464 / 1096
S449
ESTRO 36
_______________________________________________________________________________________________
Results
Both IMPT or target tailoring by excluding the proximal
uterus resulted in significant reductions of V
15Gy
, V
30Gy
,
V
45Gy
and D
mean
for bladder and small bowel. Compared to
conventional volumes, target tailoring by excluding the
non-invaded uterus resulted in an average reduction of the
primary ITV and PTV of 37% and 8%, respectively. IMPT
would have reduced the estimated NTCP for small bowel
toxicity (≥grade 2) from 25% to 18%, and would be
additionally reduced to 9% when IMPT were combined with
MRI-based target tailoring. Major NTCP reductions of >10%
were predicted in four patients (36%) by IMPT, and in six
patients (55%) when IMPT were combined with MRI-based
target tailoring. Patients benefitted most (NTCP reduction
>10%) from one of the investigated approaches if the V
45Gy
for bowel cavity was >275 cm
3
during standard IGART
alone; a similar reduction in NTCP from the combined
approached would have been obtained in patients with a
V
45Gy
for bowel cavity >200 cm
3
.
Conclusion
In patients with cervical cancer, both 1) proton therapy
and 2) target tailoring by excluding the radiologically
uninvolved part of the uterine corpus led to a significant
dose reduction to surrounding OARs, which separately
would already yield a clinically important decrease in
small bowel toxicity, which is cumulative if both
approaches would be combined.
Reference
[1] de Boer P, Bleeker MCG, Spijkerboer AM, et al. Eur J
Radiol Open. 2015;2:111–7.
PO-0834 Automated planning to reduce integral dose
in robotic radiosurgery for benign tumors
L. Rossi
1
, A. Méndez Romero
1
, M. Milder
1
, E. De Klerck
1
,
S. Breedveld
1
, B. Heijmen
1
1
Erasmus Medical Center, Radiation Oncology,
Rotterdam, The Netherlands
Purpose or Objective
Highly conformal dose distributions and minimizing
integral dose are essential in radiosurgery of benign
vestibular schwannoma (VS) tumors to avoid long term
side effects. This includes avoidance of secondary tumor
induction in these long surviving patients. High delivery
accuracy can be obtained with the robotic CyberKnife (CK,
Accuray Inc, Sunnyvale, USA) due to real time image-
guided tracking, allowing small PTV margins. However,
optimal plan quality may be hampered by the current
trial-and-error planning approach, as it strongly depends
on the planner’s experience and available planning time.
We have developed a system for fully automated
CyberKnife treatment planning. In this study, we have
used this system to automatically generate plans for
vestibular schwannoma patients (AUTOplan) and we have
compared them with plans that were manually generated
in clinical routine (MANplan), both with the IRIS
collimator.
Material and Methods
Both MANplans and AUTOplans were genereated with the
Multiplan TPS (Accuray Inc). For AUTOplanning, a fully
automatic pre-optimization was performed with our in-
house multicriterial optimizer to generate input
parameters for automated plan generation in Multiplan,
including patient-specific parameters to maximally
control integral dose. Plan comparisons were made for 15
patients. Both for automatic and manual planning, the
goal was to deliver a single fraction of 12 Gy, with planning
priorities PTV V100% ≥ 98%, Brainstem Dmax < 12.5 Gy,
while at the same time keeping the integral dose as small
as possible. For un-biased plan quality comparisons,
AUTOplans were generated such that the resulting CK
treatment time was similar to that for the corresponding
MANplan.
Results
AUTOplans were comparable to manual MANplans in terms
of PTV coverage (AUTO: 99.4 ± 0.5 %, MAN = 99.1 ± 0.5 %,
p=0.1) and treatment time (AUTO = 39.5 ± 4.7 min, MAN =
38.9 ± 5.9 min, p=0.3). On average, the brainstem D2%,
D1cc and Dmean were very similar, i.e. 9.5 vs. 9.6, 8.6 vs.
8.5, and 2.0 vs. 2.2 Gy for the AUTO- and MANplans,
respectively (p>0.2). Patient volumes receiving more than
1, 2, 4, and 6 Gy were highly reduced in the AUTOplans
for the majority of patients, as visible in figure 1 (upper),
with average reductions of 26.0% (SD= 15.4%, p < 0.001 ),
14.7% (SD=10.5%, p < 0.001), 9.8% (SD= 10.3%, p = 0.002 ),
and 6.3% (SD=10.4%, p = 0.010). Conformality was also
better in the AUTOplans, and spiky dose leakage away
from the target was less frequent and severe, as visible in
figure 2. The D2% in ring structures at 1, 2, and 3 cm
distance from the PTV were 3.6, 1.9, and 1.3 Gy in
AUTOplans vs. 4.7, 2.4, and 1.6 Gy in the MANplans (p<
0.001). For almost all patients, ring structures’ D2% were
lowest in the AUTOplan (see figure 1, lower).
Conclusion
With automated Cyberknife planning, highly patient-
specific parameters for optimal plan generation in
Multiplan are automatically established, resulting in
substantial reductions in integral dose in treatment of
benign vestibular schwannoma tumors, without degrading
PTV dose delivery, increasing OAR doses, or enlarging
treatment time.