A better knowledge of the kinematics behavior of total knee replacement (TK
R) during activity still remains a crucial issue to validate innovative pro
sthesis designs and different surgical strategies. Tools for more accurate
measurement of in vivo kinematics of knee prosthesis components are therefo
re fundamental to improve the clinical outcome of knee replacement. In the
present study, a novel model-based method for the estimation of the three-d
imensional (3-D) position and orientation (pose) of both the femoral and ti
bial knee prosthesis components during activity is presented. The knowledge
of the 3-D geometry of the components and a single plane projection view i
n a fluoroscopic image are sufficient to reconstruct the absolute and relat
ive pose of the components in space. The technique is based on the best ali
gnment of the component designs with the corresponding projection on the im
age plane. The image generation process is modeled and an iterative procedu
re localizes the spatial pose of the object by minimizing the Euclidean dis
tance of the projection rays from the object surface, Computer simulation a
nd static/dynamic in vitro tests using real knee prosthesis show that the a
ccuracy with which relative orientation and position of the components can
be estimated is better than 1.5 degrees and 1.5 mm, respectively. In vivo t
ests demonstrate that the method is well suited for kinematics analysis on
TKR patients and that good quality images can be obtained with a carefully
positioning of the fluoroscope and an appropriate dosage, With respect to p
reviously adopted template matching techniques, the present method overcome
s the complete segmentation of the components on the projected image and al
so features the simultaneous evaluation of all the six degrees of freedom (
DOF) of the object. The expected small difference between successive poses
in in vivo sequences strongly reduces the frequency of false poses and both
the operator and computation time.