Physica Medica: European Journal of Medical Physics - Volume 32 - Supplement 1 - February 2016

43 / 146

Material and Methods: All measurements were performed with high energy
photon beams of 6MV RX with an Elekta VersaHD linac equipped with high
energy photon beams (6 and 10 MV RX) in standard and FFF mode. The
measurements were performed with a PFG (Aktina Medical) attached to
the linac head and an iViewGT EPID.
The results obtained with the PFG were compared to those obtained with
an Isocentric Beam Checker (IBC II, Mick Radio-Nuclear Instruments, Inc.)
The PFG has wires of tungsten embedded in an acrylic plate along the beams
cross and inplane directions with a wire spacing at iso of 1 cm. The IBC II
has tungsten markers embedded in an acrylic plate at the centre and at
the corners of 5 different field sizes.
Portal images were taken with the collimator at cardinal angles in order
to assess whether the central wire of the PFG coincides with the collima-
tor axis of rotation. Repeated exposures were performed in order to assess
the reposition accuracy of the EPID and the PFG extracting and reposition-
ing only the former or only the latter.
Results: The results obtained with the two graticules were coherent and
showed that the centre of the central wire did not coincide with the centre
of the EPID. The automatic software routine detected the centre of the wires
with a standard deviation of 0.1 pixels in both directions. Rotating the col-
limator the central wire shifts with a standard deviation of 1.0 pixel
(0.25 mm) with a maximum difference of 2 pixels. The reposition accura-
cy of the EPID and the PFG had a maximum deviation of 1 pixel, while the
reposition accuracy of the PFG was within detection limits.
Conclusion. The results obtained with the Port Film Graticule were co-
herent with those obtained with the IBC II. Its simplicity of use facilitates
QA procedures and makes it a good substitute for the IBC II.
A.128
FORECASTING ALGORITHM TO PREDICT RE-PLANNING IN TOMOTHERAPY:
FOLLOW-UP TOXICITY CORRELATION
N. Maffei G. Guid E. D’Angel B. Meduri E. Mazze A. Ciarmatori
G.M. Mistretta P. Ceroni G. Gottardi P. Giacobazzi T. Costi a Az.
Ospedaliero-Universitaria di Modena, Modena, Italy; b Università di Bologna,
Bologna, Italy
Introduction: A re-plan strategy was evaluated correlating a predictive
neural network (NN) to clinical toxicities after radiation therapy (RT). Adap-
tive RT approach was investigated using IGRT and deformable image
registration based algorithm.
Materials and Methods: 30 Head and Neck (H&N) patients were en-
rolled. Structures of 900 MVCT studies, obtained by Tomotherapy®, were
deformed by RayStation® to investigate inter-fraction organ warping. A dose
accumulation analysis was carried out. Deformed structures and doses were
used to train a NN implemented in MATLAB® finalized to select eligible
cases for re-planning. NN output was correlated with follow-up toxicities
(i.e. dysphagia, dysgeusia, mucositis, salivation) in a 12 month period.
Results: Early toxicity occurred in 74% of patients: 40% (G1), 25% (G2) and
9% (G3); late toxicity in 41%: 30% (G1), 10% (G2) and 1% (G3). A 2nd order
polynomial correlation was detected between the medium-high grade of
early toxicity and the dose of the event; R2 value of 0.93 for G2 and 0.92
for G3. ANOVA multivariate analysis showed an increased frequency of early
(21%) and late (19%) toxicity for smoker patients. Correlating the NN pre-
diction of volumetric/dosimetric variations with weight loss, a benefit of
plan review was estimated for 89.6% of patients. 37% of patients do not need
a re-plan; 25% of them had a weight loss <5%. 63% of patients would benefit
for a re-plan: during the 2nd week for 25% of cases with a weight de-
crease <10%; during the 4th week for remaining 38% of cases, 25% of them
had a weight loss >10%.
Conclusions: To validate predictive approach based on organ warping and
dose deformation, a follow-up correlation result is useful. An increased
number of patients with systematic clinical information should be ana-
lyzed to make robust algorithm and to ensure personalization of patients’
treatment including radiobiological and toxicity data.
A.129
INTRA-FRACTION MOTION IN IMRT, VMAT AND HELICAL TOMOTHERAPY:
IN VIVO DOSIMETRY USING TLD AND LEGO PHANTOM
N. Maffei G. Guidi A. Ciarmator G.M. Mistretta P. Ceron A. Brun
G. Zambelli P. Giacobazzi G. Baldazzi T. Costi a Az. Ospedaliero-
Universitaria di Modena, Modena, Italy; b Università di Bologna, Bologna, Italy;
c Personal Dosimetry Service – Lavoro e Ambiente Srl, Forlì, Italy
Introduction: During free-breathing arbitrary phase, a mean motion re-
construction is acquired using not gated CT. The lack of knowledge of the
tumor and organs at risk (OAR) location can generate possible random/
systematic errors during RT treatment. A home-made anthropomorphic
dynamic phantom was developed to assess, by TLD, the breathing of the
lung district and to quantify the dose variation due to intra-fraction motion.
Materials and Methods: Respiratory motion was simulated by a LEGO
Mindstorms phantom, programmed in LabVIEW and equipped by 8 ribs,
1 OAR and 1 target with 4 degree of freedom. Within a treatment of 40 cGy,
3 planning strategies were compared. Static (S): CT acquired and plan de-
livered with phantom in static mode. Static-Real (SR): equal to the S
condition but plan delivered with human breathing condition. Dynamic (D):
4DCT with breathing phantom and plan based on Maximum Intensity Pro-
jection (MIP) and the ITV generated by the junction of the target contoured
in each phases. For each CT were planned and delivered an IMRT, VMAT
and Helical plan to uniformly irradiate the target. Dosimetry was made using
TLD GR-200 (LiF:Mg,Cu,P).
Results: For each technique, data were normalized to the S IMRT plan. In
SR condition, the dose delivered to the target was 89.2 ± 6.6%. Due to the
motion of the target, TLD measure confirms the uncorrected dose distri-
bution related with the plan. Between the 3 techniques, the Helical plan
allows reaching a greater coverage probably due to the kinetic behavior with
slow machine rotation. In D condition, the dose delivered to the target was
93.5 ± 5.1%. The internal motion was partially accounted with ITV density
re-assignment.
Conclusions: Respiratory motion is often assumed to be the same during
CT and RT cycles. However, due to contraction of the thoracic diaphragm
muscle, it can be slightly different. In this study, a dynamic phantom and
TLD measures have quantified the error and dose distribution in a simu-
lated lung treatment.
A.130
SUSCEPTIBLE-INFECTED-SUSCEPTIBLE MODEL APPLIED TO RT TO
PREDICT PAROTID GLANDS SHRINKAGE DURING 6 WEEKS OF THERAPY
N. Maffei G. Guidi C. Vecchi A. Ciarmatori G.M. Mistretta
P. Ceroni B. Meduri P. Giacobazzi G. Baldazzi T. Costi a Az.
Ospedaliero-Universitaria di Modena, Modena, Italy; b Università di Bologna,
Bologna, Italy
Introduction: To quantify and handle intra-organ variations during treat-
ment sessions, a novel voxel-by-voxel approach has been proposed in
adaptive radiation therapy (ART). The epidemic Susceptible-Infected-
Susceptible (SIS) model was applied to RT challenge to predict morphological
warping in the Head and Neck (H&N) region and to follow single voxel
motion.
Materials and Methods: A cohort of 12 H&N patients, previously treated
by Tomotherapy®, was analyzed. Based on ANACONDA-RayStation® algo-
rithm, Deformable Image Registration (DIR) of 360 daily MVCT studies was
performed. Applying the SIS epidemic model, parotid glands (PG) shrink-
age was evaluated considering each voxel as a single subject and the
Deformed Vector Field (DVF) magnitude as an infection. Daily deformed
mesh grid data were exported by an IronPython® script. The SIS model was
implemented and simulated by a MATLAB® home-made toolbox.
Results: 0.4 cm of voxel displacement was set as clinical threshold within
a [0÷1 cm] range of warping to discriminate S and I cases. Dynamic Time
Warping (DTW) algorithm was used to match simulated epidemic model
and daily PG shrinkage. Lower DTW distance: 2.39 ± 0.66 was reached with
a setting of 7.55 ± 0.69 and 2.45 ± 0.26 for the contact and the recovery rate
respectively. PG warping based on SIS model prediction was confirmed by
physicians in 65% of cases.
e38 Abstracts/Physica Medica 32 (2016) e1–e70