794 / 1096
S778
ESTRO 36
_______________________________________________________________________________________________
Purpose or Objective
The overall objective is to develop a 3D complexity metric
for VMAT treatments. The complexity scores will be
presented as a distribution in a 3D volume and correlate
to the fraction of penumbra dose. Regions lacking charged
particle equilibrium that might cause dose calculation
errors and regions sensitive to multileaf collimator (MLC)
positioning errors are located in the penumbra of the MLC
opening. The hypothesis is that an increased amount of
dose in a voxel that originates from a penumbra region will
correlate to the probability of increased difference
between planned and delivered dose in that voxel. In this
pilot study, 2D distributions are analyzed to validate the
correlation to differences between calculated and
measured dose.
Material and Methods
A C# software with dynamically linked MatLab®
(Mathworks, Natick, MA) libraries was developed. The
input to the software is the DICOM-file of the treatment
plan from where the MLC positions are collected, i.e. the
appearance of the beams eye view (BEV) plane.
1.
The pixels of the BEV plane (pixel size 0.25 mm)
is structured binary in open beam (1) or blocked
beam (0).
2.
The pixels of the BEV plane is also structured
binary in field edge (1) or no edge (0).
3.
The binary BEV from step 1 is convolved with a
Gaussian function normalized to 1. This will
weight the complexity score higher in regions
with higher dose and lower in the low dose
region. This is called a pseudo dose (PD)
distribution.
4.
The binary BEV from step 2 is convolved with a
box function (1 inside box, 0 outside). This will
define a region, with the width of the box, as
the region of interest where the complexity
metric will have a score ≠ 0.
5.
The convolution from step 3 is multiplied with
the convolution from step 4. This gives a 2D
distribution of complexity scores.
The 2D distributions of calculated complexity scores for
different box widths and Gaussian sigmas for 30 MLC
openings were compared to the 2D distributions of
difference (absolute values) between calculated and film
measured dose at 10 cm depth in water for the same
openings.
Results
The correlation between the ratios “mean complexity
score/mean value of PD distribution” and “mean absolute
difference between calculated and measured
dose/calculated mean dose” for 30 MLC openings is shown
in figure 1. The sigma of the Gaussian and the box width
had negligible influence on the correlation. However,
those parameters will have influence when the fraction of
penumbra dose is evaluated for each pixel separately.
They can be chosen to match the dose gradient and width
of the region of relevant dose differences at a specific
depth, see example of a 2D visual comparison between
complexity scores and differences between calculated and
measured dose at 10 cm depth in figure 2.
Conclusion
2D distributions of complexity scores were successfully
calculated and comparisons to 2D distributions of
differences between calculated and measured dose show
conformities that are promising for further development
of calculations in a 3D volume.
EP-1458 3D dose reconstruction on CBCT for daily
monitoring of delivered patient dose
K. Eilertsen
1
, F.C. Vidaurre
2
, Y. Pylypchenko
3
1
Eilertsen Karsten, Medical Physics, Lommedalen,
Norway
2
Oslo University Hospital, Medical Physics, OSLO, Norway
3
Oslo University Hospital, Medical Phjysics, OSLO,
Norway
Purpose or Objective
The ability to reconstruct the delivered 3D dose
distribution using the CBCT acquired on every fraction,
can help to verify that the dose to both target as well as
organs at risk comply with the treatment intentions
throughout the treatment course. The objective of this
work has been to study the dosimetric accuracy and
feasibility of daily dose monitoring using a novel system
for 3D dose reconstruction onto kV CBCT from electronic
portal images and machine log data acquired during
treatment execution.
Material and Methods
The tested dose reconstruction engine is based on a
collapsed-cone convolution algorithm and is an integrated
part of the PerFRACTION3D system (SunNuclear). The
method uses a forward projection of MLC leaf position
measurements from the EPID, as well as monitor chamber
dose rate and output data derived from the associated
machine log files. Two different approaches were taken to
test the concept: First, kV CBCT of the ArcCHECK
measurement array (SunNuclear) was acquired on Varian
TrueBeam and Elekta Synergy linacs. Then a number of
different patient plans and were delivered to the detector
array. The generated EPIs and log files were imported to
PerFRACTION3D, and the dose distributions reconstructed