463 / 1082
S448
ESTRO 36 2017
_______________________________________________________________________________________________
Sweden) for both initial treatment planning and online
plan adaptation. Next to the presence of a magnetic field,
also several MRL-specific beam and collimator properties
need to be taken into account that could influence plan
quality. Our aim was to investigate the influence of MRL-
specific characteristics on plan quality for rectum cancer
and benchmark MRL plans against current clinical
practice.
Material and Methods
Eight rectum cancer patients treated on a conventional
CBCT-based linac (25 x 2.0 Gy) were included in this
retrospective study. For each patient, the clinically
acquired planning CT, delineated structures and
treatment plan generated with Pinnacle
3
(dual-arc VMAT,
10MV, collimator 20°, SAD: 100.0 cm) were available. The
same CT and structure set were used to create two MRL
treatment plans with Monaco: one plan with (MRL
+
) and
one plan without (MRL
–
) the presence of a 1.5 T magnetic
field. Both MRL plans were created using a 7-beam IMRT
technique incorporating MRL-specific properties (7MV,
collimator fixed at 90°, FFF, SAD: 143.5 cm). Plan
optimization was based on a class solution and objective
values were individually optimized. Also, a quasi MRL plan
was generated with Pinnacle
3
using a 7-beam IMRT
technique and comparable MRL properties (6MV,
collimator 90°, FFF, SAD: 143.5 cm). After rescaling (PTV
V
95%
= 99.2%), plans were accepted when the clinical
acceptance criterion was fulfilled (PTV D
1%
< 107%).
Quality differences between MRL
+
, MRL
–
and quasi MRL
plans were assessed by calculating PTV D
mean
, PTV D
1%
,
bowel D
mean
and bladder D
mean
. Also, D
mean
and D
1%
to the
patient excluding PTV
2cm
(i.e. PTV + 2.0 cm) were
determined. All MRL plans were benchmarked against the
clinically delivered treatment plans and tested for
significance (Wilcoxon signed-rank test).
Results
All MRL plans were clinical acceptable after rescaling.
Figure 1 shows an example of dose distributions for the
MRL plans and the clinical plan of one patient. The 7-beam
IMRT technique used for all MRL plans resulted in a minor
decrease in plan homogeneity, indicated by an increased
PTV D
mean
(Table 1). Also, all MRL plans showed a
significant increase in D
mean
for the bladder, bowel and
body compared to clinical practice. However, the clinical
relevance of these differences is expected to be limited.
Given the similar quality of MRL
–
and quasi MRL plans,
differences between MRL
+
plans and clinical practice are
mainly induced by the MRL-specific properties. The small
difference between MRL
+
and MRL
–
plans indicated limited
influence of the magnetic field on plan quality.
Conclusion
This study demonstrates the ability of creating high-
quality MRL treatment plans for rectum cancer. Given the
differences in machine characteristics, some plan quality
differences were found between MRL treatment plans and
current clinical practice. These results support a well-
prepared clinical introduction of the MRL.
PO-0839 Personalized VMAT optimization for
pancreatic SBRT
I. Mihaylov
1
, L. Portelance
1
1
University of Miami, Radiation Oncology, Miami, USA
Purpose or Objective
Inverse IMRT planning is a very labor intensive, trial-and-
error process, aiming to find a middle ground between the
conflicting objectives of adequate tumor coverage and
sparing nearby healthy tissues. Even if a plan is clinically
acceptable, that plan is unlikely to be the best solution,
where the healthy tissue is spared as much as possible. To
a large extent the optimization process is user and
treatment planning system specific, where more
experienced users generate better quality radiotherapy
plans. This work introduces a fully automated inverse
optimization approach and its application to pancreatic
SBRT.
Material and Methods
Ten cases, treated breath-hold, were retrospectively
studied. The outlined anatomical structures consisted of a
PTV, and OARs including duodenum, stomach, bowel,
spinal cord, liver, and kidneys. In each case the
prescription was set to 35 Gy (to 95% of the PTV) in 5
fractions. The treatment plans were created by
experienced dosimetrists, following national and
international clinical protocols. Those treatment plans
were generated for VMAT delivery. For each case an
additional plan was generated with the newly proposed
automated inverse optimization. This optimization is
based on unattended step-wise reduction of DVHs, where
several DVH objectives were specified for each OAR. The
automated plans utilized the same number of arcs, with
the same parameters as the treatment plans. The
treatment and the automated plans (Treatment and Auto
hereafter) were compared on commonly used clinical
dosimetric parameters. Those parameters included D
PTV
95%
(dose to 95% of the PTV), D
Duodenum
1%
, D
Bowel
1%
, D
Stomach
1%
,
D
Cord
1%
, D
Liver
mean
, D
rt_kidney
mean
, and D
lt_kidney
mean
. The doses to
1% of the volumes of duodenum, bowel, stomach, and
spinal cord were used as surrogates for maximum doses.
The prescriptions for the Auto plans matched the
prescriptions of the Treatment plans.
Results
The first row in the table below summarizes the average
values of the tallied quantities (over the ten patients) as
derived from the treatment plans. The second row
outlines the average differences (in per-cent) between the
dosimetric endpoints as well as the range of the
differences between the Treatment and the Auto-
optimized plans. The negative differences indicate that
the Auto plans result in lower absolute doses and vice-