484 / 1082
S469
ESTRO 36 2017
_______________________________________________________________________________________________
Results
From the simulation of patient weight loss (thickness
reduction), Watchdog has less sensitivity to small patient
thickness reduction. From figure 1 left, it can imply that
dropping by 25% c pass-rate refers to 10% patient
thickness reduction or approximately 1.5 cm shrinkage. In
clinical case validation, Watchdog was able to detect the
significant patient anatomical changes that lead to the
decision to replan all four HN IMRT patients (see figure 1
right). Based on this study, we found that Watchdog
system can detect the clinically significant anatomical
change in HN IMRT based on 1) at least 3 out of 7 fields of
the fraction are below the SPC-based threshold, 2) the
lowest c pass-rate is less than 30%, and 3) a 25% c pass-
rate drop equates to approximately a 1.5 cm (-10.0%)
patient thickness reduction.
Conclusion
The Watchdog dosimetic response to HN patient
anatomical changes has been evaluated based on the
simulation of patient thickness reduction/weight loss and
clinical cases of HN IMRT replan. Using the SPC-based
threshold, Watchdog is able to detect clinically significant
anatomical changes in HN IMRT treatment.
PO-0869 A population-based estimate of proton beam
specific range uncertainties in the thorax
Y.Z. Szeto
1
, M.G. Witte
1
, M. Van Herk
2
, J. Sonke
1
1
Netherlands Cancer Institute Antoni van Leeuwenhoek
Hospital, Radiotherapy department, Amsterdam, The
Netherlands
2
Institute of Cancer Sciences- University of Manchester,
Molecular and Clinical Cancer Sciences, Manchester,
United Kingdom
Purpose or Objective
Proton therapy has great potential for locally advanced
lung cancer patients because of considerable reduction of
intermediate and low dose to the healthy tissues.
However, due to their finite beam range, proton dose
distributions are more susceptible to anatomical
variations. The purpose of this study was to derive a
population-based map of beam specific range
uncertainties due to anatomical variations.
Material and Methods
The planning CT (pCT) of 100 NSCLC patients treated
between 2010 and 2013 with (chemo-)radiotherapy were
included. To simulate realistic anatomical variations, we
used a previously developed statistical model, based on
principal component analysis for systematic variations in
the thorax. This model generates deformation vector
fields that deform the planning CT to induce systematic
differences between the anatomy of planning and
delivery. For each patient, we synthesized 1000 CTs (sCT)
representing plausible variations in treatment anatomy.
Subsequently, the water-equivalent path length
differences (∆R) between the pCT and sCTs was calculated
at the beam’s distal and proximal edge of the GTV for 13
equally spaced angles of 15
⁰
through the ipsilateral lung.
Undershoot and overshoot at the distal edge results in an
under-coverage of the target and higher dose in normal
tissues respectively, and vice versa at the proximal edge.
To summarize the results, first for each scan and angle,
the 95th percentile ∆R in undershoot (∆R
u
) and overshoot
(∆R
o
) was determined at the distal and proximal edge of
the tumor respectively. Secondly, for each angle and
patient, the 90th percentile of ∆R
u
and ∆R
o
were
calculated. Finally, median and inter-quartile ranges for
these beam-specific range uncertainties were evaluated.
Results
Figure 1 shows the median and inter-quartile range of ∆R
u
and ∆R
o
for the 13 different angles. For both ∆R
u
and ∆R
o,
the range errors of the lateral beams (around 90
⁰
) are
significant lower (paired T-test, p < 0.05) than the anterior
and posterior beams. Moreover, there is considerable
inter-patient differences in range uncertainties.
Figure 1. Median and inter-quartile range of the range
error ∆R in overshoot and undershoot at the beam’s
proximal and distal edge of the tumor respectively. The
angle of 90
⁰
is the lateral beam to the ipsilateral lung.
Conclusion
Variation in anatomy during the course of irradiation
causes variation in range, possibly leading to tumor under-
coverage and high dose in normal tissue. These range
uncertainties depend on patient and beam angle, and are
smaller for the lateral beams. Taking these beam-specific
range uncertainties into account, could improve the
robustness of proton treatment plan against anatomical
variations.
PO-0870 DIBH produces a meaningful reduction in lung
dose for some women with right-sided breast cancer
J.L. Conway
1
, L. Conroy
1
, L. Harper
1
, M. Scheifele
1
, W.
Smith
1
, T. Graham
1
, T. Phan
1
, H. Li
1
, I.A. Olivotto
1
1
Tom Baker Cancer Centre, Radiation Oncology, Calgary-
Alberta, Canada
Purpose or Objective
To determine whether deep inspiration breath hold (DIBH)
produced a clinically meaningful reduction in pulmonary
dose in comparison to free breathing (FB) during adjuvant
loco-regional radiation (RT) for right-sided breast cancer.
Subsequently, to prospectively evaluate DIBH in right-
sided breast cancer cases with a FB V20Gy ≥30%.
Material and Methods
Thirty consecutive women with breast cancer treated with
tangent pair RT following breast conserving surgery were
included. ESTRO guidelines were used to contour right-
sided IMC nodes on DIBH and FB scans, with care taken to
ensure comparability between scans. A four-field,
modified-wide tangent plan was developed on each scan
to include the right breast and full regional nodes with a
minimum dose of 80% to the IMC CTV. The junction
between the supraclavicular and tangent fields was at the
inferior extent of the ossified medial clavicle. Treatment
plans were calculated in Eclipse using Acuros algorithm
version 11. FB and DIBH plan metrics were compared using
Wilcoxon-signed rank testing. Commencing in March 2016,
as per a new institutional policy based on the above
results, all right-sided breast cancer patients with a FB
ipsilateral lung V20 ≥30% had a DIBH treatment plan
developed prior to compromising on IMC coverage. If the
absolute difference in lung V20 was ≥5% between plans,
the DIBH plan was used. The junction was moved
superiorly in only one case.
Results