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

111 / 146

total therapeutic dose changed by PET/CT information. BTV (biological target
volume) identified by PET was boosted in 28% of patients for whom a focal
dose escalation was considered clinically relevant.
Conclusions: In radiotherapy treatment planning of head and neck disease
PET/CT compared with CT provide a more precise target definition and
prevent exclusion of pathologic areas not detectable on CT. In our popu-
lation addition of PET information leads to a modification of the irradiated
target in 61% of the patients. The ability of PET/CT to identify the BTV allows
a dose escalation on the more aggressive part of the disease.
C.360
COMPARISON STUDY OF TUMOUR UPTAKE DISTRIBUTION IN PRE AND
POST TREATMENT IMAGING FOR 90Y LIVER RADIOEMBOLIZATION
E. Gallio * M. Finessi E. Richetta M. Stasi R.E. Pellerito G. Bisi
R. Ropolo a S.C. Fisica Sanitaria, A.O.U. Città della Salute e della Scienza, Torino,
Italy; b S.C. Medicina Nucleare U, A.O.U. Città della Salute e della Scienza, Torino,
Italy; c S.C. Fisica Sanitaria, A.O. Ordine Mauriziano, Torino, Italy; d S.C. Medicina
Nucleare, A.O. Ordine Mauriziano, Torino, Italy
Introduction: For unresectable hepatic malignancies, a possibility of therapy
is the radioembolization with 90Y-microspheres. The treatment is prelim-
inary simulated by SPECT 99mTc-MAA acquisitions for quantify lung shunt,
to avoid extra-hepatic abdominal shunt and to determinate the optimal 90Y
therapeutic activity. After 90Y infusion, PET-CT acquisition is performed to
verify the radioisotope distribution.
Materials and Methods: 15 patients were included in this study, 4 treated
with glass microspheres and 11 with resin ones. 99mTc-MAA SPECT-CT at-
tenuation (glass) and 99mTc-MAA SPECT-CT attenuation and scatter
corrected images (resin) were made. After treatment, a non-TOF PET-CT ac-
quisition was performed for all patients. Tumour VOI was drawn to estimate
uptakes and volumes, both for SPECT and for PET. For 90Y tumour distri-
bution a qualitative evaluation was made according to the following scores:
0 = SPECT and PET tumour uptake distributions are perfectly coincident
(within 0.2 cm), 1 = the two distributions are in accordance within 1 cm,
2 = between 1 and 3 cm, 2 = more than 3 cm.
Results: Qualitative visual comparison of tumour uptake presented score
0 for 10 patients (67%), score 1 for 1 patient (6%), and score 2 for 4 pa-
tients (27%). No patient exhibited score 3. Preliminary results about tumour
distribution volume showed a consistence difference between SPECT and
PET. SPECT 99mTc-MAA volumes were significantly lower compared to PET
ones (~−40%). These differences should be ascribed to dissimilar catheter
position between angiographic session and possible volume effects.
Conclusions: This study pointed out the attention about the importance
of PET-CT verification after radioembolization because simulation study was
performed on MAA distribution that is similar but not equal to 90Y
microspheres. At the congress, all results will be presented. These prelim-
inary results must be confirmed by further ongoing investigation with a
larger patient dataset.
C.361
CALCULATION OF TUMOR AND NORMAL TISSUE BED IN 90Y LIVER
RADIOEMBOLIZATION WITH DIFFERENT DOSIMETRIC METHODS
E. Gallio * M. Finessi E. Richetta M. Stasi R.E. Pellerito G. Bisi
R. Ropolo a S.C. Fisica Sanitaria, A.O.U. Città della Salute e della Scienza, Torino,
Italy; b S.C. Medicina Nucleare U, A.O.U. Città della Salute e della Scienza, Torino,
Italy; c S.C. Fisica Sanitaria, A.O. Ordine Mauriziano, Torino, Italy; d S.C. Medicina
Nucleare, A.O. Ordine Mauriziano, Torino, Italy
Introduction: Radioembolization with 90Y microspheres is an effective treat-
ment for unresectable liver tumors. Two types of microspheres are available:
resin (SIR-Spheres®) and glass (Theraspheres®). The aim of this study is
to compare BED (Biological Effective Dose) values obtained by three dif-
ferent dosimetric methods.
Materials and Methods: 13 HCC patients were treated: 4 with glass
(3.0 ± 0.5 GBq) and 9 with resin microspheres (1.4± 0.4 GBq) with the dose
goal of 120 Gy to lobe (glass) or lesion T (resin) and maximum 40 Gy to
normal liver NL (resin). Doses to T and NL were calculated with three
different methods (AAPM, multi-compartmental MIRD and voxel-based) and
from these BED values were obtained. For AAPMmethod, total body planar
images were used; 99mTc-MAA SPECT-TC attenuation (glass) and 99mTc-
MAA SPECT-TC attenuation and scatter corrected images (resin) were
employed for multi-compartmental MIRD and voxel methods. BED was cal-
culated from the average absorbed dose. Mean dose per activity unit and
percent differences were calculated for different methods and sphere types.
Results: Mean BED (± for lesion were: AAPM 78 ± 35, 308 ± 165 Gy/
GBq; MIRD 136 ± 97, 273 ± 235 Gy/GBq and Voxel 168± 81, 206 ± 154 Gy/
GBq for glass and resin spheres respectively. For normal liver BED were:
AAPM 5± 3, 27 ± 19 Gy/GBq; MIRD 78 ± 65, 73 ± 45 Gy/GBq and Voxel 54 ± 21,
57 ± 21 Gy/GBq for glass and resin.
AAPM doses were in mean about −35% less than other two dosimetric
methods for T and higher discrepancies between methods were found for
NL (mean 40% up to 90%). Comparison between MIRD and voxel dosim-
etry showed a smaller range of variation both for T and NL: 10% for resin
and ~25% for glass.
Conclusions:AAPMmethod seems less accurate for BED estimation. Instead,
MIRD and voxel based dosimetry are more confident with each other. These
preliminary results must be confirmed by further ongoing investigation with
a larger patient dataset.
C.362
A DIRICHLET PROCESS MIXTURE MODEL FOR AUTOMATIC 18F-FDG PET
IMAGE SEGMENTATION
M.G. Giri *, C. Cavedon. Azienda Ospedaliera Universitaria Integrata, Verona,
Italy
Purpose:To investigate the use of a Dirichlet process mixture model (DPM)
for automatic tumour edge identification on 18F-FDG PET.
Methods: TheDPMalgorithmimplemented in the statistical software package
R was used in this work. Contouring accuracy was evaluated on an IEC
phantom(spherical inserts, diameter 13–37mm) acquired by a Philips Gemini
BigBore PET-CT scanner, using 9 target-to-background ratios (TBR, range
2.5–70). Accuracywas then tested on a digital phantomsimulating spherical/
uniform lesions and on a consecutive series of 20 clinical cases (lung and
oesophageal cancer). The influence of the DPMparameters on contour gen-
eration was studied in two steps. Firstly, the IEC spheres with 22 mm and
37 mm diameter were studied, by varying the DPM parameters until the
diameters were obtained within 0.2% of the true value. In the second step,
the results obtained for this training set were applied to the entire dataset.
Results: Only one parameter was found to influence segmentation accu-
racy. This parameter was linearly correlated to the variance of the tested
ROI: a calibration curve was determined to calculate the optimal value of
the parameter from the ROI variance. This calibration curve was later used
to contour the whole dataset. Accuracy of volume estimation was better
than 6% and bias was below 2% on the whole dataset (1 SD). The overlap
between true and automatically segmented contours (dice similarity co-
efficient) was 0.93 (SD = 0.02).
Conclusions: The proposed DPM model was able to reproduce known
volumes of FDG concentration with high overlap between segmented and
true volumes. For all the analysed inserts of the IEC phantom, the algo-
rithm proved to be robust to variations in radius and TBR. The main
advantage was that no setting of DPM parameters was required in advance.
Furthermore, the algorithm did not need any preliminary choice of the
optimum number of classes and no assumption about the shape and the
tracer uptake heterogeneity was required.
C.363
OPTIMIZATION OF ACQUISITION PROTOCOL OF 11C-LABELED
METHIONINE PET/CT
T. Licciardello * P. Pisciotta G. Russo G. Sabini L. Valastro
F. Scopelliti S. Cosentino M. Ippolito a Azienda Ospedaliera Cannizzaro,
Catania, Italy; b Università degli Studi di Catania, Catania, Italy; c Istituto di
Bioimmagini e Fisiologia Molecolare-Consiglio Nazionale delle Ricerche
(IBFM-CNR), Cefalù (PA), Italy
e106 Abstracts/Physica Medica 32 (2016) e97–e115