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

69 / 146

A.215
BIOLOGICAL TARGET VOLUME SEGMENTATION FOR RADIOTHERAPY
TREATMENT PLANNING
A. Stefano S. Vitabil G. Russo D. D’Urs M. Ippolito F. Marlett
M.G. Sabini I.V. Patti S. Pittera D. Sardina O. Gambin R. Pirron
E. Ardizzone M.C. Gilardi a Istituto di Bioimmagini e Fisiologia Molecolare
– Consiglio Nazionale delle Ricerche (IBFM-CNR), Cefalù, Italy; b Dipartimento
di Ingegneria Chimica, Gestionale, Informatica, Meccanica (DICGIM), Università
degli Studi di Palermo, Palermo, Italy; c Dipartimento di Biopatologia e
Biotecnologie Mediche (DIBIMED), Università degli Studi di Palermo, Palermo,
Italy; d Azienda Ospedaliera Cannizzaro, Catania, Italy;e Università degli Studi
di Catania, Catania, Italy
Introduction: The CT gross tumor volume (GTV) is the gold standard for
target tumor delineation in head and neck cancer (HNC) radio-treatments
(RT). Nevertheless, CT images may not show the viable extension of tumors.
Vice versa, FDG-PET provides functional information, offering the oppor-
tunity to radically change the patient treatment. In RT, within or without
the GTV, it is possible to define the biological target volume (BTV) on PET
image and to apply it to plan the radiotherapy treatment. BTV delinea-
tion is challenging because of the low spatial resolution and high noise level
in PET images. For these reasons, BTV varies substantially depending on
the algorithm used. In this study, a fast, accurate, and scanner indepen-
dent BTV segmentation method is proposed.
Materials and Methods: The RandomWalks (RW) method provides a target
recognition in images in which the start seeds are specified by the user.
An enhanced RW algorithm with automatic seed detection and with an
adaptive probability threshold value, instead of the fixed one of the orig-
inal RW method, is introduced. The proposed method has the benefit to
make the algorithm not subjective and to take into account the intensity
changes between slices along the tumor volume. Phantom experiments are
performed to estimate the BTV segmentation accuracy. The same algo-
rithm is then applied in 15 HNC patients to assess its clinical feasibility.
BTV changes between pre and post RT are also evaluated.
Results: High dice similarity coefficients (>85%) and true positive volume
fractions (>87%), and low Hausdorff distance values (<3 mm) confirm the
method accuracy. In addition, the method has real-time performance
(0.3 sec) and it meets the physician’s requirements in a RT environment.
Conclusions: The proposed method is very powerful in terms of BTV seg-
mentation and time performance, and it may be used daily as a medical
decision support system to enhance the current methodology performed
by healthcare operators in RT.
A.216
A NEW 3D RADIOTHERAPY TECHNIQUE REDUCES SKIN TOXICITY FOR
LARGE BREAST AND/OR HIGH BMI PATIENTS
G. Stimato *, D. Gaudino, C. Di Venanzio, A. Mameli, E. Infusino, L. Bellesi,
E. Ippolito, S. Silipigni, C. Rinaldi, R. D’Angelillo, S. Ramella,
L. Trodella. Universita’ Campus Biomedico di Roma, Rome, Italy
Introduction: The aim of this study was to develop an alternative 3DCRT
standardized fields class solution for whole breast radiotherapy in pa-
tients with large/pendulous breasts and/or high body mass index (BMI) to
improve dose homogeneity and reduce hot spots and extra-target dose
irradiation.
Materials and Methods: Two treatment plans (tangential fields-TF and
standardized five fields-S5F) were generated for patients with large
breasts (PTV > 1000 cc and/or BMI > 25 kg/m2), supine positioned. S5F
plans consist of two wedged tangential beams, AP: 20° for right breast
and 340° for left and PA: 181° for right breast and 179° for left. A field in
field in medial-lateral beam and additional fields were added to reduce
hot spot areas and extra-target irradiation and to improve dose distribu-
tion. The total prescribed dose was 50 Gy/25 die. V95% PTV, V105% PTV,
Dmax PTV, homogeneity index (HI) and conformity index (CI) were
recorded. An opposite normal tissue structure (Body-PTV) was gener-
ated. V105%, V107%, V5 Gy LUNG, V20 Gy LUNG, V40 Gy LUNG, V50 Gy
LUNG, Dmax LUNG, V5 Gy HEART, V15 Gy HEART, V25 Gy HEART and
Dmax HEART were recorded. Statistical analysis were performed using
SYSTAT, v.12.0 (SPSS, Chicago, IL).
Results: 38 patients were included in the analysis. S5F improved HI (8.4
vs. 10.1; p ≤ 0.001) and Dmax PTV and V105% PTV were significantly reduced,
while CI and V95% PTV were not. However, the constraint V95% PTV > 95%
was achieved for all S5F plans, not for TF. The extra-target tissue irradia-
tion was significantly reduced using S5F for V105% (cc) and V107% (cc). These
reductions were of particular interest because the difference in tissue ir-
radiation was very high (46.6 cc vs 3.0 cc, p ≤ 0.001 for V105% and 12.2 cc
vs 0.0 cc, p ≤ 0.001 for V107% for TF and S5F plans, respectively). No dif-
ferences in lung and heart dosimetric parameters were recorded.
Conclusion: S5F technique allows treatment of patients with large breasts
or high BMI, increasing HI, decreasing Dmax PTV and by a factor of 15 hot
spots in extra-target tissues.
A.217
IMPLEMENTATION OF A NEW STRATEGY FOR DOSE TRACKING AND OF
NOVEL RADIOBIOLOGICAL MODELS FOR ADAPTIVE RADIOTHERAPY
S. Strolin E. Mezzeng A. Sarnelli V. Bruzzanit S. Marzi S. Ungani
G. Sanguineti L. Marucci M. Benassi L. Strigari a Istituto Nazionale
Tumori Regina Elena, Roma, Italy; b Istituto Scientifico Romagnolo per lo Studio
e la Cura dei Tumori IRST IRCCS, Meldola, Italy
Introduction: The feasibility of the dose tracking approach in Head-and-
neck (HN) cancer was tested by implementing and comparing two
deformable image registration strategies.
Material and Methods: For each of 15 patients with diagnosed Head-and-
neck (HN) cancer a planning CT (pCT) and a total of 6 weekly CT (wCT) were
acquired at our institute. The pCT, wCT, manually contoured structures and
dose distributions were imported in Raystation TPS. Two approaches were
implemented, both using the hybrid algorithm for region of interest (ROI)
mapping. In the first approach (RH), the body contour was considered as
focus ROI while manually contoured GTV, parotid gland and spinal cord
were used as controlling ROIs. The second method (H) considers the body
contour as focus ROI, without including any controlling ROIs in the anal-
ysis. DICE similarity index was used to assess the agreement of propagated
contours between the two approaches. Dose/volume statistics and radio-
biological indices were computed and compared.
Results: The RH approach provided the best results, with median DICEs
values of 0.88, 0.94 and 0.94 for GTVs, parotid gland and spinal cord, re-
spectively. The corresponding values in H strategy were 0.63, 0.82 and 0.88,
respectively. Dose differences on GTVs showed a median variation within
1% with minimal values up to 8%; tumor control probability values for
planned, RH and H approach were 63.7%, 69.7% and 61.9%, respectively. The
average normal tissue complication probabilities in planned, RH-based and
H-based accumulated dose-volume histograms (DVHs) were 36%, 46% and
34%, respectively, for homo-lateral parotids and 28%, 36% and 27% for contra-
lateral parotids.
Conclusion: The RH strategy generates structures with higher similarity
to manually contoured ROIs thus proving the good accuracy of the gener-
ated deformation matrix. This approach gives an appropriate strategy to
perform dose tracking in HN cancer patients eligible for adaptive
radiotherapy.
A.218
EVALUATION OF THE AUTOSEGMENTATION TOOL OF NORMAL TISSUE
STRUCTURES IN PROSTATE CANCER: A MULTICENTRIC STUDY
S. Strolin* L. Strigar V. Bruzzanit S. Ungani R. Nigro S. Riccardi
M. Casale A. Tonnetti M.T. Russ L. Chiatt R. Cassese N. Franza
a Istituto Nazionale Tumori Regina Elena, Roma, Italy; b OGP S. Camillo de Lellis,
Rieti, Italy; c Azienda ospedaliera Santa Maria, Terni, Italy; d Dosimetrica, Italy;
e Ospedale Belcolle, Viterbo, Italy
Introduction: In the treatment planning process, manual delineation of
organs at risk (OARs) is tedious and time consuming. The aim of this study
was to evaluate and clinically validate the feasibility and time saving of the
autosegmentation algorithm of the prowess treatment planning system (TPS)
using computed tomography (CT) images.
e64 Abstracts/Physica Medica 32 (2016) e1–e70