ESTRO 36 Abstract Book

449 / 1082

S434 ESTRO 36 2017
_______________________________________________________________________________________________
bladder and small bowel are merged to a structure that is
used as the single OAR. Next, a density override of 0.5
g/cm3 is performed on any air pockets in the PTV that are
identified using a density threshold. A dual arc VMAT plan
is set up and the dose distribution is optimized using the
Pinnacle3 Auto-Planner. After the generation of the Auto-
Plan, which takes about 45 minutes, it is presented to the
dosimetrist for approval.
The Pinnacle 3 Auto-Planner creates plans based on a set of
dose optimization goals and a number of advanced settings
called the “treatment technique”, which allows (indirect)
control over the resulting plan. The main challenge is to
develop a single treatment technique that leads to
optimal plans, which meet our precise and high clinical
demands, for a large patient population.
After having optimized the treatment technique using a
test set of 30 patients, we evaluated the Auto-Plans by
performing a blind test where 4 physicians and 4 planning
dosimetrists were asked to compare manual clinical plans
with Auto-Plans for 10 new patients.
Results
The optimized treatment technique is shown in Table 1.
On average, the mean dose to the small bowel + bladder
is 2.5 Gy lower for the Auto-Plans compared with manual
plans, at the expense of having a slightly increased dose
in the lateral direction. An example of a manual plan and
an Auto-Plan is shown in Figure 1. The result of the blind
test was a unanimous preference for the Auto-Plans (20-
0), based on a better PTV coverage and a lower OAR dose.
The slightly higher lateral dose was considered
acceptable.
Conclusion
We have successfully developed automatic rectum VMAT
treatment planning using our automation framework FAST
in combination with the Pinnacle 3 Auto-Planner. The Auto-
Plans systematically differ from the manual clinical plans
(with an average OAR mean dose reduction of 2.5 Gy) and
are unanimously preferred by physicians and dosimetrists.
This clearly demonstrates how the implementation of an
Auto-Planner system, combined with the accompanying
reconsideration of plan style and clinical trade-offs, can
lead to improved treatment plans. As a result, automatic
rectum VMAT planning has been introduced in our clinic as
of July 2016.
PO-0819 Robustness evaluation of single- and multifield
optimized proton plans for unilateral head and neck
M. Cubillos Mesías1, E.G.C. Troost1,2,3,4,5, S. Appold2, M.
Krause 1,2,3,4,5, C. Richter1,2,3,4, K. Stützer1,4, M. Baumann1
1 OncoRay – National Center for Radiation Research in
Oncology- Medical Faculty and University Hospital Carl
Gustav Carus- Technische Universität Dresden-
Helmholtz-Zentrum Dresden – Rossendorf, Dresden,
Germany
2University Hospital Carl Gustav Carus- Technische
Universität Dresden, Department of Radiation Oncology,
Dresden, Germany
3German Cancer Consortium DKTK, partner site Dresden,
Dresden, Germany
4Helmholtz-Zentrum Dresden – Rossendorf, Institute of
Radiooncology, Dresden, Germany
5National Center for Tumor Diseases, partner site
Dresden, Dresden, Germany
Purpose or Objective
To compare 4 different proton pencil beam scanning (PBS)
treatment approaches for unilateral head and neck cancer
(HNC) targets in terms of robustness, including anatomical
changes during the treatment course.
Material and Methods
Eight patients with unilateral HNC treated with double
scattered proton therapy were selected. Each patient
dataset consists in a planning CT and several control CTs
acquired by an in-room CT scanner during the treatment
course. Four different proton PBS plans with simultaneous
integrated boost and dose prescriptions of 50.3 Gy(RBE) to
the low-risk CTV and 68 Gy(RBE) to the high-risk CTV in 34
fractions were calculated: conventional PTV-based single-
field (SFO) and multifield optimization (MFO), and
robustly optimized SFO and MFO plans on CTV level,
considering ±3 mm and ±3.5% of setup and range
uncertainty,respectively.
The treatment plans were recalculated on the registered
control CTs and the cumulative doses calculated and
compared with the nominal plan.
For robustness evaluation, perturbed doses using a
probabilistic scenario-wise approach obtaining random
setup shifts through Gaussian sampling, and range
uncertainties of 0, +3,5% and -3,5% were calculated, using
planning and control CTs, considering both anatomic
changes and uncertainties. Cumulative doses from 30
different perturbed treatment courses were generated for
each plan.
Results
The target coverage for the four nominal plans was
similar, fulfilling the clinical specification of D98≥95% of
the prescribed dose (range 96.9-100.5% for low-risk CTV,
97.4-100.8% for high-risk CTV), being slightly lower on the
robust optimized plans. The doses to the organs at risk
were similar for all plans; however, for the ipsilateral
parotid, higher median doses up to 5 Gy were found on the
SFO approaches (Table 1), whereas the contralateral
parotid is completely spared. The target coverage
throughout the treatment course with slightly changing
anatomy remains in general constant.