960 / 1096
S944
ESTRO 36
_______________________________________________________________________________________________
planning of AVM’s. This negates the need to perform an
invasive localised DSA in the majority of cases thereby
reducing risks associated with this procedure.
EP-1720 Framework for Statistical Cone-Beam CT
Reconstruction with Prior Monte-Carlo Scatter
Estimation
J. Mason
1
, M. Davies
1
, W. Nailon
2
1
University of Edinburgh, Institute for Digital
Communications, Edinburgh, United Kingdom
2
Oncology Physics Department, Edinburgh Cancer Centre,
Edinburgh, United Kingdom
Purpose or Objective
Scatter from the patient and detector leads to significant
inaccuracies and artefacts in cone-beam computed
(CBCT). Monte-Carlo (MC) methods may allow the scatter
signal to be accurately estimated based on a prior scan,
but this must be matched and calculated for the new cone-
beam measurements, and incorporated appropriately into
a reconstruction method. We investigate a framework for
statistical reconstruction with these prior MC estimates,
under various work-flows.
Material and Methods
The framework consists of statistical iterative
reconstruction with knowledge of a MC scatter estimate
from the planning CT. The estimate can be generated with
the scheme proposed by Xu et al. (PMB 2015) with online
registration and approximate MC of the prior, or by
calculating an accurate scatter off-line and warping to
match the new measurements, which will be faster. We
supplement the reconstruction with regularisation
spatially and between the prior image, hereby utilising the
planning image twice.
Results
The method was applied to data from repeat CT scanning
of a neck cancer patient, with a low-dose repeat CBCT to
form the new measurements. The figure shows
reconstructions from the framework along with
alternative approaches for comparison. The norm of the
error in attenuation coefficient through the region
containing the specimen is: 2.94 for
no scatter estimate
FDK
, which has no scatter correction; 1.74 for
uniform
estimate statistical
, which is statistical reconstruction
with a simple scatter estimate; 2.41 for
proposed estimate
FDK,
with prior scatter estimation; and 0.835 for
proposed
estimate statistical
, which is the proposed framework
combining double regularised statistical reconstruction
with prior scatter estimation. We note that using the
estimation strategy of Xu et al. in our framework yields an
error of 0.792.
Conclusion
We show that the general framework of statistical
reconstruction with prior scatter estimation is accurate
under low-dose acquisitions. The choice of our off-line
scatter estimation or the on-line approximate estimate of
Xu et al. is a trade-off in computational time and added
accuracy, but either perform well. We suggest this
quantitatively accurate method should be suitable for
adaptive radiotherapy, and we are planning further
testing with more applicable data.
EP-1721 A new calibration method of an Elekta XVI
(R.5.0.2) system able to achieve superior image quality.
D. Oborska-Kumaszynska
1
1
Royal Wolverhampton NHS Trust, MPCE Department,
Wolverhampton, United Kingdom
Purpose or Objective
During acceptance testing of an Elekta XVI system, it is
standard practice to only test system performance using
SFOV. The focus of this work was to develop and introduce
an extended customer acceptance test (AT) procedure as
a foundation for introducing a new calibration method of
an Elekta XVI (R.5.0.2) system. With optimal image quality
achieved for all FOVs during AT, more appropriate
optimisation of clinical XVI protocols can then be
performed.
Material and Methods
Following a significant service of the system (X-ray tube
replacement and calibration of the detector SDD), it was
first calibrated in line with the standard manufacturer
procedure. Extended ATs were then performed for all
FOVs. A CATPHAN 600 phantom was used to assess spatial
resolution, uniformity, contrast and geometry. The
parameters used by the reconstruction algorithm can
change depending on which AT you are performing (e.g.
spatial resolution, uniformity or contrast). It was
therefore necessary to assess all combinations of FOVs and
each test (e.g. MFOV with spatial resolution, LFOV with
uniformity…).
Results
The performed ATs showed unacceptable image quality
for MFOV and LFOV. Spatial resolution tests revealed
images with significant artefacts (Fig.1). The contrast
results were worse than 3%. The uniformity test results
were worse than 5%. The calibration procedure of the
system was repeated a few times but results were still
unacceptable. Investigations pointed to a few reasons:
very rigid reconstruction algorithm (FDK) with regards to
geometry, an exactly defined geometrical relation
between X-ray source (focal spot position) and the
coordinate system of pixels of the detector for each FOV,
and correction of flexmap offset required for each FOV.
This led to additional steps in the XVI system calibration:
X-ray beam axis alignment perpendicular to the imaging
panel surface and to “the centre” of that panel for each
FOV; calibration of lateral panel position and re-setting of
pots readings; re-calibration of flexmaps, badpixel maps
and gains of the imaging panel. The spatial resolution test
results showed for SFOV=13 lp/cm, MFOV=12.5 lp/cm,
LFOV=13 lp/cm and clear visibility of the spatial pattern.
The results for contrast were 1.85%; 2.39%; 2.91%,
respectively and for uniformity were 0.52%; 2.00%; 3.64%,
respectively.
Conclusion
Introducing extended AT for the Elekta XVI (R.5.0.2)
system provides full evaluation of the system before
introducing it in to clinical practice and ensures that
image quality is acceptable for all FOVs. The new
calibration procedure was able to realise the full and
proper reconstruction of image information and resulted
in a significant improvement of image quality for all FOVs,
as assessed by the above parameters.