Home Search Collections Journals About Contact us My IOPscience

Automatic image registration performance for two different CBCT systems; variation with

imaging dose

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2014 J. Phys.: Conf. Ser. 489 012070

(http://iopscience.iop.org/1742-6596/489/1/012070)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 203.32.142.60
This content was downloaded on 05/10/2016 at 22:01

Please note that terms and conditions apply.

You may also be interested in:

Zonal cone beam CT for increased contrast and reduced dose

C J Moore, T E Marchant and A M Amer

Investigation of CBCT-CT image registration uncertainties in IGRT

J R Sykes, D S Brettle, D R Magee et al.

Dosimetry of CBCT: methods, doses and clinical consequences

J R Sykes, R Lindsay, G Iball et al.

Geometric validation of MV topograms for patient localization on TomoTherapy

Janid P Blanco Kiely, Benjamin M White, Daniel A Low et al.

Improved compressed sensing-based cone-beam CT reconstruction using adaptive prior image

constraints
Ho Lee, Lei Xing, Ran Davidi et al.

Dose assessment from an online kilovoltage imaging system in radiation therapy

Ernest K Osei, Bryan Schaly, Andre Fleck et al.

Dose calculation accuracy using CBCT for pelvic adaptive radiotherapy

Huaiqun Guan and Hang Dong
XVII International Conference on the Use of Computers in Radiation Therapy (ICCR 2013) IOP Publishing
Journal of Physics: Conference Series 489 (2014) 012070 doi:10.1088/1742-6596/489/1/012070

Automatic image registration performance for two different

CBCT systems; variation with imaging dose

J BARBER1,3, J R SYKES2, L HOLLOWAY3, D I THWAITES3

1
Nepean Cancer Care Centre, Penrith, NSW, Australia
2
St James's Institute of Oncology, Leeds, UK
3
Institute of Medical Physics, University of Sydney, Sydney, NSW, Australia

[email protected]

Abstract. The performance of an automatic image registration algorithm was compared on

image sets collected with two commercial CBCT systems, and the relationship with imaging
dose was explored. CBCT images of a CIRS Virtually Human Male Pelvis phantom (VHMP)
were collected on Varian TrueBeam/OBI and Elekta Synergy/XVI linear accelerators, across a
range of mAs settings. Each CBCT image was registered 100 times, with random initial
offsets introduced. Image registration was performed using the grey value correlation ratio
algorithm in the Elekta XVI software, to a mask of the prostate volume with 5 mm expansion.
Residual registration errors were calculated after correcting for the initial introduced phantom
set-up error. Registration performance with the OBI images was similar to that of XVI. There
was a clear dependence on imaging dose for the XVI images with residual errors increasing
below 4mGy. It was not possible to acquire images with doses lower than ~5mGy with the
OBI system and no evidence of reduced performance was observed at this dose. Registration
failures (maximum target registration error > 3.6 mm on the surface of a 30mm sphere)
occurred in 5% to 9% of registrations except for the lowest dose XVI scan (31%). The
uncertainty in automatic image registration with both OBI and XVI images was found to be
adequate for clinical use within a normal range of acquisition settings.

1. Introduction
The high uptake of kV imaging systems mounted on medical linear accelerators in recent years has
placed tools for automatic image registration at the treatment console for routine use. These automatic
matching algorithms have been shown to be accurate for matching 3D cone beam computed
tomography (CBCT) scans to planning CT scans, on both the Elekta XVI system [1,2] and Varian OBI
system [3,4]. There are some significant differences in how the two systems reconstruct 3D image
sets, such as the handling of scatter and the sharpness of image reconstruction filters used. They also
use different algorithms in their respective automatic registration software, ie – OBI uses a mutual
information algorithm, whilst XVI uses a Chamfer algorithm for bone matching and a correlation ratio
algorithm for the grey value match.
In a previous study [1], the accuracy of XVI automatic image registration is shown to have a
dependence with imaging dose. This study looks at whether there is a similar relationship in the OBI
system, and the impact of using images from both systems in a single automatic algorithm, in order to
determine any differences in performance between the two systems.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
XVII International Conference on the Use of Computers in Radiation Therapy (ICCR 2013) IOP Publishing
Journal of Physics: Conference Series 489 (2014) 012070 doi:10.1088/1742-6596/489/1/012070

2. Method
A CIRS model 801-P-B Virtually Human Male Pelvis phantom (Computerized Imaging Reference
Systems, Inc., Norfolk, Virginia, USA) was used for all images. Cone beam images were acquired on
both an Elekta Synergy (Elekta AB, , Stockholm, Sweden) and a Varian TrueBeam (Varian Medical
Systems, Palo Alto, CA). For each, the phantom was set up and positioned using the CBCT system
using a reference scan acquired on a GE Lightspeed RT scanner (GE Medical Systems, Waukesha,
WI) with 1.25mm slice thickness. The initial phantom position was corrected in 6 degrees of freedom
(DoF) on the Synergy using a Hexapod couch and 3 DoF on the TrueBeam.
Once positioned to within 0.5mm and 0.5° of the reference scan, a series of CBCT images were
acquired on each system with mAs settings chosen to cover a range both above and below that used
typically in clinical practice. The settings chosen were matched to be as similar as possible between
systems. The exposure was varied from 68 mAs to 1360 mAs with XVI (5 settings), and from 132
mAs to 680 mAs with OBI (4 settings). The effective dose for each scan was calculated using
published in-air dose/mAs factors for XVI and OBI [5].
The OBI images were imported into the XVI software, and then all 9 image sets were analysed
against the initial reference CT scan. For each image set, an initial random translational offset was
applied. Then the correlation ratio algorithm (grey value match) was run to register a mask region in 6
DoF. This mask encompassed the phantom prostate volume with an additional 5mm expansion. The
registration result was subtracted from the initial known offset to provide the residual error of the
registration algorithm. This process was repeated 100 times for each image set to simulate inter-
fraction differences, using random initial offsets sampled from a 3D Gaussian distribution based on
variations, observed clinically, in daily setup of prostate patients [6].
The mean of each image set’s residual error was subtracted as it was assumed to be the systematic
error in phantom positioning, accounting for initial positioning differences. The Target Registration
Error (TRE) was determined as the maximum distance of a point on the surface of a sphere with
residual translation/rotations to the initial sphere position [2]. A sphere of 30 mm radius was used to
approximate the surface of the prostate. This was computed using dual quaternions to combine both
translational and rotational components of the residual error and the distance denoted TRE30 [2].

Figure 1 – Series of phantom CBCT collected for varying exposure. The top row shows XVI images
and the bottom row shows OBI images. Exposure increases going from left to right.

2
XVII International Conference on the Use of Computers in Radiation Therapy (ICCR 2013) IOP Publishing
Journal of Physics: Conference Series 489 (2014) 012070 doi:10.1088/1742-6596/489/1/012070

3. Results
The image quality for the XVI and OBI images both deteriorated with decreasing dose, showing
increased noise. For the lowest dose XVI image artifacts are observed which are likely due to photon
starvation (Figure 1). Note that very low dose images could not be obtained with the OBI system to
compare to the XVI system. Image quality comparisons between the systems are rather subjective due
to the differing sharpness and noise present in each system’s images. However, for images in the
typical range of clinical doses (10-30 mGy), similar image quality was observed based on the ability to
resolve anatomical landmarks in soft tissue such as the rectal wall. 5 mm diameter plugs in the
phantom were only visible in scans of doses more than 23 mGy for both systems.

a)

b)

Figure 2 – a) Plots of residual translational error (dots) and residual rotational error (crosses) as 3D
vector magnitude in the automatic image registration process for each image-set. b) Plots of TRE30.
Each plot shows the results of 100 repeat registrations

3
XVII International Conference on the Use of Computers in Radiation Therapy (ICCR 2013) IOP Publishing
Journal of Physics: Conference Series 489 (2014) 012070 doi:10.1088/1742-6596/489/1/012070

Registration performance with the OBI images was similar to that of XVI with residual translations
<0.5mm (1σ) (Figure 2a). OBI residual rotations were typically 1.2° (1σ) compared to 0.8° for XVI
(Figure 2a). Over the clinical dose range, automatic registration provided similar residual error for
both imaging systems. There was a dose dependence on residual translation and rotation errors for the
XVI images, with both increasing for images with dose below 4 mGy, which correlated with the work
of Sykes et al [2]. A similar dose dependence could not be investigated in the OBI results, because
scans below 4 mGy could not be obtained.
The TRE30 in the clinical range was similar between the two imaging systems (Figure 2b). There
was little difference in the range of TRE30 results across image sets except for XVI scans with dose
less than 4 mGy where values increase. Using a pass/fail tolerance for any given automatic registration
of TRE30 = 3.6mm, failures of the algorithm occurred in 5 - 9% of registrations except for the lowest
dose XVI scan (31%). This indicates the algorithms can provide clinically suitable registrations more
than 90% of the time, and also agrees with other results in the literature.

4. Conclusions
The uncertainty in automatic image registration with both OBI and XVI images was found to be
adequate for clinical use, and similar between the systems, within the normal range of acquisition
settings.

Acknowledgements
The authors would like to thank Adrian Bailey, Radiation Oncology Orange Base Hospital NSW,
Australia for the loan of the phantom used in this study.

References
[1] Sykes J R, Brettle D S, Magee D R and Thwaites D I 2009 Investigation of uncertainties in
image registration of cone beam CT to CT on an image-guided radiotherapy system. Phys.
Med. Biol. 54 7263–83
[2] Sykes J R 2010 Quantification of geometric uncertainties in image guided radiotherapy
(University of Leeds) http://etheses.whiterose.ac.uk/3864/
[3] Fox T, Huntzinger C, Johnstone P, Ogunleye T and Elder E 2006 Performance evaluation of an
automated image registration algorithm using an integrated kilovoltage imaging and
guidance system. J. App. Clin. Med. Phys. 7 97–104
[4] Grams M P, Brown L C, Brinkmann D H, Pafundi D H, Mundy D W, Garces Y I, Park S S,
Olivier K R and De los Santos L E F 2013 Analysis of automatic match results for cone-
beam computed tomography localization of conventionally fractionated lung tumors Prac.
Radiat. Oncol. (Article in Press DOI: 10.1016/j.prro.2013.02.008)
[5] Sykes J et al. 2010 CEP10071: X-ray tomographic image guidance radiotherapy systems
http://nhscep.useconnect.co.uk/CEPProducts/Catalogue.aspx?reportId=450
[6] Smitsmans M H P, Wolthaus J W H, Artignan X, De Bois J, Jaffray D a, Lebesque J V and Van
Herk M 2004 Automatic localization of the prostate for on-line or off-line image-guided
radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 60 623–35