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

82 / 146

Data were collected from 4 Philips Brilliance scanners (6, 16 and 64 slices).
For each type of examination we evaluated the max, min, mean DLP and
its standard deviation, the number of examinations and SEM.
Results: Data were extracted through Gray Detector for every CT scanner
and considering the whole patients sample. The parameters described were
calculated using pivot analysis in a dedicated Excel spreadsheet. The large
number of data collected (46,000 exams/year) allowed to create a robust
database useful for statistical analysis. We believe it is of great impor-
tance to start monitoring DLP values also for examinations not included
in standard DLR evaluation.
Conclusion: The mean values of DLP are a photograph of the actual clin-
ical practice and a starting point for optimization. Comparison with data
in referenc indeed indicated that some examinations need further in-
vestigation and optimization. Despite we found an incorrect identification
on the RIS of exams with multi anatomical district, e.g. Chest/Abdomen/
Pelvis, Gray Detector is overall a valid system for supporting EU Directive
2013/59 compliance.
Reference
[1]
B.258
TEXTURE ANALYSIS IN CT AND PET: A PHANTOM STUDY FOR FEATURES
VARIABILITY ASSESSMENT
G. Feliciani *, M. Bertolini, F. Fioroni, M. Iori. Arcispedale Santa Maria Nuova,
IRCCS, Reggio Emilia, Italy
Purpose: In the field of Radiomics, the search for reliable and stable metrics
is a central theme. The aim of this paper is to analyze textural features vari-
ability on phantom varying CT and PET clinical acquisitions parameters for
the study of the Non Small Cells Lung Cancer and Head and Neck Cancer
patients.
Materials and Methods: Acquisitions from Catphan 600 were performed
on a Philips Ingenuity CT. Acquisition parameters were varied according
to AAPM lung acquisition protocols in particular changing mA, pitch,
slice thickness and reconstruction algorithm (FBP and iterative algo-
rithms. Cylindrical VOIs were segmented for 7 different uniform materials
(7 different known HU) inserts inside the Catphan and 1 VOI on its water
part keeping VOI volume constant to 1.25 ml ± 0.01 ml. PET acquisitions
were performed on NEMA IEC phantom from different scanner manufac-
turers. PET acquisitions and reconstruction parameters were varied to
match clinical protocols. Spheres inside phantom were segmented using
a 40% isocontour algorithm. After feature extraction, Features Coefficient
of Variations were calculated for each feature as COV = SD/mean × 100 in
order to classify features as stable (COV < 10%), semi-stable (10% < COV < 20%)
and unstable (COV > 20%).
Results: Textural features exhibit strong dependence both in PET and CT
on the number of pixels employed in their calculations. Reconstruction
matrix and slice thickness should be maintained as constant as possible
both in PET and CT. Strong VOI dimension dependencies arises in textural
features with large volume variations. In CT, 18 textural features out of 69
calculated, were considered stable (COV < 10%) for different materials and
varying acquisition mA. In PET, textural features calculation results show
a COV > 20% considering different scanner manufacturers.
Conclusions: Textural features show a high variability with the recon-
struction algorithm and matrix size. A standardization is necessary to
perform reliable Radiomics investigations.
B.259
IMAGE QUALITY IN DIGITAL BREAST TOMOSYNTHESIS: A PHANTOM
ASSESSMENT
C. Feoli A. Sarno F. Di Lillo G. Mettivier P. Russ a Università
degli Studi di Napoli Federico II, Napoli, Italy; b Istituto Nazionale di Fisica
Nucleare, Napoli, Italy
Introduction: The authors investigated the imaging performance of Digital
Breast Tomosynthesis (DBT) in terms of spatial linearity, visibility of
microcalcifications (m-c) and the dose-normalized contrast to noise ratio
(CNRD) of masses in a 3D breast phantom, with respect to digital mam-
mography (DM).
Material and Methods: A DBT test object (CIRS mod. 020 BR3D, a stack
of 6 slabs each 10-mm thick, simulating 50/50 breast tissue) was scanned
with a Siemens Mammomat Inspiration DBT unit at 29 kVp, 318.3 mAs,
W/Rh anode/filter, 3.43 mGy MGD. The phantom includes a slab with m-c
clusters, masses and fibers of various sizes. The position of this slab in the
stack was varied, for testing image quality as a function of the distance of
the plane containing the details from the top surface of the phantom. The
longitudinal spatial linearity and the transverse spatial linearity were evalu-
ated by locating the inclusions plane in the DBT images and by evaluating
the distance between clusters. An analysis of the reconstructed size of m-c
and the CNRD = CNR/D½ of masses is reported.
Results: In DBT: (1) the distance between clusters differs by 4% with respect
to the actual distance, at all depths in the phantom (in DM the correspond-
ing distance deviates linearly from the actual value as a function of depth,
with a discrepancy from −27% to +20%, due to the beam divergence);
(2) the reconstructed depth of the details differs from the actual depth by
1.4% ± 0.8% in the 60-mm phantom thickness; (3) the FWHM of the profile
across the largest m-c differs by less than 4.4% from the nominal size;
(4) the CNRD of 6.3-mm masses is in the range 0.27–1.27 in the range of
depths in the phantom; (5) the maximum size of visible fibers and m-c in
DBT are 0.38 mm and 0.196 mm respectively.
Conclusions: DBT images reproduce accurately within a few percent the
size and the position of the details, with no significant dependence on depth
in the phantom. The CNRD is constant with depth and lower than in
DM.
B.260
OPTIMISATION OF PAEDIATRIC FULL SPINE EXAMINATION: THE
EXPERIENCE OF CARLO POMA GENERAL HOSPITAL IN MANTOVA
C. Ferrari *, C. Minari. Azienda Ospedaliera Carlo Poma, Mantova, Italy
Introduction: Many radiological departments are provided with conven-
tional x-ray systems not dedicated to paediatric muscle-skeletal imaging.
Aim of this study is to optimise the paediatric full-spine examination in
terms of dose reduction, patient and dedicated shielding devices position-
ing, taking necessarily into account both economic criteria and available
equipment.
Materials and Methods: The optimisation process consisted of different
steps. First, based on a literature review and experimental measure-
ments, the most appropriate equipment has been selected, in order to
achieve the lowest exposure and effective dose level as possible. Then a mul-
tidisciplinary group, including the physiatrist, the radiologist, the
radiographer and the medical physicist, was created in order to highlight
the critical aspects of the whole process and optimise the radiographic pa-
rameters on the basis of diagnostic requirements and sheared criterions.
Another important step in the optimisation process was the education and
training of all personnel involved in clinical practice. Finally, for a moni-
toring period of six months, through the information stored in the PACS
system, every examination has been analysed and recorded data have been
compared with data obtained before optimisation.
Results: Major improvements in the optimisation process involved posi-
tioning of the patient, use and positioning of shielding devices, collimation,
automatic exposure control, kVp and filtration settings. The analysed data
showed a significant reduction in terms of dose, especially in case of frontal
projection.
Conclusions: Cooperation of all the different professionals involved in the
workflow, in situ training and continuous monitoring are essential for ac-
quisition and consolidation of appropriate radiological technique. Wide dose
radiation reduction according to ALARA principle for paediatric proce-
dures can be achieved by means available in a general purpose radiological
department.
e77 Abstracts/Physica Medica 32 (2016) e71–e96