147 / 1082
S134
ESTRO 36 2017
_______________________________________________________________________________________________
plan for delivery. Gamma pass-rate and mean Gamma
(
γ
mean
) were evaluated for 380 daily adaptive fractions for
over 5 months. The independent MC dose engine was
verified for site specific class solutions (Pancreas, Liver,
Lung and Prostate) using film dosimetry before clinical
implementation.
Results
Online patient specific QA for daily plan adaptation,
including 3D MC dose calculation, gamma evaluation and
automatic check of plan parameters was performed on
average in 1 min 45 sec. A typical result is shown in Figure
1. Average gamma pass-rate and γ
mean
over 380 fractions
was 99.5 (95% CI [97.1, 100]) and 0.38 (95% CI [0.33,0.44]),
respectively. All daily adapted plans passed criteria for
approval.
Verification of MC dose engine for site specific class-
solutions exhibited an excellent agreement with film
dosimetry, with average gamma pass-rate and γ
mean
of
100% of 0.14, respectively.
Conclusion
A patient specific QA procedure for online adaptive MR-
guided radiotherapy was successfully implemented at our
institution. This procedure, which takes less than 2
minutes, include online independent 3D MC dose
calculation after daily plan adaption and automatic check
of plan parameters while the patient is in treatment
position. A very good agreement with dose distribution
from the TPS was found for all adaptive fractions.
Proffered Papers: Variabilities in volume definition
OC-0263 Single vs. multi-atlas auto-segmentation for
prostate RT: Comparison of two commercial systems
A. Gulyban
1
, P. Berkovic
1
, F. Lakosi
2
, J. Hermesse
1
, P.A.
Coucke
1
, V. Baart
1
, D. Dechambre
1
1
Liege University hospital, Department of Radiation
Oncology, Liege, Belgium
2
Health Science Center- University of Kaposvar,
Radiation Oncology, Kaposvar, Hungary
Purpose or Objective
Using atlas-based auto-segmentation during treatment
planning has the potential to reduce the workload of the
staff while improving delineation consistency. Our aim was
to evaluate two commercial systems for prostate
treatment planning: evaluating volumetric accuracy 1)
while completing the atlas (learning curve), 2) using the
full atlas (performance) and 3) determining dose volume
histogram parameter (DVH) variations between of the
auto-generated and reference contours.
Material and Methods
Forty random prostate patient cases were selected for this
study. Each dataset consisted of a CT, a structure set
(including prostate, rectum, bladder, anal canal and
penile bulb) and the original dose matrix. Two systems
were used, the single-atlas based Raystation ('RS”,version
5.0.2, Stockholm, Sweden) and the multi-atlas based RTx
('MIR”, version 1.6.3, Mirada Medical, Oxford, UK). The 1-
5
th
case was used as base atlas. The learning phase was
completed in an incremental way, where each new auto-
contours generated using an atlas consisting of all former
cases (6
th
case used the atlas of 1-5
th
, 7
th
the 1-6
th
etc.)
until the 20
th
case. Performance was evaluated using the
complete (1-20
th
) atlas on another 20 cases (21-40
th
).
Analysis included the Dice Similarity Coefficient (DSC),
Jaccard index (JI), commonly contoured volumes (CCV),
volumetric ratios (VR) and 95% of the Hausdorff distance
(HD95%). Furthermore using the dose matrix, DVHs were
generated for all volumes and the differences of relevant
organs at risk specific parameters were compared. Mean
values and standard deviations (SD) were used for the
descriptive statistics and paired t-test to compare the MIR
vs. RS performance, while the Root mean squares (RMS)
were compared for the dosimetrical differences.
Results
For volumetric comparison (DSC, JI, and CCV, VR, HD95%)
21 vs. 14 out of 24 parameters improved from the learning
to the performance stages for MIR vs. RS respectively
(table 1). For rectum, MIR underestimated the volume (VR
for MIR 0.75 vs. RS 0.79, p<0.001), while for all other
parameters outperformed RS (p<0.001). Results for
bladder and prostate showed superior performance of MIR
with the exception of bladder CCV, which was not
significantly better compared to RS. Anal canal and penile
bulb showed poor agreement (for both systems). RS was
more accurate to estimate the anal canal volume; similar
results were obtained for penile bulb’s CCV, while all
other parameters were significantly better with MIR. RMS
values of MIR vs. RS resulting in 10.1 vs. 11.9 cc of V50Gy
and 4.7 vs. 8.0 Gy of mean dose for rectum, for bladder
V55Gy 22.4 vs. 27.3 cc and mean dose 8.6 vs. 8.6 Gy, while
for anal canal V20Gy of 6.9 vs. 6.3 cc and mean dose of
9.1
vs.
12.5
Gy
respectively.
Conclusion