829 / 1082
S813
ESTRO 36 2017
_______________________________________________________________________________________________
C. Cases
1
, A. Latorre-Musoll
1
, P. Carrasco
1
, N. Jornet
1
, T.
Eudaldo
1
, A. Ruiz-Martínez
1
, M. Lizondo
1
, P. Delgado-
Tapia
1
, M. Ribas
1
1
Hospital de la Santa Creu i Sant Pau, Radiofisica i
Radioprotecció, Barcelona, Spain
Purpose or Objective
For SBRT lung treatments accurate 4D dose calculations,
accounting for heterogeneities and respiratory motion,
are crucial to determine an optimal ITV beyond a purely
geometric ITV. We propose a model to predict an optimal
ITV from a single figure computed from the Probability
Density Function (PDF) of the breathing waveform.
Material and Methods
We used the QUASAR Respiratory Motion Phantom (Modus
Medical Devices) with a cylindrical mobile wood insert as
lung substitute and an inner 30mm diameter sphere as
tumour substitute (GTV). We acquired 21 independent
scans (CT
z
) by axially shifting the mobile insert from z = –
10 to z = 10mm in 2mm steps. We generated 6 ITV: from
ITV
0mm
(static case, equal to the GTV of CT
0mm
) to ITV
10mm
(overlap of all GTV from CT
–10mm
to CT
10mm
). We planned a
SBRT treatment collimating to each ITV (from PLAN
0mm
to
PLAN
10mm
) in Varian Eclipse (AAA v13.5) using a 6MV non-
coplanar 3DCRT technique. Due to the
in silico
nature of
the study we added no extra margins to the ITV.
We considered 3 breathing patterns: sinusoidal (provided
by QUASAR software), free and trained (obtained by the
Varian RPM from real patients). We rescaled every
waveform to amplitudes from 2 to 10mm in 2mm steps.
We built the actual 4D dose distribution for every PLAN
z
considering
all
combinations
of
breathing
patterns/amplitudes. We first copied the original
treatment (planned at CT
0mm
) to the remaining CT
z
scans
and recalculated them by using fixed MU. Then we copied
the resulting dose matrices back to the CT
0mm
scan and we
shifted them axially by -z. Later, we summed the dose
matrices using weights derived from the PDF of the
underlying waveforms.
We defined a Quality Index which balances GTV coverage
and healthy tissue-sparing as QI = (V
100%,GTV
)
2
/V
PD
, where
V
100%,GTV
and V
PD
stand for the percentage of GTV covered
with the Prescribed Dose (PD) and the volume of the PD
isodose respectively. The optimal plan, PLAN
opt
, was the
highest QI scoring plan for each breathing
pattern/amplitude. Finally, we assessed the PDF’s
measure of central tendency that best predicts PLAN
opt
irrespective of the breathing pattern/amplitude.
Results
Figure 1a shows the QI for the sinusoidal movement. Every
breathing pattern’s maximum QI scores project an optimal
curve to the x-y plane (figure 1b). For any breathing
pattern/amplitude, PLAN
opt
was found to be conformed to
an optimal ITV smaller than the purely geometric ITV.
We found that the integral of the PDF between ±x, i.e.,
the time fraction on which the GTV is on the central part
of the respiratory excursion, was the best predictor of
PLAN
opt.
Irrespective of the breathing pattern/amplitude
all PDF’s integrals collapsed to a unique curve for x = 3mm
(figure 2).
Conclusion
Based on 4D dose calculations, we propose a QI to reduce
the ITV while maintaining the GTV coverage for SBRT lung
treatments. We provide a model to predict the optimal ITV
from the integral of the PDF of the breathing waveform.
Partially financed by FIS PI15-00788 grant.
EP-1533 Modulation complexity assessment in VMAT
plans from different treatment planning systems.
P. Winkler
1
, A. Trausnitz
2
, J. Schroettner
2
, A. Apfolter
1
,
K. Kapp
1
1
Medical University of Graz, Department of Therapeutic
Radiology and Oncology, Graz, Austria
2
University of Technology, Institute of Health Care
Engineering, Graz, Austria
Purpose or Objective
Modulation complexity (MC) in Linac-based VMAT plans
might influence the accuracy of dose calculation and
dose delivery as well as the precision of dose delivery.
However, MC is not a single-parametric property, but
rather consists of several different influencing factors,
e.g. average leave speed (ALS), leave sequence
variability (LSV), mean field aperture area (FAA),
aperture area variability (AAV), gantry acceleration
(gantry-speed variation, GSV) and dose rate variability
(DRV), which might predominantly be correlated with
uncertainties in either dose calculation or delivery. In
our clinical treatment plans we observed, that different
TPS accomplish strong modulation in a noticeably
different way, forcing either ALS, LSV and AAV whilst
retaining moderate GSV and DRV, or vice versa. The aim
of this study is to present several distinct modulation
complexity indices (MCI), describing the different aspects
of modulation complexity in VMAT plans, and to assess
the characteristic ranges of these MCI for different TPS.
Material and Methods
We established six MCI, parameterising the magnitude of
ALS, LSV, FAA, AAV, GSV and DRV in a VMAT-arc, and
implemented their calculation in our automated plan-QA
software tool (in-house development). The MCI for 200
randomly selected clinical beams were calculated for the
TPS Eclipse (Varian) and Pinnacle (Philips),
respectively. Additionally 20 phantom plans (37 arcs)
with increasing modulation complexity were generated
for both of the two TPS using best possible matching of
optimization criteria, and were subsequently analysed.
Results
In the phantom plans, the Pinnacle-optimized arcs showed
significantly higher average leaf speed compared to
Eclipse-optimized arcs (10.9 and 6.7 mm/sec,
respectively). Aperture – opening was 40% larger for the
Eclipse-arcs. Consequently, the number of monitor units
was smaller in the Eclipse plans (-32%). Whereas the
differences for LSV and AAV were rather small (figure 1),
DRV and GSV differed significantly, revealing a more
pronounced modulation in the Pinnacle plans as far as dose
rate and gantry acceleration are concerned. Findings for
retrospectively analysed clinical plans and (non-biased)
phantom plans were
similar.
Figure 1: Modulation complexity scores (LSV: leave
sequence variability, AAV: aperture area variability, GSV:
gantry-speed variation, DRV: dose rate variability) for 200
VMAT plans, calculated with Eclipse and Pinnacle TPS,
respectively. Box-plots showing median, first and third
quartile and range.
Conclusion
Modulation complexity in VMAT plans has a potential
impact on dose-calculation and –delivery accuracy. We
found considerable differences for two different TPS in
multi-parametric assessment of MC features, indicating
the diverging algorithms of the different optimizers. A