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

103 / 146

of this study was to measure FDG PET standardized uptake values (SUVs)
change over time in three physiological compartments: liver, lung and Blood
Pool (BP).
Material and Methods: 100 patients enrolled in Dual Point clinical trial
on Hodgkin disease were retrospectively evaluated. PET/CT imaging was
performed at either 1 h (PET1) and at 2 h (PET2) after FDG administra-
tion. DICOM headers were analyzed to collect information on the scanning
procedure. A three-dimensional region of interest (ROI) was drawn on the
fused PET image to measure the mean SUV of the three compartments at
PET1 (SUV1) and at PET2 (SUV2). DeltaSUV ((SUV2-SUV1)/SUV1) and nor-
malized SUV (SUV/((SUV1 + SUV2)/2)) were calculated for each pair of data.
Linear model has been applied to the data. Kolmogorov–Smirnov (KS) test
was used to compare the model to the data.
Results: Uptake time mean values were (72 ± 17) min and (133 ± 18) min
for PET1 and PET2 respectively. The liver’s SUV1 and SUV2 for all patients
were (1.82 ± 0.40) g/mL and (1.52 ± 0.35) g/mL respectively while the BP’s
SUV1 and SUV2 were (1.37 ± 0.39) g/mL and (1.07 ± 0.36) g/mL respective-
ly. Lung’s SUV1 was (0.40 ± 0.10) g/mL and SUV2 was (0.33 ± 0.09) g/mL. The
mean liver, lung and BP SUVs were (−16 ± 8)% (range: −35% to 0%),
(−23 ± 12)% (range: −57% to −1%) and (−18 ± 10)% (range: −61% to −1%). Mean
SUV variations were respectively 0.25, 0.10, and 0.27 SUV/hour for liver,
lung and BP SUVs. The normalized SUV mildly correlated to the uptake time
(Pearson correlation coefficients were 0.63, 0.58 and 0.53 for liver, BP and
lung respectively). SUV1 was recalculated using the linear model and com-
pared to measured data. KS values were 0.10(p = 0.99), 0.12(p = 0.99) and
0.13(p = 0.99) for liver, lung and BP, respectively.
Conclusion: SUV decreases over time in the physiological compartments.
C.333
APPLICATION OF LIDEA (LIVER DETECTION ALGORITHM) FOR PET/CT
SCANS
E. Bertone F. Bergesi A. Terull P. Cerell S. Chauvie a Medical
Physics Unit, ASO S. Croce e Carle, Cuneo, Italy; b School of Medical Physics,
University of Torino, Torino, Italy; c INFN, Torino, Italy
Introduction: The aim of this work was to study the distribution of liver
SUV in a large population of patients undergoing whole-body 18FDG Pos-
itron Emission Tomography/Computed Tomography (PET/CT).
Material and Methods: LIDEA (LIver DEtection Algorithm) is an algo-
rithm that automatically detects the liver position in whole-body 18F-
fluorine-deoxyglucose PET/CT and calculates the mean and the standard
deviation of voxels’ Standardized Uptake Value inside it. Within this work
we added other two metrics to describe image noise. Image Roughness (IR)
measures the pixel to pixel variability in the single image, as the average
in different ROI of the coefficient of variation (COV) of the pixel values. The
Background variability (BV) is the noise quantification proposed in the NEMA
NU-2 instructions and it is calculated as the COV of the ROI averages across
the ROIs.
Results: In the 630 analyzed PET scans, coming from more than 40 PET
centers distributed in 21 countries, the algorithm identified the liver in 98.9%
of the cases. The liver SUV was (2.01 ± 0.55). The distribution of liver SUV
of the patients was not Gaussian (p = 4e−8 with the Shapiro–Wilk test). The
values of the noise indexes were SD (0.32 ± 0.12), IR (0.16 ± 0.06) and BV
(0.07 ± 0.05).
Conclusion: The algorithm was used to retrieve average liver values in a
large population of PET/CT scan and to analyze image noise with different
metrics. These would be used to automatically define good/bad statistics
scan within a framework of multi-center clinical trials.
C.334
DEPTH-OF-INTERACTION DISCRIMINATION TO CORRECT PARALLAX
ERROR IN HIGH RESOLUTION PET
M. Bettiol E. Preziosi C. Borrazzo M.N. Cinti A.J. Gonzale R. Pan
a Department of Molecular Medicine, Sapienza, Roma, Italy; b Doctorate School
of Biology and Molecular Medicine – SAIMLAL Department, Morphofunctional
Sciences – Biophysics, Roma, Italy; c Institute for Instrumentation in Molecular
Imaging, I3M-CSIC, Valencia, Spain; d Department of Sciences and Medical and
Surgical Biotechnologies, Sapienza, Roma, Italy
Introduction: High resolution PET systems are often limited by the par-
allax error due to the lack of information about the Depth of Interaction
(DoI) inside the crystal of the incoming gamma-photons.
In this work, an effortless DoI estimator suitable for monolithic scintilla-
tion crystals has been proposed. This development was done in the
framework of the MindView PET/MRI brain imaging European Project.
Materials and Methods: A test PET module with a 50 × 50 × 20 mm mono-
lithic Lutetium Yttrium oxyorthosilicate (LYSO) crystal coupled to a 12 × 12
SiPM array has been employed.
Experimental measurements from a 45 degree slanted pencil beam of
511 keV gamma photons have been performed.
The purpose of the slanted beam is to obtain an image in which there is a
direct correlation between the interaction position and the DoI.
In addition, for calibration and validation of the method, the Geant4 sim-
ulation has been also used.
DoI estimation has been performed via a novel parameter related to the
shape of scintillation light distribution produced by the gamma photon in-
teraction inside the crystal. Furthermore, we show an innovative, DoI-
dependent, position algorithm based on the manipulation of the same
distribution.
Results: The proposed method provides a DoI resolution of about 5 mm.
This leads to a parallax error reduction of about 75% in case of 30 mm di-
ameter PET system. Moreover, thanks to the DoI-dependent position
algorithm, the detector spatial resolution is improved of a factor of 15%.
As a consequence of these results, the overall spatial resolution improves
from 20 to 60% in respect to a PET system without DoI-dependent
corrections.
Conclusion: In respect to a high efficiency PET module without DoI-
dependent corrections, the proposed methods allow an improvement in
overall spatial resolution ranging from 20% to 60%, depending on source-
ring center distance. Moreover, this method is particularly engaging since
it is simple to implement and does not require additional hardware
equipment.
C.335
GEOMETRY VARIABILITY IN THE MEASUREMENT OF GAMMA, BETA AND
ALPHA EMITTERS IN RADIONUCLIDE CALIBRATORS
C. Bianchi S.A. Abd Hami P. Colleon S. Andreol a A.O. Papa Giovanni
XXIII, Bergamo, Italy; b The Brunei Cancer Centre, Brunei Darussalam, Brunei
Introduction: The aim of this work is to study the variability of radionu-
clide calibrators outcome with different source volume and container.
Measurements were carried out together with a medical physicist of The
Brunei Cancer Centre, during a IAEA fellowship in our hospital.
Material and Methods: The variation in calibrator response was studied
varying the solution volume in a glass vial and syringes (from 1 to 10 mL).
Activity measured in the syringe was compared with activity difference
in the vial before and after withdrawal and with the value determined
from the vial activity concentration. Volumes were determined by weight-
ing vial with a precision scale. Measurements were made with 2 calibrators
(Capintec CRC15R and 15PET) and with 5 isotopes: 99mTc, 123I, 131I,
223Ra, 90Y. Because 90Y is incorporated in a resin microsphere and the
withdrawal of a uniform solution is difficult, no syringe measurements
were made.
Results: As expected, with 99mTc, the calibrator outcome does not sig-
nificantly vary with the change in volume (<1% from 0.5 to 30 mL) and with
different volume syringes (<1.2%). The same behavior is observed for 223Ra,
in which the α radiation does not influence the calibrator response with
different volume and syringes. Variation in the outcome is observed for 123I,
because of a production of characteristic x-rays. The maximum variation
is 4% for volume in the vial ranging from 0.5 to 13 mL and the activity mea-
sured in syringes deviates from the vial one not less than 20%. The 90Y source
self attenuation is important: for a volume of 1 mL the calibrator outcome
is 19% higher than 5 mL; over 6 mL volume the variability is lower. With
131I β emissions calibrator response slightly changes between vial and sy-
ringes (<4%).
e98 Abstracts/Physica Medica 32 (2016) e97–e115