable of contrasting the healthy from OA knees. The congruity is in general
highest in healthy knees and decreases with onset of OA and further with disease
severity (AUC to separate KL
1 vs. KL
0.0001). The same
framework could be also applied to other knee compartments and maybe to other
load bearing joints in the body. Future studies could include the meniscus while
computing the overall MTF congruity using the methods presented to elucidate the
congruity changes with disease progression.
>
1 was 0.73,
p
<
Acknowledgements The authors would like to thank the Center for Clinical and Basic Research
for providing the MRI scans and radiographic readings. This work has been funded by the
Danish Strategic Research Council through the grant “Learning Imaging Biomarkers” (grant no.
09-065145).
