=Paper=
{{Paper
|id=Vol-1477/paper26
|storemode=property
|title=2D-3D reconstruction-based implant migration measurement
|pdfUrl=https://ceur-ws.org/Vol-1477/Proceedings_CURAC_2013_Paper_26.pdf
|volume=Vol-1477
|dblpUrl=https://dblp.org/rec/conf/curac/ThelenBSNZ13
}}
==2D-3D reconstruction-based implant migration measurement==
https://ceur-ws.org/Vol-1477/Proceedings_CURAC_2013_Paper_26.pdf
2D-3D reconstruction-based implant migration measurement
B. Thelen, S. Balestra, S. Schumann, L. Nolte, G. Zheng
Institute for Surgical Technology & Biomechanics, University of Berne, Switzerland
Contact: benedikt.thelen@istb.unibe.ch
Abstract:
This paper presents a phantom-based study for validating a newly developed 2D-3D reconstruction-based method for
measuring implant migration after total hip arthroplasty (THA). Based on a mock-up setup, three different methods
were used to determine the cup orientation with respect to the anterior pelvic plane (APP) of a plastic pelvis: (a) the
optical tracking method; (b) the fiducial-based method; and (c) the 2D-3D reconstruction-based method. It was found
that the incremental anteversion and inclination angles measured by the newly developed 2D-3D reconstruction-based
method are comparable with those measured by the other two methods.
Keywords: implant migration, 2D-3D reconstruction, optical tracking
1 Background
The initial and the secondary implant positions after the cementless total hip arthroplasty (THA) documents the stability
of the implanted components. Postoperative migration of implants is therefore one important parameter for initial and
follow-up quality documentation. Due to the concerns on cost and radiation, postoperative migration is typically meas-
ured on two-dimensional (2D) postoperative anteroposterior (AP) pelvic X-ray radiographs. The existing methods either
determine only a 2D migration in the AP pelvic X-ray radiograph plane [2] or require invasive implantation of metal
fiducials to determine precise three-dimensional (3D) migration [1]. Recently, a precise X-ray image calibration method
as well as a 2D-3D reconstruction method has been reported for a robust and accurate reconstruction of a 3D patient-
specific model of acetabular surface from 2D X-ray radiographs [3]. However, there is little known work reported to use
2D-3D reconstruction for implant migration measurement. The aim of this work is to determine the incremental orienta-
tion variation of a cup implant using 2D-3D reconstruction techniques and to validate the methodology using optical
tracking and implanted metal fiducials.
2 Materials and Methods
Mockup setup: First a mockup was built that allowed simulating cup migration with respect to a fixed pelvis. The
mockup was composed of a cadaver pelvis that was not part of the statistical shape model (SSM) population but repre-
sented by it (male, caucasian), a plastic cup and a fixation frame made from X-ray translucent Plexiglas. The construc-
tion allowed the cup to move on a spherical joint by hand and to hold the position during acquisition, while the pelvis
was fixed permanently to the frame (see figure 1).
Four metal markers where placed coplanar with the cup opening plane. The placement was designed to provide reliable
landmarks in the X-ray images. Furthermore ten metal nails with indents that allowed pointer tool picking were placed
on the pelvis to mark the anterior pelvic Plane (APP) and six other prominent landmarks. During acquisition the orienta-
tion of the cup with respect to the APP was determined using an infrared passive optical tracking camera from Northern
Digital Inc. Ontario, Canada (see fig. 1).
107
Figure 1: (a) Drawing of the mockup, (b) experimental setup showing the cadaver bone, the phantom (fixed to the sa-
crum), the base construction with the joint allowing the cup to rotate and the optical markers.
X-ray image acquisition: The acquisition started with the fixation of a calibration phantom to the pelvis. A pointer digit-
ization based registration of the optical tracking space and the fiducails as well as the cup opening plane was performed.
Next, 15 datasets were acquired. Each dataset was composed of three images (AP, outlet, 30° oblique). For each dataset
the cup was oriented by hand with the help of the optical tracking software. All images were acquired with a standard
clinical Philips X-ray machine. During acquisition, the orientation of the cup was tracked by the camera giving absolute
readings for the anteversion and inclination angles at a rate of approx. 20 Hz. All positions recorded by the camera
where stored to allow detailed analysis during evaluation. The images were stored in DICOM format for further pro-
cessing.
Evaluation: The experimental setup provided three sources of measurements for the anteversion and inclination angles.
(a) The optical tracking system provided a real time estimation of the angles using the optical tracking markers placed
on the cup and the base frame. These values were stored onto hard disk. Measurements obtained by this method are la-
beled as ‘optical’. (b) Through a semi-automatic algorithm using Hough transform the X-rays where calibrated and a
custom software was used to pick the markers on the X-ray’s and determine the 3D position of the markers. Next, these
3D points where used to determine the APP orientation. Since the plastic cup also contained a metal ring parallel to the
Figure 2: (a) Fiducial based APP with AP radiograph, (b) reconstructed pelvis model with APP (blue) and fiducial
based APP overlayed (green) for comparison
108
cup plane it was used to determine the 3D orientation of the cup. (c) With the help of a SSM-based 2D-3D reconstruc-
tion algorithm [3] the pelvis was reconstructed for each dataset from the three calibrated X-ray images. The APP was
then extracted based on four vertices determined on the mean model (see fig.2b). In the SSM-based 2D-3D reconstruc-
tion method four 3D points where determined from the reconstructed shape model using fixed vertices. The cup orienta-
tion was determined as in method (b). During the study 15 reconstructions where performed and the reconstructed sur-
faces where compared to a reference surface from a computer tomography (CT) scan of the used bone, the mean
Hausdorff distance of the surfaces was 1.44 ± 0.15 mm. The anteversion and inclination angles between these planes are
labeled ‘reconstructed’.
3 Results
An extensive comparison of the incremental angles (difference of angle in a dataset to the angles of the first dataset)
was conducted. The following plots show the correlation of the datases measured with the three different methods.
Fiducial-Optical Reconstruction-Optical Reconstruction-Fiducial
Anteversion [°] Inclination [°] Anteversion [°] Inclination [°] Anteversion [°] Inclination [°]
Mean 1.95 0.45 2.18 -0.24 -0.23 0.69
STD 0.81 0.35 0.96 0.42 0.56 0.33
Max 3.12 0.92 3.95 1.35 1.13 1.18
Min 0.29 0.05 0.48 0.05 0.07 0.03
Table 1: The differences of incremental measurements done with the different methods
109
Correlation of anteversion (Absolute)
Optical Fiducial Reconstructed
Optical 1 0.51 0.48
Fiducial ‐ 1 0.93
Reconstructed ‐ ‐ 1
Correlation of inclination (Absolute)
Optical Fiducial Reconstructed
Optical 1 0.96 0.95
Fiducial ‐ 1 0.92
Reconstructed ‐ ‐ 1
Correlation of anteversion (Incremental)
Optical Fiducial Reconstructed
Optical 1 0.93 0.88
Fiducial ‐ 1 0.97
Reconstructed ‐ ‐ 1
Correlation of inclination (incremental)
Optical Fiducial Reconstructed
Optical 1 0.96 0.94
Fiducial ‐ 1 0.96
Reconstructed ‐ ‐ 1
Figure 3: The linear correlation between the incremental measurements done with the tree methods shown graph-
ically and a table containing the correlation values for the incremental and the absolute angles of the three methods.
4 Discussion
As the main interest in this work is the migration of the cup, the absolute angles obtained by the three methods play a
secondary role. The biggest focus is given to the incremental angles of each method. When comparing the absolute and
the incremental angles of all three measurement types we noticed that the difference of the obtained results is smaller
for the inclination than for the anteversion. Table 1 shows the differences between the incremental measurements. The
difference between the reconstruction based method and the fiducial based method is smaller than the difference of the
optical method to any of the two methods. When comparing the correlations of the incremental and absolute angles as
shown in figure 3, we notice that reconstructed and fiducial based angles correlate very well, while the optical method
do not always correlate. The reason for this is the calibration which is the basis of both the reconstructed and the
fiducial measurement method. On one hand the size of the used calibration phantom limits the quality of the X-ray cali-
bration. On the other hand the close correlation indicates that the error produced by the 2D-3D reconstruction is rela-
tively small.
110
5 Conclusion and future work
In this paper, we propose a method to determine the cup orientation from calibrated X-ray images and SSM based 2D-
3D reconstruction. From the results we conclude that the non-invasive SSM-based 2D-3D reconstruction method pro-
vides a valid method to determine the cup orientation as compared to fiducial based methods.
The main error contribution of our method at this point is the calibration of the X-ray images. Since the used phantom
was originally designed for smaller scaled applications like C-arms, it needs to be redesigned in the future. For the de-
termination of the cup orientation we used a metal ring inside the plastic cup instead of the initially implanted markers.
While this method produced more repeatable results than using only the four markers it is also applicable in real surgi-
cal cases where the orientation of the cup in a planar x’ray may only be determined form the silhouette, but it also
makes in plane rotation difficult to track.
6 Acknowledgements
Special thanks to Urs Rohrer (ISTB) and Patrick Moser (ISTB) for their help in developing the model frame. We would
also like to thank the MTRA team from the Inselspital Bern under direction of Prof. Dr. Dr. J. T. Heverhagen for the ac-
cess to the X-ray equipment.
7 References
[1] Kiss J, Murray DW, Turner-Smith AR, Bithell J, Bulstrode CJ, Migration of cemented femoral components after
THR: Roentgen stereophotogrammetric analysis, J Bone Joint Surg Br, 78(5) 796-801 (1996)
[2] Wilkinson JM, Hamer AJ, Elson RA, Stockley I, Eastell R, Precision of EBRA-digital software for monitoring
implant migration after total hip arthroplasty, J Arthroplasty, 17(7) 910-916 (2002)
[3] Schumann S, Liu L, Tannast M, Bergmann M, Nolte LP, Zheng G, An integrated system for 3D hip joint recon-
struction from 2D X-rays: A preliminary validation study, Ann Biomed Eng, in press, 2013.
111