389 / 1082
S374
ESTRO 36 2017
_______________________________________________________________________________________________
ITVs to ensure adequate coverage of remaining
treatments. The adapted mid-treatment PTVs (mtPTVs)
were then compared to the pretreatment PTVs (ptPTVs).
Results
Four patients undergoing adaptive EBRT for cervix cancer
were eligible. In total, 100 CBCT images were analysed.
The mean CTV volume decrease was 42% by the end of
treatment, with 44% of that change (21-63%) occurring
within the first week. CTV movement was patient specific,
with the largest variations noted anteriorly (1.1cm to
5.5cm) and superiorly (0.67cm-4.3cm). In patients with
greater than 1cm variation anteriorly and superiorly, most
variation (63%-100%) occurred within the first 2 weeks of
treatment.
Given that both volume and movement changes occurred
predominantly within the first 2 weeks, mtPTVs were
created based on a week 2 ITV. As movement was mainly
superiorly and anteriorly, an anisotropic 1cm ITV margin
with 0.5cm posteriorly formed the mtPTV. 93% of
treatments from week 2 to 5 were covered, with 3 patients
demonstrating up to a 28% reduction in volume compared
to the ptPTVs. If the mtPTV was created using a 1.5cm
anisotropic ITV margin (0.5cm posteriorly), treatment
coverage increased to 98%, however treated volume was
up to 28% larger than ptPTVs. mtPTV margins based on a
week 3 ITV, showed no significant coverage difference
from those based on week 2.
Reduced OAR dose was also noted, with treated bowel
volume using the mtPTV decreasing by 20.7% compared
the ptPTV.
Conclusion
Preliminary analyses of patients undergoing adaptive
external beam treatment for cervical cancer, shows a
dramatic reduction in volume and movement within the
first 2 weeks of treatment. mtPTVs based on week 2 ITVs
with an anisotropic 1cm margin (0.5cm posteriorly),
resulted in adequate treatment coverage in more than 90%
of remaining treatments. Treated bowel volume also
decreased by 20%. These results support mid-treatment
plan adaptation, although greater patient numbers are
required for this to be validated.
PO-0721 Prediction of local recurrence using
pretreatment 18FDG PET/CT radiomics features in
cervical cancer
F. Lucia
1
, M. Desseroit
2
, O. Miranda
1
, D. Visvikis
2
, J.
Malhaire
1
, P. Robin
3
, O. Pradier
1
, M. Hatt
2
, U. Schick
1
1
University Hospital- Brest- France, Radiation oncology,
Brest, France
2
LaTIM- INSERM- UMR 1101- University of Brest, LaTIM-
INSERM- UMR 1101- University of Brest, Brest, France
3
Nuclear Medicine department- University Hospital-
Brest- France, Nuclear Medicine department- University
Hospital- Brest- France, Brest, France
Purpose or Objective
Adequate prediction of tumor response to definitive
chemoradiotherapy (CRT) in cervical cancer (CC) patients
is important to offer a personalized treatment. The aim of
this study was to determine if radiomics features in
18
F-
fluorodeoxyglucose
(FDG)
positron
emission
tomography/computed tomography (PET/CT) could help
in predicting local recurrence in CC.
Material and Methods
Sixty-nine patients with International Federation of
Gynecology and Obstetrics (FIGO) 2014 stage I2B-
IVA cervical squamous cell carcinoma receiving synchronic
CRT from August 2010 to May 2015 were enrolled in this
study. Six (9%), nineteen (27%), thirty-five (51%) and nine
(13%) women presented with FIGO stage I, II, III and IV,
respectively. 18F-FDG PET/CT examination was performed
on each patient before CRT which consisted of external
beam radiotherapy (45 to 50.4 Gy in 1.8-2 Gy per fraction)
combined with weekly platinum salts, followed by a high-
dose-rate brachytherapy boost (24-26 Gy in 4 fractions).
Radiomics (intensity, shape and textural) features of the
primary tumor volumes delineated with the fuzzy locally
adaptive bayesian (FLAB) algorithm in the PET images
were extracted. For textural features three different
quantization approaches were considered. The predictive
value of clinical parameters and radiomics PET features
regarding local recurrence-free survival (LRFS) was
evaluated.
Results
Median follow-up was 1.8 years (0.26-5.6, range). Local
relapse occurred in eleven patients (15.9%). In univariate
analysis, FIGO stage (I-II vs III-IV)(p = 0.0005), T status (T1-
2 vs 3-4)(p < 0.0001), one shape feature (sphericity)(p =
0.001) and two textural features (busyness (p = 0.02), and
Grey Level Non Uniformity (GLNU)(p =0.01) were
significantly correlated with local recurrence. The
combination of two parameters, such as busyness and
sphericity (96.8 % vs 38.2%, HR, 23.51, p<0.0001) or GLNU
and sphericity (94.6% vs 23.8%, HR, 20.55, p<0.0001) led
to highly accurate models (sensitivity and specificity of
90.9%-77.3% and 81.8%-88.6% respectively).
Conclusion
In cervical cancer, radiomics features such as sphericity,
busyness and GLNU from FDG PET images are significant
independent predictor factors for local recurrence.
Further research in refining the predictive value of these
models is needed to justify dose escalation in these
patients, which is already ongoing on a prospective cohort
treated at our institution.
PO-0722 Symptomatic pelvic insufficiency fracture in
women after pelvic RT- is there a dosimetric correlate?
Z. Horne
1
, M. Dohopolsky
1
, S. He
1
, B. Gill
1
, S. Beriwal
1
1
University of Pittsburgh Cancer Institute, Radiation
Oncology, Pittsburgh, USA
Purpose or Objective