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

101 / 146

B.326
LOW CONTRAST DETECTABILITY ASSESSMENT THROUGH A STATISTICAL
METHOD: HOMEMADE SOFTWARE AND DEDICATED PHANTOM
R. Villa C. Spadavecchia A. Radice C. Pasquali N. Paruccini
E. De Ponti A. Crespi a Scuola di Specializzazione in Fisica Medica, UNIMI,
Milano, Italy; b A.O. San Gerardo, S.C. Fisica Sanitaria, Monza, Italy
Introduction: Low contrast detectability (LCD) refers to the ability of dis-
tinguishing small objects having low contrast.
LCD analysis was frequently achieved through visual assessment of human
observers, thus the evaluation is subject to human perception and deci-
sion criteria. To overcome this limit automated software packages have been
developed.
Materials and Methods: The aim of this work was to provide an objec-
tive tool of LCD assessment, based on a statistical method combined with
the use of a dedicated phantom. This was made up of an acetate sheet
equipped with a central uniform aluminium region (3 × 3cm2, thickness
0.49 mm) and an aluminium step wedge (0.6 × 0.8 cm2 each, with increas-
ing thickness: 0.2–1 mm) for linear conversion from pixel values to
millimetres of aluminium.
A Matlab program that automatically analyses images of the dedicated
phantom was developed. LCD was estimated according to Chao’s statisti-
cal method, analysing a square ROI defined on the uniform aluminium
region, then subdivided into a matrix of subROIs. Standard deviation of
subROIs mean values was calculated.
Normality of subROIs means distribution, basic assumption of statistical
method, was evaluated both with Lilliefors and Jarque-Bera normality
tests. The dependence on the square ROI and sub ROIs dimensions was
analysed.
Results: Normal distribution resulted to be dependent particularly on the
square ROI dimension, but even on the sub ROIs dimensions, possible bad
pixels and inhomogeneity.
Conclusions: The method described turned out to be a reliable, rapid and
repeatable method for LCD evaluation.
The aluminium step wedge was useful for obtaining LCD in a more mean-
ingful unit than the conventional grey value, allowing comparisons among
different devices. Applications include routine assessment of equipment
performance and optimization studies for digital radiography images.
B.327
MULTICENTRIC EXPERIENCE OF DOSE MONITORING FROM
MAMMOGRAPHY SYSTEMS USING A COMMERCIAL DOSE REGISTRATION
SOFTWARE
D. Zefiro R. Errico S. Pin S. Mazzocchi A. Girell F. Foppiano a S.C.
Fisica Sanitaria, ASL N.5 Spezzino, La Spezia, Italy; b Universita’ degli Studi di
Genova, Genova, Italy; c S.C. Fisica Sanitaria, Azienda Sanitaria di Firenze, Firenze,
Italy
Introduction: The Dose Archive and Communication Systems (DACS) are
useful software tools that allow radiological data collection and patient dose
monitoring.
The aim of this work is to present the preliminary results of the analysis
performed on the data obtained from the mammography systems in-
stalled in two centers. The dose index data have been extracted using a DACS
(Radiation Dose Monitor, Medsquare, France), tested within a project of the
Italian Society of Medical Physics.
Materials and Methods: Data were collected between January and Sep-
tember 2015 from a Hologic Selenia Dimensions (ASL5 Spezzino) and a
Siemens Inspiration (ASL Firenze) mammography. An initial descriptive anal-
ysis was performed on the 5500 patients (23,000 unique projections)
constituting the database.
As a first step, the relationship between tube parameters and breast thick-
ness was studied to detect unexpected trends. Then, for each breast thickness
value, the registered Average Glandular Dose (AGD) was compared to the
achievable levels of the European Guidelines.
Results: The preliminary analysis allowed to exclude non-standard data
from the collected database and to isolate special cases such as patients
with prosthesis which were studied separately.
The AGD as a function of compressed breast thickness showed similar trends
between the two centers: for high compression values the registered AGD
was significantly lower than the achievable levels of the European Guide-
lines, in agreement with the behavior reported in literature and with
measurements in phantom for digital mammography systems.
Conclusion:The data collection allowed to perform an efficient patient dose
monitoring in mammography but it required a preliminary organization
of the database by the operator, which should therefore have the neces-
sary statistics and dosimetric skills. For this reason, the role of physicists
is crucial in this data analysis.
B.328
VIRTUAL MONOENERGETIC IMAGES ON DUAL ENERGY CT: AN IN
PHANTOM COMPARISON OF HU ACCURACY IN TWO DIFFERENT
SCANNERS
S. Zucca *, S. Loi, G.M. Argiolas. Azienda Ospedaliera Brotzu, Cagliari, Italy
Introduction: The purpose of this study was to estimate the accuracy of
material decomposition of two dual energy CT scanners installed in our
Radiology Department.
Material and Method: Images of CATPHAN600 CT phantom were ac-
quired using dual energy scan protocols: protocol GSI-31 (Large
FOV,CTDI = 12.9 mGy) on a GE single source CT scanner with fast kilovoltage-
switching (Discovery 750HD) and protocol 100–140Sn (Large
FOP,CTDI = 12.7 mGy) on a Siemens dual source CT scanner (Definition Flash).
The free insert slot of CPT404 was filled with iodine contrast medium
(10 mg/ml). Virtual monoenergetic images (VMI) from 40 keV to 140 keV
were reconstructed. CPT404 section images were tested for HU accuracy
recording the mean pixel value on a ROI placed on the material insert of
known chemical composition. These values were compared with the the-
oretical CT number calculated using material chemical composition furnished
by manufacturer and NIST linear attenuation data. Image noise was de-
scribed through the standard deviation of a background ROI. Moreover,
standard polyenergetic images (PI) were acquired at 100 kVp with the same
CTDI in order to evaluate CNR ratio between VMI and PI.
Results:For the 40–140 keV range, RMS errors between themeasured and true
CT numbers are 26.8 HU and 13.4 HU for GE and Siemens scanner respective-
ly, reaching a maxim value of 43.9 and 35.8 for high Z material. VMI noise
strongly increases for energies <65 keV. CNR ratio varies with energy and ma-
terial, with a potential enhancement for optimal energy choice. For CM, the
maximum CNR ratio is 0.97 at 40 keV (GE) and 1.14 at 75 keV (Siemens).
Conclusion: The accuracy of CT numbers on VMI varies with material com-
position and energy. For tested protocols, significant differences were found
for high Z material and energy <60 keV. CNR enhancement can be achieved
through optimal energy choice on virtual monoenergetic images taking ad-
vantage of an improved material decomposition capability.
e96 Abstracts/Physica Medica 32 (2016) e71–e96