835 / 1096
S819
ESTRO 36
_______________________________________________________________________________________________
Conclusion
Plan generation for breast with locoregional nodes was
successfully automated using the Eclipse scripting API to
create a workflow that integrates the RP knowledge-based
planning system, and a combination of different
techniques: open fields, slip zone, RA. Automated
generation of treatment plans is anticipated to lead to
more consistent and efficient planning. It may also
facilitate the transfer of complex treatment planning
techniques between centers.
EP-1525 Automatic treatment plan generation for
Prostate Cancer
S. Agergaard
1
, C.R. Hansen
1,2
, L. Dysager
3
, A. Bertelsen
1
,
H.R. Jensen
1
, S. Hansen
2,3
, C. Brink
1,2
1
Odense University Hospital, Laboratory of Radiation
Physics, Odense, Denmark
2
University of Southern Denmark, Faculty of Health
Sciences, Odense, Denmark
3
Odense University Hospital, Department of Oncology,
Odense, Denmark
Purpose or Objective
Automatic treatment planning is of high interest, since the
optimization process is highly complex and the current
plan quality is dependent on the treatment planner. In a
clinical setting where time for treatment planning is
sparse, automatic treatment plan generation would be
desirable. This study evaluates automatic treatment
planning for high risk prostate cancer in comparison to a
current clinical plan quality.
Material and Methods
All patients (#42) treated for high risk prostate cancer
during 2015 at our clinic were replanned using the
Autoplan module in Pinnacle® (ver. 9.10). Similar to the
manual plan (MA) the autoplan (AP) was generated for an
Elekta® Synergy linac, consisting of one full VMAT arc and
using 18 MV photons. All APs were calculated by the same
medical physicist. There was no comparison of the MA and
AP in the plan generation process. Using a template model
it took on average 90 sec to start autoplanning, which took
approximately 1 hour to complete optimization. Hereafter
it took on average 173 sec (range 45 to 550) of active
planning for one or two post-optimizations with 15
iterations per run to fine-tune the plan to meet the
acceptance criteria.
The plan quality was evaluated by comparing DVHs, dose
metrics, delivery time and dose accuracy when delivered
on an ArcCheck phantom.
For each patient the MA and AP were blindly evaluated
side-by-side by a radiation oncologist, who concluded
which plan was better, and if the differences were
predicted to be clinically relevant.
All differences were tested for statistical significance with
a Wilcoxon signed rank test (p<0.05).
Results
The DVHs show small but significant differences in the
doses to both CTV and PTV. The APs spared all OARs
significantly. For the rectum the average of the mean
doses is reduced from 42.6 Gy to 31.8 Gy. The reduction
in rectal dose is significant between 1 Gy and 73 Gy (figure
1). Table 1 shows the results for targets as well as OARs,
their standard deviations (std) and the corresponding p-
values.