Accepted Manuscript Determining 3D kinematics of the hip using video fluoroscopy: guidelines for balancing radiation dose and registration accuracy Fabio D’Isidoro, Patrik Eschle, Thomas Zumbrunn, Christian Sommer, Stephan Scheidegger, J. Ferguson Stephen PII:
S0883-5403(17)30475-8
DOI:
10.1016/j.arth.2017.05.036
Reference:
YARTH 55903
To appear in:
The Journal of Arthroplasty
Received Date: 10 November 2016 Revised Date:
19 April 2017
Accepted Date: 16 May 2017
Please cite this article as: D’Isidoro F, Eschle P, Zumbrunn T, Sommer C, Scheidegger S, Ferguson Stephen J, Determining 3D kinematics of the hip using video fluoroscopy: guidelines for balancing radiation dose and registration accuracy, The Journal of Arthroplasty (2017), doi: 10.1016/ j.arth.2017.05.036. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Title page Type of Submission: “Original Paper”
RI PT
Title:
Determining 3D kinematics of the hip using video fluoroscopy: guidelines for balancing radiation dose
M AN U
Author names and affiliations:
SC
and registration accuracy.
D’Isidoro Fabioa, Eschle Patrikb, Zumbrunn Thomasa, Sommer Christianb, Scheidegger Stephanb, Ferguson Stephen J.a
ETH Zurich, Institute for Biomechanics, Hönggerbergring 64, 8093 Zürich, Switzerland.
b
ZHAW, School of Engineering, Technikumstrasse 9, 8400 Winterthur, Switzerland.
EP
TE D
a
Contact of corresponding author:
AC C
E-mail:
[email protected]
Address: HPP O 14, Hönggerbergring 64, 8093 Zürich, Switzerland Tel: +41 44 633 40 60
ACCEPTED MANUSCRIPT Title page
2
Title:
3
Determining 3D kinematics of the hip using video fluoroscopy: guidelines for balancing radiation dose
4
and registration accuracy.
RI PT
1
5 Abstract
7
Background
8
Video fluoroscopy is a technique currently used to retrieve the in vivo 3D kinematics of human joints
9
during activities of daily living. Minimization of the radiation dose absorbed by the subject during the
10
measurement is a priority and has not been thoroughly addressed so far. This issue is critical for the
11
motion analysis of the hip joint, due to the proximity of the gonads. The aims of this study were to
12
determine the X-ray voltage and the irradiation angle that minimize the effective dose and to achieve
13
the best compromise between delivered dose and accuracy in motion retrieval.
14
Methods
15
Effective dose for a fluoroscopic study of the hip was estimated by means of Monte Carlo simulations
16
and dosimetry measurements. Accuracy in pose retrieval for the different viewing angles was
17
evaluated by registration of simulated radiographs of a hip prosthesis during a prescribed virtual
18
motion.
19
Results
20
Absorbed dose can be minimized to about one sixth of the maximum estimated values by irradiating
21
at the optimal angle of 45° from the posterior side and by operating at 80 kV. At this angle, accuracy
22
in retrieval of internal-external rotation is poorer compared to the other viewing angles.
AC C
EP
TE D
M AN U
SC
6
1
ACCEPTED MANUSCRIPT Conclusion
24
The irradiation angle that minimizes the delivered dose does not necessarily correspond to the
25
optimal angle for the accuracy in pose retrieval, for all rotations. For some applications, single-plane
26
fluoroscopy may be a valid lower dose alternative to the dual-plane methods, despite their better
27
accuracy.
RI PT
23
28 Keywords:
30
Video fluoroscopy; Hip Joint; Kinematics; Effective Dose; Image registration; Total Hip Arthroplasty.
M AN U
SC
29
31 Introduction
33
Video fluoroscopy is an accurate and minimally-invasive technique currently used for the analysis of
34
in vivo joint kinematics. A sequence of fluoroscopic images of the joint of interest is collected while
35
the subject performs a specific motion task. For each frame, the 3D pose of the joint is retrieved by
36
matching a 3D model of each segment to its 2D projection in the corresponding image, i.e. 2D-to-3D
37
registration [1]. The movement is finally reconstructed from the sequence of registered poses.
38
Fluoroscopy of the hip has been performed to analyze gait on a treadmill [2–5], stair climbing [6] and
39
isolated abduction [2,7].
40
As with conventional radiography, fluoroscopy imaging involves exposure of the subject to ionizing
41
radiation. According to the ALARA (As Low As Reasonably Achievable) radiation safety principle, all
42
reasonable methods should be employed to minimize radiation dose. In video fluoroscopy, a trade-
43
off exists between image quality and radiation exposure. Acquisition of image data at high frame
44
rates for dynamic events requires greater X-ray exposure to increase brightness and obtain the image
45
quality required to achieve sufficient registration accuracy [8]. Furthermore, X-ray imaging of the hip
AC C
EP
TE D
32
2
ACCEPTED MANUSCRIPT requires typically higher exposure compared to imaging other peripheral joints of the body, because
47
the image contrast is limited by a greater amount of surrounding tissue. However, higher X-ray
48
exposure is associated with increased dose absorbed by the subject [9]. This issue is particularly
49
critical for the hip due to the proximity of the gonads: the risk of biological damage to the gonads
50
attributed by the International Commission on Radiation Protection (ICRP) is two times the risk
51
assigned to bladder, liver and thyroid glands and eight times the risk assigned to the bone surface,
52
skin and brain [10]. The radiation dose is doubled when dual-plane fluoroscopy is used to acquire a
53
pair of simultaneous images [4,5] as a means to improve the accuracy in pose retrieval.
54
Many in vivo fluoroscopic studies of the hip lack information concerning radiation exposure
55
[2,6,7,11–14] or report only the fluoroscopic settings without dose estimations [4]. The studies that
56
provided values for the absorbed dose [5,15–17] did not explain the method used for their
57
estimations and did not characterize the dependency of the dose on the imaging parameters.
58
Effective dose can be estimated from the product between the kerma-area product, usually provided
59
by the fluoroscope, and a reference dose conversion factor (DCC) [18] that is specific for the
60
diagnostic procedure and the irradiated organs [19]. However, the values for the kerma-area product
61
are not validated with in-situ dosimetry measurements and the reference values for the DCC,
62
although derived from Monte-Carlo simulations using anthropomorphic digital phantoms, are not
63
specific for the subject (gender, BMI), for the measurement geometry (distance and position of the
64
patient with respect to the beam) and for the imaging settings (voltage, current, pulse width). For
65
example, Le Heron [20] estimated that for the trunk region the DCC for the lateral and the posterior-
66
anterior (PA) radiographic projections are approximately half those corresponding to anterior-
67
posterior (AP) projection.
68
Considering the safety implications of radiation exposure, and the discrepancies in reported radiation
69
dose for current fluoroscopy studies, the first aim of this study was to accurately estimate the
70
radiation dose during hip imaging and to determine the imaging parameters that minimize the dose
AC C
EP
TE D
M AN U
SC
RI PT
46
3
ACCEPTED MANUSCRIPT while providing acceptable image quality. The variable parameters investigated for this analysis were
72
the X-ray voltage and the irradiation angle with respect to the pelvis.
73
For single-plane use, the fluoroscope must be positioned carefully to obtain sufficient bony details
74
from optimized viewing angles, while minimizing occlusion of surrounding tissues and out-of-plane
75
movement for the analyzed activity [15]. Hence, registration accuracy is influenced by imaging angle
76
and must be considered together with dose minimization. The second aim of this study was to
77
quantitatively investigate the dependence of the registration accuracy on the viewing angle of single-
78
plane fluoroscopy, for a specific motion task.
SC
RI PT
71
M AN U
79 Material and methods
81
Estimation of dose
82
Effective dose (Sv) for a fluoroscopic study of the hip was estimated by means of Monte Carlo
83
simulations and dosimetry measurements with an Alderson phantom. Estimations were obtained for
84
two different X-ray voltages, 80 kV and 100 kV, and five different irradiation angles described in Fig.
85
1. The values for the X-ray voltages were comparable to those used in previous X-ray studies of the
86
hip [4,5,15–17,21]. On the clinical fluoroscope incorporated in the imaging system, tube current is
87
not independently adjustable. The tube current was automatically set by the fluoroscope to 12 mA
88
for both voltages.
89
Monte Carlo methods with the Geant4 library [22] were used to simulate the X-ray irradiation of a 3D
90
human model with 73 different organs and a voxel size of 2x2x2 mm3. Material properties of each
91
organ were assigned according to the tissue composition provided by the ICRP [23]. The X-ray beam
92
used in the simulations was modeled according to the specifications of the used fluoroscope (BV
93
Pulsera, Philips Medical Systems, Switzerland). The beam aperture was 17°. The output of the
94
simulation was the average absorbed dose (Gy) for each organ, a deterministic quantity
AC C
EP
TE D
80
4
ACCEPTED MANUSCRIPT accounting for the amount of deposited energy. The simulated effective dose was computed as
96
the weighted sum of the absorbed doses over all organs of the body (Equation 1). The organ-specific
97
weighting factors account for the stochastic biological risk associated with radiation exposure
98
and are defined by the ICRP. = ( 1)
99
RI PT
95
Simulations at different irradiation angles (Fig. 1) and different X-ray voltages were performed. With experimental setups analogous to each simulation, the entrance dose to an Alderson phantom
101
placed at a distance of 70 cm from the X-ray source was measured while operating the fluoroscope at
102
a frame rate of 25 Hz and pulse width of 8 ms (Fig. 2). For each case, the dose-area product (Gy*m2)
103
was measured with a dosimeter. The simulated effective dose was then scaled with the ratio of the
104
measured dose-area product to the simulated dose-area product (Equation 2), in order
105
to retrieve the estimated effective dose for each measurement setting.
M AN U
SC
100
( 2)
TE D
=
Due to the different position between the female and the male gonads, simulations for both a female
107
subject (58 Kg, 1.66 m, BMI 21) and a male subject (80 Kg, 1.73 m, BMI 27) were carried out.
108
Additionally, simulations for a shielded male patient were performed in order to quantitatively
109
evaluate the benefit brought by gonad protection. Shielding was modeled with a 2 to 4 mm thick lead
110
sheet placed in front of the testes.
111
Estimation of registration accuracy
112
Pose accuracy for the different viewing angles was evaluated by registration of simulated
113
radiographs of a hip prosthesis during a prescribed virtual motion. Since accuracy is expected to
114
depend on the specific motion task, level walking and sitting on a chair were chosen as activities
AC C
EP
106
5
ACCEPTED MANUSCRIPT 115
covering typical Range of Motion (ROM) of the hip joint. For each of these, the following procedure
116
was carried out:
118 119
a) Realistic motion of the hip was retrieved from the public database www.OrthoLoad.com [24] and applied to 3D models of the cup and the femoral stem of the prosthesis;
RI PT
117
b) A simulated fluoroscopic image was generated by a perspective projection model and for each viewing angle a sequence of simulated radiographs of the hip prosthesis during motion
121
was analysed (Fig. 3);
123 124 125
c) Implant pose at each motion step was determined from the images by means of 2D-3D registration;
M AN U
122
SC
120
d) Accuracy was evaluated by calculating the registration error, as the difference between the registered pose and the prescribed pose over the whole motion sequence. 2D-3D registration was performed with the GUI provided by the open source software Joint Track
127
[25,18]. Manual matching was performed to exclude the influence of a specific automatic registration
128
algorithm. Additionally, a reproducibility study of the results over multiple manual registrations was
129
carried out to evaluate the inter-user variability. One single-position was chosen and registered 10
130
times per viewing angle.
131
The coordinate systems of the acetabular cup and the femoral components were defined according
132
to the standard recommendation of the ISB [26]. For each motion step, the implant pose relative to
133
the reference standing position was described as a sequence of consecutive flexion, abduction and
134
internal rotation of the femoral stem with respect to the cup followed by translation. Registration
135
error was computed for each of the three rotational and the three translational components
136
describing the implant pose. However, due to the symmetry of the cup around its geometrical axis,
137
the rotational position for the cup cannot be fully determined. Therefore, the registration errors for
138
rotation were evaluated for the femoral component only.
AC C
EP
TE D
126
6
ACCEPTED MANUSCRIPT 139
Accuracy for each viewing angle was evaluated by computation of the RMS value of the registration
140
error over a whole motion sequence. Statistical differences between the registration errors for each
141
angle were evaluated with a one-way ANOVA (significance level of 0.05) followed by a multi-
142
comparison test with Tukey’s correction.
RI PT
143 Results
145
Estimation of the dose
146
The dose is generally lower for lower voltages and for irradiation from the back (Fig. 4). Values
147
estimated for 100 kV were higher than those predicted for 80 kV, by a factor of 2.0 – 2.2. The highest
148
doses were found for AP and APO irradiations, where the gonads are more exposed to the beam and
149
less shielded by the surrounding tissues. Effective dose can be reduced to about one sixth of the
150
highest value by collecting X-ray images with a voltage of 80 kV and a 45° posterior angle (PAO).
151
The values estimated for a female phantom were higher than those for a male phantom by a factor
152
ranging from 1.3 to 1.9. The main contribution to the higher effective dose comes from the ovaries:
153
these are more directly exposed than the testes, since they are located at the same height as the hip
154
joint.
155
Simulations for a male patient showed that gonad protection with the metal shielding modeled in
156
this study does not significantly reduce the dose. Only the effective dose for AP irradiation was
157
reduced by 10% relative to the unshielded case.
158
Estimation of the registration accuracy
159
RMS of the registration error over a whole motion sequence at the different viewing angles is
160
reported in Table 1 for each of the three rotational components describing the pose of the femoral
161
component of the hip prosthesis.
AC C
EP
TE D
M AN U
SC
144
7
ACCEPTED MANUSCRIPT Generally lower accuracy was observed for the rotations involving larger out-of-plane movement. For
163
gait, flexion was registered from the lateral view with the smallest RMS error of 0.17° while RMS
164
errors for the PA and the PAO views were 1.34° and 1.41° RMS, respectively. Abduction was
165
registered with the smallest RMS error of 0.16° from the PA view, while RMS errors for the PAO and
166
the L views were 1.26° and 2.89°, respectively. For sitting down on a chair, abduction was retrieved
167
with the smallest RMS error of 0.2° from a PA view, while poorer performance was seen from the
168
other angles. However, no substantial differences in accuracy were found for the flexion component,
169
since RMS errors for all angles were around 0.5°. For internal rotation, the oblique view (PAO) was
170
found to provide the worst accuracy for both gait and sitting down on a chair, showing a RMS error
171
of 2° and 1.7°, respectively; on the contrary, sub-degree errors were observed from the PA and the L
172
views.
173
Results from the reproducibility study (Fig. 5) over multiple registrations (n =10) showed statistically
174
significant differences in registration error for different irradiation angles. The observed trends in
175
pose accuracy over the full motion cycle were therefore confirmed also for reproducibility: the
176
frontal view (PA or AP) provides statistically better results for abduction (p